System for detecting, tracking, and reconstructing signals in spectrally competitive environments

[0001] The present invention relates to a system for three dimensional multiple signal tracking and reconstruction for use in connection with search and rescue, surveillance, storm and severe weather alerting, animal and bird migration, subsurface mapping, anti-terrorism, conventional warfare, etc.

[0002] Analog beamforming and antenna steering using loop, parametric and/or phased arrays, are methods employed for direction finding of acoustic and electromagnetic waves belonging to quasi-continuous signals. In one approach, a directional antenna (such as a parasitic array or loop antenna) is rotated towards the source for maximum reception. The detector may be a single receiver, as in the case of electromagnetic devices, or might be human ears receiving the output of two microphones mounted in the centers of large megaphones. The antenna's ‘beam’ could also be steered at a given frequency of operation by artificially introducing appropriate time delays into the detected signals from different antenna elements. These analog forms of beamforming were employed in World War II (radar and acoustics), for tracking aircraft and for transmitter hunting, and traditionally by radio amateurs in hunting hidden transmitters in practice for emergencies for sport, and for the elimination of undesired signals.

[0003] The time of arrival method has also been applied to locating sources of short duration signals, such wave pulses generated by earthquakes, shocks, blasts from explosions, cell phones, etc. Time of arrival techniques generally seek absolute and arrival time differences between signal arrivals at antenna transducer elements. Such systems typically obtain information from two or more antennas to obtain directions of incoming signals and employ triangulation techniques to estimate points of origins.

[0004] For quasi-continuous signals, certain beamforming and beam steering methods use fixed antenna arrays of transducer elements and digital signal sampling and processing to convert signal conditioned antenna output signals into digitized time series. These methods generally do not require antennas to be rotated, but instead typically make use of second order statistical functions obtained from cross-correlations and cross-spectral densities of time series of the different antenna elements. The second order statistical procedures allow for the determination of preferred signal directions and signal power spectra. Artificial channel time delays using digital processing can be used to numerically steer “antenna beams” and can also incorporate antenna beam modifications for adaptive processing and removal of undesired sources.

[0005] In contrast, the present system is based upon the creation and manipulation of signal vectors providing for the construction of mathematical models of physical wave fields at or in the vicinity of the antenna.

[0034] Generally, the present invention includes a method and apparatus for determining selected information from signals contained in a physical wave field, the properties of the physical wave field being detectable over three spatial dimensions and a temporal dimension, and the method including characterizing the physical wave field by a signal vector providing a mathematical model of the physical wave field, and determining the selected information from the signal vector.

[0035] The necessary equations, theory, and computational tools required for constructing embodiments of the present invention for the different physical wave types are historical and well known and can be established using the following texts (which are incorporated herein by reference): Stratton, Electromagnetic Theory, 1941 Chap 9; Jackson, Classical Electrodynamics, 1962 Chapter 16; Roger G. Newton, Scattering of Waves and Particles, 1966; (Jenkins and White, Fundamentals of Optics, 1957); Thorne Lay & Terry C. Wallace, Modern Global Seismology, 1995; and Oppenheim and Schafer, Digital Signal Processing, 1975; Mathworks, Matlab and Signal Processing Toolbox (User's Guides), 1998; Roman, Theory of Elementary Particles, 1961. As a consequence the details of these theoretical notions, equations, and the standard mathematical tools (Mathworks) are not discussed in detail herein.

[0036] The general natures of the approaches of the present invention are applicable to linear physical wave types or those which can be sufficiently treated as linearized waves. Because of the large number of applications of the invention to different physical media and wave types, for the sake of brevity, illustration of specific applications of the preferred embodiment are limited herein to several demonstrative acoustic and electromagnetic applications, and it is to be understood that the present invention is not to be limited to the applications and embodiments disclosed herein.

[0037] The conceptual basis for the present invention lies in properties of solutions to scalar and vector field wave equations describing physical wave propagation in different physical media and/or in the fields induced by such propagating waves in certain material media. The mathematical models chosen determine the available information. The choice of which physical wave fields are of interest, the propagating wave and/or the waves it induces on surfaces, depends upon what transducers are employed and what is required to attain the desired selected information in the given situation. Under proper circumstances propagating wavefields in the vicinity of a spatial surface are determined by sufficient knowledge of wave fields and their derivatives on that surface over some time interval (see Jackson or Stratton). Kirchoff Surface-Integral representations provide such connections (the treatment of diffraction is such an example) and so can the wave equation itself (with sufficient knowledge of the field along the surface power series expansions can be defined to define the mathematical model in the vicinity of the surface). With these cases, more extensive selected information can be obtained. In other cases, the ability to construct a mathematical model on the surface itself may suffice. Whatever the choice for selected data, data collection of band limited signals through discrete spatial and temporal sampling using properly selected transducers placed along such a surface according to well understood criteria can provide the required information, as the Nyquist sampling theorem shows. The signal vector is either a collected data set containing such required sample information or the equivalent, and therefore contains all the information obtainable from the chosen mathematical model. The selected data, considered already established as obtainable from the field, can be extracted then from signal vectors or the chosen wave fields through the data reduction procedures described below.

[0038] Depending on the nature of the embodiment's antenna, what is being sampled via data collection, the choice of transducers, and the nature of desired selected data, signal vectors can characterize physical wave field fluctuations and/or fluctuations in physical “charge” and “current” density fields on surfaces. The latter can arise in electromagnetism. Thus, the signal vector characterization of an incoming wave field can be indirect. The type of signal vector for the most extensive selected information is that dictated by the procedures for mathematical wave field model determination in the antenna vicinity, as described above, for knowledge of the most complete description of the propagating wave implies knowledge of the most complete information.

[0039] The present invention includes, in certain embodiments, systems and methodologies using antenna arrays of transducer elements, together with electronic signal processing hardware, signal conditioning units, signal processing software, and reporting electronics. The present invention includes a method of deriving signals obtained from properly configured antenna elements (transducers or sensors) which convert acoustic, seismic, fluid dynamic, electromagnetic and/or electric charge or current density wave fluctuations into fluctuating voltages or currents. These signals are amplified and filtered as necessary, and then digitized. The resulting time series of the signals constitute a signal vector providing for the construction of a mathematical representation of the physical wave chosen over the bandpass of interest. The signal vector is cast in representation or form suitable for attainment of desired selected information. The signal vector is then manipulated for desired selected data and/or to obtain the signal vectors characterizing resolved sources of incoming signals. A reversal of the process provides for the construction of a mathematical model for the signals of resolved sources using their individually characterizing signal vectors, from which the information content they carry can be determined through digital signal processing and/or through proper digital to analog (“D/A”) conversion and demodulation.

[0040] Pre-constructed libraries of directional and/or modeling signal vectors and the superposition principle enable the determination of desired signal information from data processing algorithms. In essence, the goal of being able to approximately construct the individual signals of resolved sources of interest as they would be received at the antenna, as if each such resolved source was the only signal present, serves as the most basic design criteria for embodiments of the present invention operating at the highest level. The path to the goal permits the incorporation of new techniques for the extraction of selected information (including that of detection, tracking, and modulation content) and the development of new criteria for when such information is obtainable. However, the system of the present invention can operate without reaching the above-stated goal in the event of partial hardware failure or breakage, when conditions and desired selected data allow a relaxation of system configuration constraints, or when the chosen mathematical model is not necessary for the attainment of certain selected information.

