CDMA receiver with weighted interference cancellation

[0002] The present invention relates to a code division multiple access communications receiver, more particularly to a receiver for use in the base stations and mobile stations of a code division multiple access mobile communications system.

[0003] Code division multiple access (hereinafter, CDMA) is undergoing intensive development as a technique for using bandwidth more efficiently in mobile communications systems. CDMA uses a spectrum spreading process to enable multiple signals to share the same frequency band. A desired signal is extracted from the shared band by a de-spreading process.

[0004] In a direct sequence CDMA system, the different signals are distinguished by the use of different spreading codes. When a signal is extracted, the extracted signal usually contains interference from other signals, due to imperfect orthogonality of the spreading codes and other factors.

[0005] A known method of canceling this so-called co-channel interference estimates the signal received from each transmitting station on each transmission path, and subtracts, from the combined received signal, the signals estimated to have been received from stations other than the desired station. The subtractions may be performed one after another in a serial manner, or they may be performed in a parallel manner after all signals have been estimated. In estimating the received signals, it is also necessary to estimate the behavior of the radio channels, which distort the phase and amplitude of the signal on each transmission path from each station.

[0006] Accurate radio channel estimation is difficult, however, particularly in the parallel case. Therefore, co-channel interference continues to be a problem in CDMA communications.

[0007] It is accordingly an object of the present invention to reduce co-channel interference in the reception of a CDMA signal.

[0008] Another object of the invention is to improve the estimation of symbol values in the reception of a CDMA signal.

[0009] Another object of the invention is to improve the estimation of the radio channels in the reception of a CDMA signal.

[0010] The invented CDMA receiver receives a combined signal from a plurality of transmitting stations, despreads the received signal, thereby estimates the symbol values respreads the symbol values, thereby estimates the signals received from each transmitting station, and subtracts these estimated signals as estimated interference from the received signal, thereby producing a residual signal. The CDMA receiver also weights the estimated symbol value by multiplication by respective weighting factors when the estimated symbol values are respread. The weighting factors are adjusted at certain intervals.

[0011] The weighting factors are calculated so as to reduce the power of the residual signal. According to one aspect of the invention, the weighting factors are calculated from the values of the residual signal, using a mathematical technique such as the least mean square-method. According to another aspect of the invention, a hard decision is made as to the symbol values before the symbol values are weighted and respread, and the weighting factors are calculated from the differences between the estimated symbol values before the hard decision and the values after the hard decision.

[0012] The weighting factors may be used as digital filter coefficients, by convolving the symbol value or values estimated to have been received from the same transmitting station over a certain period of time with two or more weighting factors.

[0013] Besides estimating the symbol values transmitted by each station, the invented CDMA receiver preferably estimates the distortion in each transmission channel, and in estimating the signal received from each station (the estimated interference due to the station), the CDMA receiver preferably includes the estimated distortion. In this case, the weighting factors may be applied to the estimated phase and amplitude distortions, thereby indirectly applying the weighting factors to the estimated symbol values.

[0014] The CDMA receiver may combine the received signal to allow for path diversity from each transmitting station. In this case, separate weighting factors are calculated for each path, a separate estimated interference value is obtained for each path, and the estimated interference values for different paths from the same transmitting station are added together to obtain the estimated interference due to that station.

[0015] The process of estimating the symbol values and estimating and subtracting interference may be carried out in multiple stages, the residual signal output by each stage becoming the input signal to the next stage.

[0016] Subtraction of estimated interference may be carried out in either a serial or a parallel manner. In the parallel case, the estimated interference values are subtracted all at once, after the interference due to all transmitting stations has been estimated. In the serial case, the estimated interference due to each station is subtracted as soon as the estimate has been made. In the serial case, when the weighting factors are calculated from the residual signal, the residual signal from which the weighting factors for each transmitting station are calculated may be the intermediate residual signal left immediately after subtraction of that station's estimated interference. Alternatively, the weighting factors for all transmitting stations may be calculated from the same residual signal, this being the residual signal left after all stations' estimated interference has been subtracted.

