Acousto-Optical Classifier

The present invention relates to high speed classification systems, and more particularly to an acousto-optical processor for providing a Fourier transform power spectrum which is detected in a detector array, digitized and classified in a digital classifier.

Presently, the Fourier transform is one of the most powerful transforms which can be used as the input to classifiers for the classification of signals. Computer processing compatibility requires the signal to be in digital form. Present systems first digitize the radar signal, then perform the digital Fourier transform, extract the appropriate coefficients and perform the classification. Analog to digital converters are required which are bandwidth limited such that wide bandwidth radar signals can only be processed with difficulty. Such converters are costly, complex and large in size.

Accordingly, the present invention provides a method for overcoming the bandwidth limitation of the present systems through the use of an acousto-optic cell oriented at the Bragg angle to a collimated laser beam. Wide bandwidth radar pulses are converted to acoustic pressure waves by an input piezo-electric transducer which is mechanically coupled to the acousto-optic cell.

Pressure waves are generated in the cell which interact with the laser beam to produce optical diffraction patterns that are proportional to the power spectrum of the input radar pulses. The patterns are detected by a photodiode array, digitized and fed to a digital classifier for real time classification.

The Figure is a block diagram of an acousto-optical classifier.

Referring now to the Figure, a collimated laser beam 12, pulsed or continuous, is aligned with an acousto-optic cell 10 at the Bragg angle, .theta.. With the pulsed or continuous beam an accurate time synchronisation between incoming radar signal and readout can be achieved. A radar pulse to be classified forms the input through a transducer 14 to the acousto-optic cell 10 where it appears as an acoustic pressure wave. As the acoustic pressure wave propagates through the acousto-optic cell 10, it produces changes in the index of refraction, making the cell appear as a phase grating to the laser beam 12. Brillouin scattering occurs by the interaction of the laser beam 12 with the acoustic pressure wave, and since the acousto-optic cell 10 is oriented at the Bragg angle, the scattered light is deflected at the Bragg angle to form a diffracted beam 16.

The loser part of the Figure shows a laser 4 emitting a collimated beam 6 which passes through expanding optics 8 and becomes laser beam 12. Any expanded laser beam output system can be used. One possible system is U.S. Pat. No. 3,609,590.

The diffracted beam 16 carries a spatial intensity modulation proportional to the temporal modulation of the acoustic pressure wave, and hence, the radar pulse to classified. Since the far-field diffraction pattern of a collimated coherent light beam is proportional to the spatial Fourier power spectrum of the spatial intensity variations in the light beam, Fourier transform optics 18 are used to produce the far-field diffraction pattern for the steered beam 16 in the form of a continuous light pattern whose intensity is proportional to the power spectrum of the radar pulse. This light intensity pattern 20 is sensed by a photodetector array 22 and stored as an analog pattern of voltages.

A digitizer 24 converts the analog voltage pattern to a digital pattern which is fed via coupling means 26 to a digital classifier 28. The digital classifier 28 can immediately use the Fourier power spectrum transform to classify the radar pulse. "Introduction to Mathematical Techniques in Pattern Recognition", by H. C. Andrews, Wiley-Interscience, 1972, describes classification techniques.

The undiffracted beam 30 is absorbed by a light trap 32 to avoid interference by backscattering with the diffrated beam 16.

Also, optical scattering, or noise, due to imperfections of the optical system, can be subtracted by digital techniques following the photodetector array 22 before the classification is performed. With this subtraction, the signal to noise ratio of the signals for classification can be significatly improved. Since classification can be achieved with only a small number of features or Fourier coefficients, only an array of a few detectors are required for the photodetector array 20.

Subtractor 34 contains a memory that records the signal present when the laser is off. This is background light that includes scatter from cell 10, lens 18, other optical components and dust particles present in the system. Over short periods of time this background light is invariant. A background reading should usually be good for at least one day. Once this background signal is stored in memory it is subracted in the conventional electronic fashion from the signal present when the laser is on. Thus, subtractor 34 provides a means for digitally subtracting the effects of undesirable optical scattering.

The present invention provides a real time Fourier transform process with an overall processing time considerably faster than that realizeable by discrete Fourier transform implementation in digital computers and able to process signals with a very wide bandwidth.