[0041] There are additional fundamental application independent notions underlying the concept of the apparatus or invention. One such notion is that hardware/software/processor configurations and the dimensionality of the vector space in which the signal vectors are defined are coupled to the resolving power specifying the ability to differentiate neighboring signals having identical frequencies. Yet another is that antenna sensors can be configured and sampled such that for a given frequency bandpass their time series constitute sufficient data to allow determination of a signal vector representing the incoming field. Another is that different equivalent representations for a signal vector can be chosen, such as those (but not limited to) of the temporal, frequency, and partial wave domains, the choice being application dependent. Another fundamental notion is that signal vectors of any domain can be linearly and quasi-linearly manipulated for noise reduction and for the attainment of selected information, including but not limited to, that for detection, tracking, and the classification of multiple spectrally competitive sources. The superposition principle applies to signal vectors as well as to the wave fields. Another is that the library element signal vectors of preconstructed libraries can be used to construct the individual signals of resolved sources, and that those of overlapping data set time segments can be merged to provide for determination of the information content of such signals within the operating bandpass. However, certain elements of desired selected information may not require complete construction of the signal vectors of each source making up the sampled signal vector.

[0042] Different applications of the invention to different wave processes require certain changes in hardware components. For example, for acoustic waves in air, microphones could be employed as antenna transducer elements. For acoustic waves in water, hydrophones could be employed, and for seismic waves, the antenna elements would be geophones or vector magnetometers, as would also be the case for tsunamis. For electromagnetic waves, inductors could be employed or capacitors mounted on or in metallic surfaces. For oceanic internal waves, the transducers might be thermistors and/or magnetometers. Correspondingly and depending upon wavelength and frequency, different signal conditioning and data acquisition units could be used. When necessary, a single mixer may be used to simultaneously lower the frequency of the incoming signals on all channels so that the relative phase and amplitude characteristics of the signals of antenna sensor elements are preserved and sampling requirements reduced. The construction of a signal vector requires proper choices for antenna sensor placement and configuration, the frequency bandpass to be sampled, computational word size, and the mechanism for determining the signal vector of a chosen representation jointly typically define an embodiment of the present invention. However given an application, once the signal vector has been determined, the procedures for computation of selected information are nearly independent of the physical type of application. The possible embodiments of the invention include composite systems capable of simultaneously handling all three classes of physical phenomena, as well as a data reporting system possibly using the utility's antenna structure as part of a microwave or RF transmission remote reporting system.

[0043] Scalar products between signal and library vectors and constructs of the latter provide for classes of selected information, depending upon the signal to noise ratio and the system resolving power. The identification and detection process also provides bearing and elevation angles of individual sources. Multiple source tracking and differentiation is based upon the resolving power of the embodiment. An adaptive noise based metric tensor characterizing the environmental noise background redefines scalar products to provide for the adaptive removal of unwanted and interfering sources and a map between apparent and true estimates for directions of incoming signal waves, alleviating a beam maintenance problem found with adaptive beamforming approaches. The linear nature of the physical wave field to signal vector mathematical map enables signal enhancement through signal vector averaging and the attainment of relatively high resolving power through use of high element density arrays and/or a consequent reduction in numerical processing demands.

[0044] The system can incorporate use of digitized contour maps of the environment. Such digitized data permits incoming signal directions to be traced back to their points of origin when such tracebacks intersect physical surfaces characterized in digitized contour maps. In this way, range and source location can be determined from individual antennas. When antennas large enough to measure wavefront curvature are employed, and wavefront curvature is incorporated as a library category, range can also potentially be determined. When neither of these procedures can be utilized, source range can be determined from triangulation techniques using more than one system of the present invention with baselines of proper length, while Doppler observations can be used to estimate trajectory directions and closest points of approach.

[0045] The system can incorporate calibration and self-testing for verification of performance criteria. The self calibration and self-testing capabilities allow for automatically switching to alternate modes of operation, with perhaps some reduction in signal detection and resolving capability, should some signal channels become damaged or disconnected.

[0046] Using subsets of predetermined signal vector libraries, three dimensional systems can be operated in two dimensional modes when incoming signals have their wave propagation vector in a common plane. With a 3D system, this can be done in any plane. Thus, with the system of the present invention, optimum performance may be obtained for waves propagating parallel to the surface of the ground even if the antenna mast is not vertical. The 3D algorithms can be used with 3D subset libraries to operate high transducer density 2D antennas in 2D modes. The present invention can use its antenna structure for housing utility electronics and microwave remote reporting gear and also have it serve as a radio or TV transmitting antenna.

[0047] Given a chosen bandpass, the accuracy of the underlying mathematical model is determined by the accuracy of the sampling, the level of system electronic noise, the level of environmental random noise, and signal processing accuracy, and how well data sampling criteria are met. Environmental considerations, the physical process, the anticipated scenario, and operational constraints may dictate the choice of antenna transducer system to be used. Each embodiment should preferably employ a self consistent configuration, so that the system outputs reliable selected information to a reporter for decision making by a human or other external device.

[0048] The foregoing, as well as other objects of the present invention, will be further apparent from the following detailed description of the preferred embodiment of the invention, when taken together with the accompanying drawings and the description which follows set forth this invention in its preferred embodiment. However, it is contemplated that persons generally familiar with mathematical wave propagation models and with communications and direction finding systems and their theory will be able to apply the novel characteristics of the structures illustrated and described herein in other contexts by modification of certain details. Accordingly, the drawings and description are not to be taken as restrictive on the scope of this invention, but are to be understood as broad and general teachings.

[0049] Referring now to the drawings in detail, wherein like reference characters represent like elements or features throughout the various views, the high resolution, vectorized three dimensional multiple target tracking and signal reconstruction system of the present invention is illustrated in various embodiments.

[0050] The general structure of the physical components of the present invention is described below, as is the basis for the realization of signal vectors, stated with examples. Also described is the basis for configuring embodiments constructed in accordance with the present invention and procedures for obtaining selected information.

[0051] The overall structure of one preferred signal reconstruction and direction finding system, generally 10, constructed in accordance with the present invention is depicted in FIG. 1. Transducers 12 of antennas, such as acoustic antenna AC, electromagnetic antenna E, and/or seismic antenna S, convert physical wave fluctuations at the location of each transducer into electrical fluctuations. Mixer and signal conditioning units, generally SCU, can amplify, filter, and mix the signals of each transducer as requirements dictate. The SCU outputs are then passed to analog-to-digital converters, generally A/D, to digital filters, generally DF, operating consistent with temporal Nyquist sampling criteria (Oppenheim and Schafer). Further signal processing takes place in an array processor AP (or computer CPU, possibly a parallel processor) on time series data output from the A/D converters, and/or digital converters DF, and data reduction take place under software. The array processor AP feeds a reporter, generally RPTR, for transmission of results on selected information in suitable form to humans.

[0052] System 10 is illustrated having acoustic, seismic, and electromagnetic capability. It includes software for an optional calibration signal feed running from the computer processing unit AP through a calibrating signal generator, generally CSG, to transducers 12. The software for the signal calibration feed provides signals for calibration of the system 10 and for verification of system performance within allowed tolerances. The above system in diagram 10 can also be used with an appropriate hydrodynamic sensor (not shown) for use in detecting hydrodynamic wavefields.

[0053] Although FIG. 1 only illustrates applications for acoustic, electrodynamic, and seismic wave types, possible solutions and wave types can include scalar, vector, and tensor fields associated with coordinate systems appropriate for different possible antenna symmetries and physical processes (acoustic, seismic, electromagnetic, and hydrodynamic) and environmental conditions. Each such model is associated and consistent with implementation of the conceptual basis for the embodiment as described in the above Summary as applied to the pertinent type(s) of physical wave propagation, or consistent with an incomplete implementation if the desired selected information allows.