[0017] In the attached drawings:

[0018] FIG. 1 is a general block diagram of the invented CDMA receiver;

[0019] FIG. 2 is a more detailed block diagram showing the internal structure of the interference canceler in a first embodiment of the invention;

[0020] FIG. 3 is a still more detailed block diagram showing the internal structure of the signal estimators in the first and second embodiments of the invention;

[0021] FIG. 4 is a block diagram showing a variation of the internal structure shown in FIG. 2;

[0022] FIG. 5 is a block diagram showing a variation of the internal structure shown in FIG. 3;

[0023] FIG. 6 is a block diagram showing the internal structure of the interference canceler in the second embodiment of the invention;

[0024] FIG. 7 is a block diagram showing the internal structure of the signal estimators in a third embodiment of the invention;

[0025] FIG. 8 is a graph illustrating the combined effect of the second and third embodiments;

[0026] FIG. 9 is a block diagram showing the internal structure of the signal estimators in a fourth embodiment of the invention;

[0027] FIG. 10 is a block diagram showing the internal structure of the signal estimators in a fifth embodiment of the invention; and

[0028] FIG. 11 is a block diagram showing the internal structure of the signal estimators in a sixth embodiment of the invention.

[0029] Embodiments of the invention will be described with reference to the attached illustrative drawings, after an explanation of basic principles.

[0030] The baseband signal r_i(n) transmitted by the i-th transmitting station at an arbitrary time n can be represented as follows, where d_iis the symbol data transmitted by the station at that time, PN_i(n) is a pseudorandom noise code assigned to the i-th station as a spreading code, the possible values of d_iand PN_iare plus and minus one, and i is an integer identifying the station.

[0031] r_i(n)=d_i×PN_i(n)

[0032] When there are M transmitting stations, and the signal from the i-th station undergoes a distortion z_iin the transmission channel, the combined received signal r(n) can be expressed as follows. $r\left(n\right)=\sum _{i=1}^M{r}_i\left(n\right)\times z_i$

[0033] A symbol transmitted by the i-th station is extracted by correlating the received signal r(n) with the spreading code PN_i(n) during the symbol interval. If the symbol interval is from n=1 to n=G, where G is the process gain of the code, the calculation is performed as follows. $\begin{array}{c}{R}_i=\left(1/G\right)\times \sum _{n=1}^G\left({\mathrm{PN}}_i\left(n\right)\times \sum _{j=1}^M{r}_j\left(n\right)\times z_j\right)\\ =\left(d_i\times z_i\right)+\left[\left(1/G\right)\times \sum _{n=1}^G\left({\mathrm{PN}}_i\left(n\right)\times \sum _{\begin{array}{c}j=1\\ j\ne i\end{array}}^M{r}_j\left(n\right)\times z_j\right)\right]\end{array}$

[0034] The first term (d_i×z_i) on the right side of the above equation represents the desired signal, while the second term represents co-channel interference.

[0035] In the present invention, the distortion z_ion each transmission path is estimated, by use of a pilot signal, for example, and the extracted symbol value R_iis corrected accordingly. This operation can be represented as follows, where z_i′ is the estimated distortion, z_i′* is the complex conjugate of z_i′, and Δ z_iis the product of z_iand z_i′*. $R_i^{\prime }=\left(d_i\times \Delta z_i\right)+\left\{\left(1/G\right)\times z_i^{\prime *}\times \sum _{n=1}^G\left[{\mathrm{PN}}_i\left(n\right)\times \sum _{\begin{array}{c}j=1\\ j\ne i\end{array}}^M{r}_j\left(n\right)\times z_j\right]\right\}$

[0036] The value of R_i′ differs from the correct symbol value d_ibecause of the error Δz_iin estimating the distortion, and because of co-channel interference. Bit errors therefore occur when the R_i′ values are further demodulated by making a hard decision as to the symbol data (e.g. plus one or minus one) represented by each R_i′.

[0037] The bit error rate can be reduced by estimating the interfering transmitted signals and subtracting the estimated interference. For example, when a hard decision d_i′ has been made for a symbol transmitted by the i-th station, this symbol can be respread and subtracted from the received signal, so that it will not interfere with the demodulation of other signals. The estimated distortion z_i′ is re-included in the respreading operation. The respreading and subtraction are performed as follows to produce a residual signal e(n). $\begin{array}{c}e\left(n\right)=r\left(n\right)-\left[d_i^{\prime }\times {\mathrm{PN}}_i\left(n\right)\times z_i^{\prime }\right]\\ =\left[\sum _{\begin{array}{c}j=1\\ j\ne i\end{array}}^M{r}_j\left(n\right)\times z_j\right]+\left[\left(d_i\times z_i\right)-\left(d_i^{\prime }\times z_i^{\prime }\right)\right]\times {\mathrm{PN}}_i\left(n\right)\end{array}$

[0038] If the hard decision is correct (d_i′=d_i), and the estimated distortion z_i′ is close to the true distortion z_i, then most of the interference caused by the i-th station can be removed in this way before other stations' received signals are extracted. If the hard decision is incorrect, however, or if the estimated distortion z_i′ differs greatly from the true distortion z_i, then subtraction of the respread signal d_i′×PN_i(n)×z_i′ becomes less effective as a method of canceling interference, and may even worsen the interference.