[0054] The propagating physical wave and the detected or transducer sampled physical wave may differ. For example, if the propagating wave is electromagnetic, the transducer detected wave may be that of electric charge and current densities on the surface of a conductor as induced by the propagating wave. Mathematical models used in accordance with the present invention provide the map between sampled data and the chosen physical wave fields to model.

[0055] The present invention includes procedures for obtaining signal vectors. Transducers can be considered as placed upon a surface, real or imaginary. The type of transducers utilized should be consistent with the chosen mathematical model for the chosen physical wave field being sampled. The transducers should preferably be small compared with relevant wavelengths, so that measurements can be associated with specific transducer locations on the transducer surface. If transducer spacing, number, and the temporal sampling rate for wave field fluctuations meet Nyquist sampling criteria for waves on that surface, then the mathematical representation for the wave fluctuations is preferably uniquely determined over that surface for the period data is collected. Furthermore, even if a choice is made for complete mathematical model for the incoming wave field, and transducer type and sampling is consistent with this choice, a mathematical model for the field can also be determined within the vicinity of that surface. Hence, a proper discrete set of samples can define a signal vector, and the Fourier Transforms of these temporal sample sets can also define signal vectors.

[0056] Signal vectors can also be obtained as expansion coefficients of partial wave expansions (PWEs). The latter are defined by separable or nearly separable solutions to wave equations defining the mathematical models for the relevant physical wave fields, those formed through the process of separation of variables. Each partial wave is labeled by the separation constants defining it. For a properly configured system, the signal vectors defined by expansion coefficients of suitably truncated partial wave expansions are equivalent to those described above, and can be related by mathematical transformation. Integral forms for solutions of wave fields, such as those developed by Sommerfeld and Weyl (Stratton) can also be used to define signal vectors.

[0057] The procedures allowing for the construction of a chosen mathematical model for a physical wave field in the neighborhood a surface follow from knowledge of the field and required spatial and temporal derivatives on the surface, and mathematically established uniqueness criteria for the field's mathematical model tell what is required. Nyquist criteria tell when a sufficient number of samples at a sufficient number of spatial locations are taken for a given bandpass and time period. Consequently, with sufficient temporal sampling rates the model for the chosen physical wave field everywhere on the surface can be constructed, and the field on that surface can be numerically differentiated, so that temporal Fourier transforms in time are also known thereon in principle. For certain choices physical wave fields and mathematical models and types of transducers, implied knowledge of spatial and temporal derivatives along the surface allows construction of the wave field model in the neighborhood of the surface, when the latter is needed as selected information. The dispersion relation giving wave number as a function of frequency is known from the wave equation, so the maximum number of wavelengths present along the surface can be determined for each frequency (considering various possible directions for incoming waves). This determines required antenna transducer spacing according to the sample criteria and the angular resolving power as well, which is given by the angular aperture of the shortest possible wavelength along the surface, a result consistent with Lord Rayleigh's criteria for “just” resolution in optics (Jenkins and white).

[0058] Signal vectors can be the ordered set of samples (time series data) collected from the transducers via A/D conversion over some temporal window. Signal vectors also can be the set of Fourier coefficients obtained discrete Fourier transforms of this time series data. Or, they can be the expansion coefficients of a suitably truncated partial wave expansion (associated with a coordinate system for which the wave equation is separable). Other signal vectors can also be obtained, for example, from a discretization of integral solutions to a wave equation. For all such cases, the set of signal vector components will be symbolically represented by {s_α} and the signal vector by s, where α is a collective index.

[0059] The signal vector index α differentiating components of the signal vector may characterize the set {jt_k} of the original samples (j labels the transducer where the sample at time t_k was taken), as such, it characterizes the original sample set time series. It may also characterize the frequencies and transducer label of the Fourier transformed time series data. In the case of partial wave expansion, it characterizes the indices labeling the distinct partial waves, such as the traditional (k,l,m) of the expansion for spherical partial waves, etc. The particular choice of signal vector representation depends upon the selected information being sought and the nature of the environment in which the system is to be used. For example, for sources at long range, the plane wave PWEs provide a useful signal vector representation for spectral components of incoming signals, while spherical wave PWEs are useful for descriptions of same in spherical polar coordinates when output for bearing and elevation angles of sources is desired.

[0060] For a carefully configured embodiment the different signal vector representations are equivalent, a situation occurring when embodiment resolving power, temporal and spatial sampling, and hardware processing are consistent. The Fast Fourier Transform (FFT) relates the spectral domain signal vector to that of the temporal domain. The partial wave functions evaluated at transducer locations define the transformation from one of the former to one of the partial wave representations of a signal vector. In fact the spectral domain signal vector is that of a partial wave expansion in harmonic functions. The aforementioned representations of signal vectors are of course equivalent to some specifiable accuracy, for example, the accuracy with which the truncations of the infinite PWEs are valid, or that to which a low pass filter is effective in meeting Nyquist sampling criteria.

[0061] The mathematical description of a physical wave field may be a multi-component entity, as with seismic and electromagnetic waves. In the latter case, the signal vectors may be obtained through electric and current vector densities on a metallic surface. The Kirchoff type representations and associated uniqueness theorems for each choice of field for which an embodiment is to be constructed provides a guide to what data need be sampled. The same information can be obtained through consideration of what is required to create a power series expansion for the mathematical model using the wave equation. Use of the wave equation and proper temporal and spatial application of the sample theorem provide the latter. For a complete specification of a mathematical model, depending on desired selected data, more than one type of transducer may be required (for example, through use of both velocity and pressure sampled microphones in acoustics, capacitors and/or inductors in electromagnetics). Seismic and hydrodynamic waves treatable with linear mathematical models exhibit analogous properties.

[0062] The choice for a signal vector representation should preferably be made on the basis of numerical efficiency in obtaining the selected information and upon the dictates of the appropriate uniqueness theorems. The most efficient are typically those for which the coordinate system reflects the symmetry of the antenna system and environment and for which the wave equation allows for the separation of variables. Example antennae of such “efficient” systems are illustrated in FIGS. 2–6.

[0063] If the antenna presence significantly disturbs the propagation of the physical wave, then the model solution to the wave equation for waves of each incoming direction should preferably include the scattered wave as well. The signal vector should preferably as much as possible characterize the total wave field of both the incoming and scattered wave, as required by the selected information sought, for accurate results.

[0064] The antenna is the physical device containing transducers through which the fluctuations of the physical wave or induced “charge densities” are converted into electrical fluctuations. The specific form of the conversion process is assumed to be understood and linear, so that the superposition principle applying to composite waves of multiple sources can be tracked in the electrical fluctuations. Transducers define a map between physical wave field fluctuations Ψ and voltage or current fluctuations σ.

[0065] After amplification, A/D conversion, signal conditioning, and digital bandpass filtering, any decimation, and possible Fourier transform (via FFTs) of properly windowed time series from each transducer channel, various representations of the signal vector are obtained from linear equations of the form:

[0066] ${\sigma }_A=\sum _{\alpha }{s}_{\alpha }{K_A^{\alpha }}_{}j=1,\dots ,N_s.$
In this signal vector equation, σ_A may be proportional to the A^th component of the wave field, as might be in the acoustic application with use of microphones, or to surface charge or current density components, as in the electromagnetic case (related to the normal components of the electric field and the tangential components of the magnetic field at a conducting surface, respectively). The coefficient matrix elements K^α_A may coincide with the identity matrix if the signal vector represents the samples themselves, or may be the partial wave functions evaluated at transducer locations in the case of PWE representations of the signal vector.

[0067] Antennae with microphone transducers may be used for sensing pressure fluctuations (FIGS. 2A, 2B, and 6), with capacitive transducers for sensing surface charge densities (FIGS. 3A, 3B, 5A, and 5B), and with inductive transducers for sensing surface current densities (FIGS. 4A, 4B, 4C, 5A, and 5B). For the acoustic case, depending on the level of mathematical model the signal vector should define, the transducers might include both pressure and velocity type microphones. In the electromagnetic case, antennas may combine capacitive and inductive sampling using capacitive transducers 54 and inductive transducers 56. In the latter case when the two types of transducers 54, 56 are taken together, sampling would allow the polarization of incoming waves to be addressed as selected information and also allow field reconstruction given proper scattering models for the antenna structure. For the capacitive electromagnetic antenna of FIG. 3A, only the surface charge density sampled by transducers 34 of FIG. 3B, related to the normal component of the electric field at the surface, is required for determination of the directions of incoming waves, as an analysis of electromagnetic vector partial wave solutions can show. However, if signal reconstruction and or polarization information is required, additional sampling may be needed, as with using the inductive transducers of 56 of FIG. 5D, or a combination of one inductive transducer 56 and one capacitive transducer 54 set therein. Note that any of the antenna shapes may be used for acoustic or electromagnetic sampling in FIGS. 3A, 4A and 4B, and in FIGS. 5A and 5B and FIG. 6.

[0068] Other possible transducer types not specifically illustrated in the Figures above include hydrophones (sampling acoustic pressure fluctuations in fluids), geophones (sampling displacement velocities in seismic media), total field magnetometers (sampling magnetic fluctuations parallel to the earth's magnetic field), vector magnetometers (sampling fluctuations in the magnetic field components), and thermistors (embodiments sampling temperature fluctuations of oceanic internal waves). Magnetometers can also indirectly sample magnetic fluctuations in oceanic tidal, internal, tsunamic, and surface waves. The present invention is not limited to these types, it can be used with any transducer set sampling entities associated with waves mathematically modeled by linear equations. The transducers should preferably, however, be relatively small in comparison to the shortest wavelength envisioned in the process, so that samples may be associated with points on the closed transducer surfaces. Antennas with open surfaces may also be closed using suitable spatial windowing techniques analogous to those used in temporal sampling.

[0069] Depending upon whether the dimensionality of the signal vector coincides with the number of transducers utilized (generating utilized data chains) there exists a number of numerical methods that can be utilized for the inversion of the signal vector equation (Math works). Fast Fourier transforms, and matrix inversion and least square techniques have been employed in embodiments to invert the truncated partial wave expansions of different coordinate systems and physical waves. However, if the inversion process is not one-to-one, then there can be ambiguity in any subsequent signal reconstruction, with a potential for subsequent loss of information.

[0070] The system configuration sets the resolving power, the size of the antenna, the number of required transducers, and the dimensionality of the signal vector. A minimal configuration provides for the minimum number of transducers allowing for the definition of a signal vector (the number of its effective independent components thereof), as required by the Nyquist criteria, and as discussed above this also establishes the resolving power. The latter, according to Rayleigh's criteria for “Just resolved” follows when the maxima of a surface wave of one source is at the first null of that of another, a criteria developed for consideration in optics when two sources are present. The criteria coincides with the minimal condition set by the Nyquist sampling criteria, that there must be at least one transducer for each half wavelength along the surface. Here, we define the embodiment's resolving power, generally RP, is defined as the minimum number of resolvable sources on any closed circle centered on the antenna.

[0071] Assuming that any incoming direction is possible, the maximum projection of the wavevector k corresponds to the shortest wavelength at any point on the surface. Then according to Rayleigh's criteria for “just resolved”, the resolving power is: RP=2kR. But, this is just twice the number of wavelengths around the imaginary circle. Accordingly, the Nyquist criteria also establishes Ns=2kR as the minimum number of required transducers to be uniformly aligned on such a circle, or the number needed per unit length if the transducer surface is not made of circles. Ns also coincides with the utilized dimensionality of the signal vector required for the highest frequency considered for a given direction, also according to the sample theorem. For a spherical surface, the number of required transducers and signal vector dimensionality is then (2kR)² if the resolving power is independent of loop orientation. If not every incoming direction is possible, the uniformity requirement for resolving power can be relaxed.

[0072] The considerations on resolving power and number of required sensors refer to the σ_A of the signal vector equation for a scalar field. For vector and tensor fields, the result may differ. For electromagnetic waves, if the embodiment is needed to carry out a complete signal reconstruction, the number of required sensors is twice that above, as there are two polarizations. The result is then that of a sum over the two polarizations, which would be the same as that for a scalar field. Note that for propagating electromagnetic waves not all three of the components of the electric and magnetic fields are independent.

[0073] Data reduction consists of determining the selected information desired. It involves detection, direction finding, source classification through spectral decomposition, and signal reconstruction. It makes use of the superposition principle, pre-constructed libraries of mathematical models of possible sources, and certain geometrical entities. The latter include scalar products, signal surfaces, energy surfaces, detection surfaces, and adaptive directional maps.

[0074] The superposition principle regards the signal vectors s of received composite signals as the sum of those of possible sources s_q and noise. This applies to any form of the signal vector, whether in the sample, frequency, partial wave, or integral form domains.

[0075] The most fundamental signal vector library is the directional library (DL), which is created and developed from signal vectors of possible directions and frequencies within the bandpass of interest. Let s_k denote such a library element, the wavevector k corresponding to the incoming wave spherical polar angle direction (θ_k, φ_k). Such a library can form the set of signal vectors of point sources at very long range, so that the modeled incoming waves are plane when they pass over the embodiment's antenna. The directional library can be constructed using analytical input (as obtained from the partial wave expansion of plane waves) and/or numerical models.

[0076] A signal surface is the surface defined by scalar products s(θ_k,φ_k)=(s,s_k) between the signal vector of the sampled wave and library elements with the incoming signal vector for all wave directions k. The “radial” coordinate for each point on the surface corresponding to a DL element is the value of the scalar product corresponding to that direction of k. The metric for this scalar product is that which is appropriate for the signal vector representation. For the sample and frequency domains representations, the metric is diagonal, and each such diagonal element has unit value.

[0077] To each signal surface there corresponds an “energy surface” denoted by s²=|s(θ_k,φ_k)|². Note that each library element can have its own “signal surface”, defined by scalar products of the form (s_k,s_k′) for all possible wavevector directions k′. Each such element also therefore has its own “energy surface” s_k².

[0078] The “adaptive signal surface” is obtained by weighting the value of the signal surface for the direction of k by the inverse of the average of the noise background signal energy surface for the same direction, the average being taken over a number of temporal windows or realizations.

[0079] The adaptive signal surface is a distorted signal surface so that the direction associated with maxima in the signal surface's energy surface may not define the correct direction to a source. The “noise adaptive signal surface” reduces the contributions of noise sources to the signal surface for directions for which the average noise background is large relative to other directions. Such noise sources may include known sources whose presence is also known and understood but not of interest. Thus, the adaptive signal surface procedure is particularly useful for reducing the effect of background interference from known sources not of tracking interest. In such instances where distortion exists, an adaptive map should preferably be used to obtain the true directions for vectors associated with sources. This map will be defined below.