[0039] In the present invention, accordingly, the respread signal is multiplied by a weighting factor w_i, which is adjusted so as to minimize the energy of e(n). That is, the residual signal e(n) is calculated as follows and is treated as an error signal, and w_iis adjusted, by the least mean square method, for example, to minimize the mean square error |e(n)|². $e\left(n\right)=\left[\sum _{\begin{array}{c}j=1\\ j\ne i\end{array}}^M{r}_j\left(n\right)\times z_j\right]+\left[\left(d_i\times z_i\right)-\left(d_i^{\prime }\times z_i^{\prime }\times w_i\right)\right]\times {\mathrm{PN}}_i\left(n\right)$

[0040] The accuracy of the demodulated data can be improved by subtracting as many respread and weighted signals as possible before extracting the signal of the desired station. In a so-called Rake receiver that combines the received signal with different delays to allow for path diversity, the respreading and weighting operation should be performed separately for each path.

[0041] Further gains in accuracy are possible by repetition of this interference cancellation process in a second processing stage. At the end of the first stage, when the interference signals of all transmitting stations have been subtracted, the residual signal e(n) is given as follows. $e\left(n\right)=\sum _{j=1}^M\left\{\left[\left(d_j\times z_j\right)-\left(d_j^{\prime }\times z_j^{\prime }\times w_j\right)\right]\times {\mathrm{PN}}_j\left(n\right)\right\}$

[0042] The signal received from the i-th station can now be more accurately extracted by first adding the i-th respread and weighted signal d_i′×z_i′×w_i×PN_i(n) to e(n) and correlating the resulting sum with PN_i, or more simply, by correlating PN_i(n) with e(n) and adding d_i′×z_i×w_ito the result. The only interference present this time is the interference left in the residual signal e(n), which is usually less than the interference present in the original received signal r(n).

[0043] After the signals from all stations have been re-extracted in this way, and have been respread, weighted, and subtracted from e(n), the same process can be repeated in further stages to gain further accuracy.

[0044] In the descriptions that follow, identical or similar constituent elements occurring in different embodiments will have the same reference numerals.

[0045] Referring to FIG. 1, the first embodiment is a CDMA receiver 10 comprising an antenna 11, a frequency demodulator 12, an interference canceler 13, and a data decoder 14. FIG. 2 shows the internal structure of the interference canceler 13. The interference canceler 13 comprises K stages 21, where K is a positive integer. Each stage 21 comprises M signal estimators 22 and M adders 23, coupled alternately in series as shown, M being an integer greater than one. FIG. 3 shows the internal structure of the m-th signal estimator 22, in the k-th stage 21, where k and m are positive integers not exceeding K and M, respectively. The constituent elements are a spreading-code generator 24, a despreader 25, a correlator 26 internal to the despreader 25, a channel estimator 27, a conjugator 28, a hard-decision unit 29, a respreader 30, and a weighting-factor calculator 31.

[0046] The frequency demodulator 12 and data decoder 14 in FIG. 1 comprise well-known circuits, descriptions of which will be omitted. The interference canceler 13 can be configured by interconnecting separate arithmetic and logic circuits as shown in FIGS. 2 and 3, or by providing a general-purpose processor such as a microprocessor or digital signal processor with software for executing the functions of the individual elements shown in FIGS. 2 and 3. Separate signal estimators 22 may be provided as shown in FIG. 2, or a single signal estimator 22 may be provided, the single signal estimator 22 repetitively performing the same process described below for each transmitting station in turn.

[0047] Next, the operation will be described.

[0048] Referring again to FIG. 1, the antenna 11 receives a combined signal from M transmitting stations. The frequency demodulator 12 filters, amplifies, and demodulates the received signal to obtain a baseband signal r(n). The frequency demodulator 12 includes an analog-to-digital converter (not visible), and outputs the baseband signal r(n) as a digital signal. The interference canceler 13 estimates each transmitting station's symbol values and interference, and subtracts the interference from the received signal. The data decoder 14 performs a decoding process such as a Viterbi convolutional decoding process on the final estimated symbol values to obtain output data.