[0080] There are two classes of detection criteria. One is energy detection, appropriate for the situation when only a single source of interest is known to be present, or the source is already known to be well resolved, and the signal-to-noise ratio is good. The second is that of directional detection, useful when the presence of multiple sources are suspected or anticipated, the sources may not be all resolved, and/or when the signal to noise ratio is low.

[0081] There are also different levels to be reached in declaring a detection. The first is that of a probable detection. A probable direction is preferably automatically assigned to every detection. The notions for probable detection and probable direction apply regardless of whether the signal surface utilized is adaptive or not.

[0082] Energy detection is a threshold detection based upon the energy surface s²(θ,φ;ω. If the maximum value for the latter as a function of (θ,φ) exceeds that of the averaged noise background “energy” by some preassigned number, then a probable detection is declared. The preassigned number can be based upon a false alarm rate following from a model for expected detections from some noise model.

[0083] Directional detection utilizes signal and library direction vector ray direction properties. With directional detection, signal vectors (including directional library vectors) are normalized such that their signal surfaces have value unity for the direction for which the corresponding energy surface has a maximum value. A probable detection is declared if the absolute value of the difference surface resulting from the difference between the signal surface of the incoming wave field and that of a directional vector integrated over all directions to points on the difference surface is below some threshold. Thus, the detection is based upon a model for the shape of the signal vector of a given source, instead of upon the signal level relative to some background and threshold.

[0084] For energy detection, the directions of the probable sources are those for which the signal energy surface has maximum values exceeding the threshold. For directional detection, the assigned probable direction is that for which the total “energy” of the difference surface corresponding to a particular direction vector has its minimum value. Probable directions associated with the adaptive signal surface are preferably assigned only after the adaptive map is employed.

[0085] The adaptive map transfers the directions assigned to sources of signal surfaces constructed using the adaptive scalar product to true directions. The directional criteria can result in the signal surface and adaptive signal surface assigning different directions to the same incoming signal detected if the time averaged noise background “energy surface” is not isotropic. The map can be computed off line by noting the adaptive signal surface directions assigned to the modeled incoming plane wavevector directions used to define the directional library vectors using the two metrics, and constructing a table lookup for this correspondence.

[0086] For “energy detection,” the direction assigned to the probable detected source is that of the maximum of the signal surface, the value of (θ,φ) belonging to the vector k of the s_k defining the signal surface maxima. This assignment can be made to other maxima provided those are at least just resolved, according to the criteria discussed above.

[0087] For directional detection, the assigned direction is that belonging to the s_k producing a minimum value for the signal vectors satisfying the criteria for probable directional detections, provided the minimum is at least just resolved from other minima.

[0088] If multiple sources are encountered, it is preferable to ascertain whether they are resolved, and if so, incorporate the resolving capability into the detection criteria. A local resolving power as a function of direction is ordinarily readily determined.

[0089] The local resolving power is found by observing the functional dependence of the scalar product (s_k,s_k′′) and determining the angular separation γ_RP between the directions for which this scalar product has its maximum value and its first minimum away from this central maxima. The local resolving power is then defined by: LRP=2π/γ_RP, where γ_RP is the angle just described, hereafter referred to as the resolving angle. Conceptually, LRP is typically an estimate of the number of equally spaced sources that could be resolved along a circle. For a two dimensional transducer surface, the local resolving power can be defined as the geometric mean of the local resolving powers for two orthogonal directions.

[0090] The detection rate surface is the surface defined by the rate at which detections occur per unit solid angle, as defined by probable directions, the rate determined over a number of data collection windows. Noise interference effects generate detections even when sources of interest may not be present. The detection surface for background noise alone is denoted by DR_N, and in the presence of signals of interest, by DR_S+N. The detection rate surfaces of probable detections can be defined after a sufficient number of realizations has taken place to yield a sufficiently smooth function of the spherical and polar angles.

[0091] Detection measure surfaces provide for the declaration of a detected source. One such measure could be the surface defined by the noise averaged detection ratio (DR_S+N−DR_N)/(DR_S+N+DR_N). This definition requires that a number of realizations be employed to compute the average, and that the ratio be considered a function of angle.

[0092] Declaration of a detected source is made if the measure function of direction has maxima exceeding a detection threshold (DT); if the shape of the maxima is judged consistent with noise properties; and if the angular distance between such maxima is consistent with the embodiment's resolving power. Thus, the concept of resolving power plays a role in the declaration of multiple source detections.

[0093] In some instances, such as when the “energy” (in the sense discussed above) is sufficient and only one source is known to be present, a detected physical source can be declared directly. The detection measure-based procedure can potentially permit detection at the background level less the antenna array gain, even when multiple sources are present.

[0094] Signal averaged detection follows from a different process. It is particularly applicable to impulse signals such as those of projectile shock and blast waves, and in estimating the directions of continuous signals with identifiable spectra, such as those associated with harmonic chains of piston engines.

[0095] For impulse waves belonging to short duration single occurrences, the detection rate surfaces and temporal averaging processes described above may not be particularly useful. But such impulse signals are broadband in nature, and the signal surfaces can be averaged over frequency. The noise fluctuations of the different frequencies have different signal surface oscillation wavelengths, which average out in the summation process, while the surface functions at their maxima are normalized to unity and add coherently within the resolving angle. In this type of average, the normalized signal surfaces (normalized by their maxima as described above) are summed and divided by the total number of spectral components involved. Either the energy detection or directional detection (appropriately modified) process can be applied to the result, the averaging process also can be limited to frequencies having probable detections. This technique has been found to yield excellent results in detecting gunfire and shooter directions, as well as those shock waves associated with the projectile. It will also be useful in the tracking of turbine engines, which also have broad band spectra. The detection rate surface process could also be applied to the spectrally averaged signal surface if the events are repetitive.

[0096] In the case of piston engine harmonics, the signal averaging summation is carried out only for elements of the harmonic chain. This is a useful technique since the Doppler shift of elements of the harmonic chain yields another harmonic chain (with different spacing for chain elements) and direction finding, and detection of individual elements of the chain may not yield accurate results if the signal levels are too low. Again, the signal surfaces involved must be normalized to unity in the direction they have their maxima.

[0097] For sources at very long range, or for where they are at rest, or when their directions are changing little with time, the properly normalized signal surfaces could be averaged over data collection temporal windows. Also, an average over both temporal windows and frequency could be made, depending upon the anticipated nature of the sources and the environment. Again, detection rate surfaces based detection criteria are applicable.

[0098] If the noise based adaptive signal surface is used in signal averaged detection, it should preferably first be mapped into a standard signal surface using the adaptive map for each frequency before signal averaging takes place, if the noise background changes with time. This is because the signal surfaces of different frequencies may have differing distortions.

[0099] The choice of which signal detection process is used depends upon the anticipated signal to noise ratio, whether or not the signal has broad spectral content or a harmonic chain, and the duration of the signal. The signal surface averaging processes are particularly useful because the effects of random sources and fluctuations generated by random processes affecting paths tend to average to zero, in sharp contrast with averages involving “energy” type functions. Source directions come with the detection processes.

[0100] For noise averaged detection, the assigned direction are the maxima of the detection rate measure surface meeting the detection measure threshold, provided the maxima is at least just resolved from other maxima as per the discussion on noise averaged detection above.

[0101] For signal averaged detection, the direction assigned is that of the maxima of the signal averaged signal surface exceeding some threshold, as with directional detection. There can be more than one such direction if there is more than one maxima exceeding the detection threshold, provided the maxima involved are just resolved.

[0102] For noise averaged signal averaged detection, the assigned direction is that of the maxima in the measure surface. There can be more than one such assigned direction if there are other maxima meeting the detection threshold criteria, provided those maxima are at least just resolved.

[0103] Source spectral content is established once the s_q have been determined for a resolved source. Given the s_q weighted of their contributions as defined by the scalar products defining signal surfaces, the s_q can be transformed to the frequency domain. The frequency domain s_q coincides with the Fourier coefficients of the incoming time series data of the q^th source, and its spectra is defined. This amounts to executing the signal vector equation with the σ_A as the result and the weighted s_q as input. The resulting |σ_A|² then provides the power spectra. The source level of a particular source q can be estimated from a propagation model for the given environment, taking into account attenuation and geometrical spreading.

[0104] Signal reconstruction and signal information content are carried out by transforming from the derived representation for the weighted frequency domain s_q, and then using the inverse Fourier transform (discrete) to go to the equivalent sample representation of the s_q, This must be carried out for a series of overlapping temporal data collection windows, where in the set of field values τ indexes the sample set or sample window of the j^th transducer at the listed coordinates for each time t_k of the window. If this is done for a series of overlapping temporal windows, windowing functions W(t_k−τ) can be employed to construct from these sets a smooth time series over an extended time period. This process permits determination of information content or reconstruction of received signals using digital signal generators carrying out digital to analog conversion, provided the embodiment bandwidth is sufficient for the resolved signal and all significant spectral components of the resolved signal are included.

[0105] If the proper set of transducers are employed in conjunction with the wave equation or surface integral expressions of the Khirkoff type, as per associated uniqueness theorem criteria, a mathematical representation of the field in the neighborhood (or vicinity) of the antenna can be carried out. In this case selected, information can be obtained after the mathematical model of the wave field has been constructed from obtained signal vectors. This can be done using the well known Neother's theorem techniques applied to mathematical model Lagrangians to obtain energy and momentum vector densities (Roman) explicitly carrying directional and frequency content.

[0106] Note that it is not necessary to temporally sample at twice the highest transmitted frequency to get desired modulation information as selected information. The processed signal need only be heterodyned and sampled at a rate of at least twice the highest modulation frequency, since the signal vector equation can be scaled so that all the transducer signals are sampled relative to one.

[0107] Several example embodiments and deployments are illustrated in FIGS. 2–14, but it is to be understood, however, that the present invention is not limited to the examples and embodiments herein disclosed.

[0108] For each antenna illustrated, transducer spacing depends upon the desired resolving power, which should be compatible with the number of transducers allowed. The resolving power need not be the same for all directions.

[0109] For a spherical antenna having spherically symmetric resolving power, the Nyquist criteria require N_s=(2kR)²=(LRP)². The partial waves defining signal vector components can be the spherical harmonic-based set of consisting of the j_l(k,R)Y_lm(θ,φ), where j, denotes the Bessel function, Y_lm the spherical harmonics, k is the wavenumber, and R is the spherical radius. The signal vector index a consists of the indices k, l and m for each frequency. The series is truncated for a maximum valuel_max=(LRP−1). The Fourier and spherical harmonic representations for the signal vector are equivalent. Near equal angular spacing of transducer elements can be achieved by having element planes with near equal polar angle spacing containing groupings of (2l+1) transducers, equally spaced within.

[0110] The spherical antenna, generally 22, of FIG. 2A is of the type which is appropriate for acoustic microphone transducers obeying these placement rules. The insert in FIGS. 2A and 2B indicates generally how the electronics, such as SCU, A/D, DF, AP, and RPTR, can be mounted within the mast. With thin structural elements and transducer dimensions relative to a wavelength, it has not yet been found necessary to include the influence of the element scattered waves on the transducers. Other coplanar grouping of transducers could involve √{square root over (N_s)} equally spaced transducers in √{square root over (N_s)} planes, with equal polar angle spacing, but the result does not give near uniform resolving power for all directions.

[0111] For the cylindrical antenna 30 of FIG. 2B, the partial waves consist of products of ordinary Bessel functions of the radial coordinate, the exponential function along the axis, and harmonic functions for the bearing coordinate, if a cylindrical partial wave representation is used to define a signal vector. This antenna would require windowing in the axial coordinate in order to simulate the condition that a closed surface contains the transducers, since practical systems cannot be infinite in length. A maximum spacing of λ(2 sin φ) between transducers is preferably required along the cylinder's axis, where λ is the wavelength at the highest frequency and φ is the highest anticipated angle of elevation over the plane of a ring. For a bearing resolving power of LRP=2ka, the number of microphones along equally spaced rings should be equal to or larger than 2ka, where k is the wave number of the highest frequency to be studied and a is the cylindrical radius, and consistent with the above, 2ka such rings 26 should be used at a minimum. Note that the first and last ring need not be at the “ends” of the cylinder where the windowing function vanishes. With this type of configuration, generally there is an equivalence between the Fourier and cylindrical partial wave representations of the signal vector.

[0112] FIG. 3A illustrates a spherical electromagnetic antenna, generally 32, with capacitors, generally 34, whose construction is illustrated in FIG. 3B, having an insulator 38 and another plate 39. Capacitors 34 can have one capacitive transducer plate being the surface 36 of the highly conducting metallic spherical shell. The same transducer placement considerations apply as discussed for FIG. 2A. The capacitor's impedance over the bandpass should be consistent with electronics and sampling requirements on time constants, and the capacitors should preferably be small in comparison to a wavelength. At cell phone frequencies, such antennas are capable of high density transducer placement and resolving power and portable for placement with individuals or on vehicles.

[0113] FIG. 4A is an electromagnetic cylindrical ring antenna 40, the electromagnetic counterpart of the antenna of FIG. 2B. The transducers are inductors 42 as illustrated in FIG. 4C and are equally spaced on rings 44 (FIG. 4B). The inductors' 42 are insulated windings over the highly metallic ring surface 46, as illustrated in FIG. 4C. Inductor axial and angular placement is the same as the transducer placement as per the discussion on FIG. 2B. The inductors should preferably also be small in comparison with utilized wavelengths and their electronic impedance should be consistent with electronic impedance requirements over the bandpass. At cell phone and other high frequencies, such antennae can have high density, high resolution, and be mounted as “whip” antennas above vehicles.

[0114] Use of high density antennae also permits adaptive reduction of vehicle noise using the adaptive noise metric, as discussed above, or noise reduction using Gram-Schmidt orthogonalization applied to the signal vectors. Such antenna when mounted as tall structures could potentially support highly accurate systems for measuring elevation angle as well as the bearings of low elevation angle incident radiation. Note that this type of structure, with microphones or hydrophones replacing inductors 42, also can be used as acoustic systems in air and water. In the latter case, when embedded in the ground with hydrophones, as in a well, they can be used to map noises in underground facilities, when proper modeling and boundary conditions are employed. If the overall antennae dimensions of FIGS. 4A and 4B are small compared to utilized wavelengths, antenna scattering need not be accounted for in acoustics, but must be accounted for in the electromagnetic case when metallic surfaces are utilized. Some scattering models exist in the literature reference above.

[0115] FIGS. 5A and 5B depict an ellipsoidal, or prolate spheroidal, antenna, generally 50, and a cylindrical shell antenna 52, respectively, associated with separable coordinate systems and also capable of supporting PWE based signal vectors. They can be used with hydrophones for water applications, with microphones for acoustic applications in air, with capacitive transducers 54 as in FIG. 5C, or with inductive transducers 56, as shown in FIG. 5D. In the case of inductors, with two “polarizations” for the inductive transducers as 55 in FIG. 5C, such an antenna can determine the polarization of incoming waves and be used for electromagnetic signal reconstruction, as discussed above.

[0116] With the antennae 50, 52, and the spherical shell antenna 32 of FIG. 3A, the modeling library signal vectors should include shell scattering effects for both electromagnetic and acoustic waves. If desired, the electromagnetic transducers could be replaced with acoustic microphones.

[0117] A theoretical analysis shows that the signal vectors of incoming electromagnetic waves at long range are characterized by the expansion coefficients of vector spherical harmonics, so that capacitive transducers suffice if the desired selected information is confined to tracking information. The sampled charge and current densities are those of signals including the possible polarizations. If the selected information also requires wave polarization information or individual signal source reconstruction, then preferably, the complete set of transducers 54, 56 depicted in the FIGS. 5C and 5D can be employed. The spherical antenna would use spherical Bessel functions and vector harmonics for its vector partial wave expansion, while the ellipsoidal antenna 50 could use “corresponding” prolate spherical coordinates and functions. If the minor axis of the ellipse is chosen to be the one rotated about, then the oblate spheroidal coordinates and the corresponding PWE would be used. The complete number of required transducers 54, 56 is double the number required in the acoustic case for the spherical antenna of FIG. 2A for the same wavelength, a consequence of the two polarizations. In any case, transducer spacing should preferably be determined by the maximum projection of the wavevector for the highest frequency onto the surface containing the transducers 54, and the spacing should preferably be less than or equal to half the length of the shortest corresponding projected wavelength.

[0118] Supporting electronics can be built into the antenna masts and/or spheres and/or ellipsoids as desired. Such antennas are very practical for electromagnetic frequencies running from short wave and VHF to cell phone frequencies. The spheroidal antenna 50 would use vector wave solutions as well, but would employ cylindrical vector harmonics as in Stratton (referenced above), the vector harmonics and the vector wave solutions involving derivatives of the aforementioned functions.

[0119] With the capacitive antenna, the transducers 54 and 34 sample surface charge densities on the surface, while with the inductive antenna surface, surface current densities are sampled. Complete scattering solutions for incident plane waves with the proper polarizations can yield the necessary connection between incoming fields of sources at long range and induced surface charge and current densities. The connection is defined by well know boundary conditions relating the surface charge and current densities to the normal components of the electric field and the tangential components of the magnetic field. For the electromagnetic antennas mounted on the metallic surfaces, the capacitive and inductive transducers should also have dimensions and structure compatible with the impedances required by the electronics and sampling criteria, i.e., time constants associated with the sampling.

[0120] FIG. 6 illustrates an antenna 60 suitable for treating coupled incident and earth surface reflected acoustic waves in regions where the earth and/or mounted surface is flat on a scale large in comparison to a wavelength. Again, use of thin structural elements 62 can make it unnecessary to determine the waves scattered by the antenna. The coupling of the incident and surface scattered waves introduces relationships between components of the signal vector, reducing the dimensionality of the processing space and decreasing computational demand.

[0121] FIG. 7 illustrates an application of the present invention, wherein where small high density antennae, generally 70, can be deployed on a vehicle 72. Shock, blast, and electromagnetic waves typically have short wavelengths in comparison to vehicle dimensions, and higher frequencies can be utilized. For a given natural local resolving power LRP, the antenna size can be reduced. A traceback program tracking back along ray propagation paths can be used to locate resolved sources of gunfire 74 and shock waves, provided that contour map data is also supplied for the environment. Selected information can also be obtained for cell phones 76 using electromagnetic antenna 78.

[0122] FIG. 8 illustrates deployments of the present invention where high density, high resolution antennae, generally 80, can be used with signal processing to help null out certain noise, such as noise from a vehicle 82, as in removing vehicle-generated noise coming from the vehicle 82 itself. Use of the noise metric and the associated mapping technique would allow tracking of sources of interest, particularly shock and blast waves, at wavelengths which are short in comparison with vehicle size, even with significant vehicle noise. Small spherical and cylindrical high density arrays can be useful on relatively noisy vehicles. Traceback programs can also be employed for the location of sources of projectile 84 firings, given contour map or triangulation data. For lower acoustic frequencies, microphones and electromagnetic transducers can be placed along the body of the vehicle. In this case the boundary conditions associated with scattering off the vehicle surface should preferably be accounted for in the wave propagation model used to construct the directional libraries, as is done with the electromagnetic antennas shown in FIGS. 3A through 5B. The noise metric and associated adaptive mapping techniques can be employed to obtain true directions.

[0123] Other useful antenna system deployments can be made in accordance with the present invention and can include severe storm 90 tracking with building-mounted spherical acoustic and electromagnetic antennas, collectively generally 92, as depicted in FIG. 9. Combined acoustic and electromagnetic systems can provide range data for near simultaneously resolved thunder and lightning 94 bursts, while visitors can be tracked and monitored with acoustic arrays and traceback programs.

[0124] FIG. 10A depicts a towed, or, surface-suspended antenna array 100, and FIG. 10B depicts a bottom-mounted antenna array 102, with both being constructed in accordance with the present invention. Such a system can employ hydrophones, thermistors, or magnetometers for detecting ship engines, internal waves of swimmer delivery vehicles, tsunamis, shipping traffic, etc., depending upon antenna and wave scales. The towed antenna 100 for side acoustic scan sonar applications of the types illustrated in FIGS. 5A and 5B, with electromagnetic transducers therein having been replaced by acoustic transducers, is illustrated in FIG. 10B. Such side scan operations, using active pulses 104, can provide 3D views of submerged objects, such as mine 106 on the bottom or a tethered mine 108.

[0125] A spherically symmetric and/or “tree” antenna 110 based system, constructed in accordance with the present invention, for a harbor security monitoring is depicted in FIG. 11A, along with a cylindrical monitoring array 112 as shown in FIG. 11B. The cylindrical antenna 112 may serve as an acoustic or hydrodynamic wave antenna (surface, internal, etc.), and could monitor for self-propelled and/or swimmer vehicles 114 as well.

[0126] Mountain deployment of an acoustic spherical antenna 120 is depicted in FIG. 12, and such a deployment could be useful for remote traffic 122 monitoring for situations where the terrain is not planar in the neighborhood of the antenna and where the antenna need not be mounted with a particular orientation relative to the earth's surface. A traceback program tracking back propagation paths can be used to locate sources of signals, such as from a sniper 124, emanating from the mountain sides and/or other structures.

[0127] The wavelengths of interest and the nature of the propagating medium place restrictions on seismic wave antenna systems. Attenuation in the earth for interior waves is higher at higher frequencies, so low frequencies are generally more productive when making application to waves propagating through the earth. Embodiments such as those depicted in FIG. 13 might be employed in seismic exploration and/or facility monitoring. They can involve antennas, generally 130, with elements burrowed into the earth's surface 134 with waves being generated using an active mode, such as a bomb 136, as in FIG. 13, or planar antennas 140, 142, as depicted in FIG. 14, for the detection of seismic surface waves, as detected by antenna 142, and for acoustic waves, as detected by antenna 140. Geophones and/or magnetometers (not shown) can be used as transducers for measuring fluctuations in seismic media or magnetic induced charges by metallic vehicles, respectively.

[0128] In simple seismic models, the wave equation modeling the physical wave can be characterized by elastic moduli, and the waves are then best described by the P and S wave Hemholts decomposition (Lay and Wallace). In this case, both longitudinal and vector harmonics are needed. Use of small bore holes 138 relative to wave dimensions of interest permits ignoring scattering by the illustrated seismic antenna structure, so that, in one embodiment, only scattering of waves at the earth's surface should preferably then be accounted for. A spherical antenna with an active explosive source 136 would permit mapping of tunnel structures and underground facilities, and multiple antennae would permit passive monitoring and mapping of the facility.

[0129] FIG. 14 illustrates a combined acoustic, generally 140, and a seismic antenna array, generally 142, constructed in accordance with the present invention for measuring waves propagating 1 (in the neighborhood of the array) along the earth's surface.

[0130] The antennae described above (with the exception of that of FIG. 13) can typically be readily constructed and can provide for a minimal number of floating point operation, since the constant value of one of the coordinates of the coordinate system describes each transducer surface. Other antennae are possible, though, and could use a random placement of transducers, as might be done if the antenna must be placed remotely, such as, for example, in the scenario illustrated in FIG. 13 if providing small bore holes were not possible. If transducers are randomly placed (such as by air drop, use of projectiles, missiles, or the like), software, and knowledge of their random positions through us of the global positioning system (GPS), could potentially allow a subset of the transducers to be sampled via radio link, or some other communications method, for the purpose of defining signal vectors. A subset of the transducers consistent with the Nyquist sampling criteria could then be selected. In such an embodiment of the present invention, the thus formed randomly-shaped, or near arbitrarily-shaped, antennae could also be used, but there would likely be a higher demand for signal processing power for certain types of selected information, such as with local wave field reconstruction. Calibration of such antennae could be done in various ways, and could include the detonation of one or more explosive devices at a predetermined location(s) within the vicinity of the transducers and reviewing the outputs of the respective transducers in response thereto.

[0131] Note that with configurations of antenna constructed in accordance with the present invention and meeting minimal Nyquist sampling requirements, such antennas can also be used in the sample and frequency domains without reference to partial wave expansions, particularly if only detection and tracking information is required.

[0132] The present invention also includes a self calibrating system, generally 150, shown in FIG. 15, which may include the self-calibrating system contained in FIG. 1. Calibration of the system 150 corrects for deviations from desired component tolerances and verifies that the system is operating within prescribed tolerances. The linear nature of the processing provides employment of simple techniques. Signal output from D/A converters (not shown) driven by time series from an array processing unit AP drives a calibrating signal generator CSG inputting a common signal to transducer elements. The complex amplification for each channel, generally A, for this signal can be determined, and deviations from a standard (which can be one of the antenna transducers or a reference signal) can be used as a basis for numerical correction of inputs. This insures that the system channels effectively are identical on each frequency over the bandpass. This procedure can be also be used to check that voltage outputs, generally V, (and/or current) of each channel are within system tolerances. The process can be carried out automatically each time the system is initiated or as desired by a controlling operator. The system operator can also execute a manual calibration and verification of system performance. Isolating impedance or isolation devices, generally Z, assist in insuring that there is little “cross-talk” between input calibration signals. Required phase and amplitude corrections for each frequency can be incorporated into the processing by modifying the matrix K from which the signal vector is obtained in the signal vector equation.

[0133] Operator calibration can be carried out if the operator is uncertain about the functioning of internal calibrator. FIG. 16 is a single channel microphone signal driver, generally 160, for providing an acoustic pressure wave reference to a standard signal. The system can be calibrated one microphone channel at a time, and includes a microphone 162, an acoustically isolating microphone receptacle 164 having a pressure wave channel 166, and a driving earphone/speaker 168.

[0134] FIG. 17 is a schematic representation of calibration device, generally 170, which can feed identical simultaneous signals to each channel. It can be used to determine simultaneously the relative complex amplification of each acoustic channel, again circumventing the need for electronically matched signal components, and includes a microphone 172, a driving earphone/speaker 174, a symmetric acoustic wave channelizer 176, and an acoustically isolating microphone receptacle 178. In the calibration devices shown in FIGS. 16 and 17, the acoustic signal feed dimensions should preferably be small in comparison to a wavelength.

[0135] FIG. 18 illustrates a composite seismic and acoustic system, generally 180. Microphones 182 and/or geophones 184 can be simultaneously sampled as with two of the subsystems of the composite system 10 of FIG. 1. A computer 186 is preferably used as an array processor. This embodiment is useful as a multimode detection and tracking device and for range determination, since seismic and acoustic waves typically have differing propagation speeds.

[0136] FIG. 19 illustrates a schematic representation of a method of the present invention for remote calibration of microphone positions, and is useful when antenna dimensions are large making position calibration difficult. Using a field survey device 192 transducer locations on an antenna 190 can be verified can be verified or recalibrated when the positions have been altered by environmental events.

[0137] The following example is provided for illustrative purposes, and it is to be understood that the present invention is not to be limited or constrained by such example, as the present invention could be practiced in a variety of other applications and configurations.

[0138] In a representative example using the present invention, a deployment of a simple embodiment of the present invention was carried out to validate the capability for locating sources of blast and shock waves in a mountainous environment using the signal averaging technique. In one test 7 elements of a 16 element spherical array were employed as a 3D antenna operating in reduced horizontal mode, after a field validation verified a partially broken data cable. The utilized signal vector library then consisted only of a subset of the 3D library corresponding to incoming waves near parallel to the earth's surface.

[0139] The array was deployed on one side of a bent canyon having canyon wall grades running from zero to 60 degrees, with numerous rock outcroppings, protrusions, and canyon twists and turns. The width of the canyon ranged from approximately 150 to 300 ft in the vicinity of the array at the array altitude, as viewed in different directions. Three shooter positions were employed, all along a N-S line, the first being on the opposite canyon wall at 179 meters along a line nearly parallel to the canyon's average axis, the remaining 2 were 32 meters north and 76 meters south of the antenna Targets for the shooters were spread along a line on the opposite canyon wall, from south to west. All shooters had roughly the same elevation as the antenna array. All of the consequent blast and shock wave time series were processed using the signal averaging technique discussed above and appropriate for short duration single events.

[0140] A total of 39 rifle firings were observed. With the temporal processing window centered on the blast or shock wave pulse, only 5 rifle firings resulted in error estimates of shooter positions above 5 degrees, with the array having a resolving power of approximately 53 degrees. All of these 5 were associated with one shooter position, the one directly north of the array and out of line of site from the array itself. That position was characterized by the shooter shooting from along an irregular rock wall ridge line, generating numerous echoes. The maxima error observed in this instance was 14 degrees. The overall mean absolute bearing error was 3.3 degrees, with 15 of 39 within 2 degrees. A movement of the temporal window to place the blast and shock wave pulses nearer to the end of the window to reduce the effect of reverberation reduced the maximum error to 8 degrees, and 1 to 2 degree fluctuations in other bearings were noted, characterizing the accuracy of the process. With the new temporal window, the overall error was nearer 3.5 degrees, with two thirds of the shootings associated with errors of 3 degrees or less. Use of high density arrays of the type proposed here is anticipated to greatly enhance accuracy.

[0141] Identical processing results were obtained for the partial wave and Fourier signal vector representations. In all cases, data processing involved 1 sec “snapshots” for data collection and 1 sec for processing. With use of an 18 element transducer array, the same acoustic “pulse” processing technique was found to generate valid results inside buildings.

[0142] While preferred embodiments of the invention have been described using specific terms, such description is for present illustrative purposes only, and it is to be understood that changes and variations to such embodiments, including but not limited to the substitution of equivalent features or parts, and the reversal of various features thereof, may be practiced by those of ordinary skill in the art without departing from the spirit or scope of the following claims.