[0049] Referring again to FIG. 2, each signal estimator 22 receives a pair of residual signals Es and generates an estimated symbol value Ds and an estimated interference signal Ps. The numbers in parentheses identify the stage and station to which the signals pertain. For example, Ds(K, M) is the estimate obtained in the K-th stage of the value of the symbol transmitted by the M-th station. The letter “s” indicates that the signal estimators 22 are coupled in series.

[0050] The first adder 23 subtracts the first estimated interference signal Ps(1, 1) from the input signal r(n), by adding the negative of Ps(1, 1) to r(n), to obtain a first serial residual signal Es(1, 1). Thereafter, each adder 23 subtracts an estimated interference signal from the preceding residual signal to obtain the next serial residual signal. For example, the last adder 23 subtracts the last estimated interference signal Ps(K, M) from the serially preceding residual signal Es(K, M−1) to obtain the last residual signal Es(K, M).

[0051] The received signal r(n), the serial residual signals Es(k, m), and the estimated interference signals Ps(k, m) are all functions of the same time variable (n), even though this is not explicitly indicated for Es(k, m) and Ps(k, m).

[0052] The operation of the signal estimators 22 will be described with reference again to FIG. 3, taking the m-th signal estimator 22 in the k-th stage as an example. The letter “s” is omitted from the signal names in FIG. 3, because this drawing also applies to other embodiments, described later, in which the signal estimators 22 are coupled in parallel. In the present embodiment, D(k, m), E(k, m), and P(k, m) in FIG. 3 are the same as Ds(k, m), Es(k, m), and Ps(k, m) in FIG. 2.

[0053] The spreading-code generator 24 in FIG. 3 generates the spreading code PN_massigned to the m-th station. This code PN_mis the same in all K stages. The correlator 26 in the despreader 25 correlates the residual signal E(k, m−1) from the preceding adder 23 with the spreading code PN_m, thereby despreading the component of the residual signal that is due to the m-th station. The despreader 25 then adds the estimated symbol value D(k−1, m) from the preceding stage and supplies the result to the channel estimator 27.

[0054] As a special case, when m=1, the residual signal from the preceding adder 23 is not E(k, m−1), but is E(k−1, M) if k>1, and is r(n) if k=1. Also, in the first stage (k=1), there is no estimated symbol from the preceding stage; D(0, m) is treated as zero.

[0055] The channel estimator 27 uses, for example, a pilot signal inserted into the transmitted signal to obtain an estimate z′(k, m) of the distortion produced in the transmission channel between the m-th transmitting station and the CDMA receiver 10. The conjugator 28 takes the complex conjugate Z′*(k, m) of this estimated distortion. The despreader 25 multiples the sum of the despread residual signal and the preceding estimated symbol value by this complex conjugate Z′*(k, m), thereby producing an estimated symbol value V from which the estimated distortion has been removed, and supplies this estimated symbol value V to the hard-decision unit 29.

[0056] The hard-decision unit 29 outputs the coded symbol value that is closest to the estimated value V received from the despreader 25. If the possible symbol values are plus and minus one as mentioned earlier, the hard-decision result will be either plus one or minus one. If another type of code is used, the output of the hard-decision unit 29 may represent, for example, a complex number with an absolute value of one.

[0057] The respreader 30 multiples the output of the hard-decision unit 29 by the estimated channel distortion Z′(k, m) and a weighting factor W(k, m) to produce a new estimated symbol value D(k, m), which is output to the next stage. The weighting factor W(k, m) is supplied by the weighting-factor calculator 31. The respreader 30 then subtracts the previous estimated symbol value D(k−1, m) from the new estimated symbol value D(k, m) and respreads the resulting difference by multiplication with the spreading code PN_m, thereby producing the estimated interference signal P(k, m).

[0058] P(k, m) is an estimate of the m-th station's contribution to the input residual signal E(k, m−1), this contribution being interference from the point of view of the other stations. An adder 23 (shown in FIG. 2) subtracts P(k, m) from E(k, m−1) to cancel this estimated interference, thereby producing the next residual signal E(k, m). As this process is repeated, if all goes well, the power of the residual signal will steadily decrease.

[0059] To help ensure that all does go well in this sense, the weighting-factor calculator 31 calculates a weighting factor W(k, m) that tends to minimize the power of the residual signal E(k, m), and adjusts W(k, m) at certain intervals. These intervals are preferably chosen according to the rate of fading of the signal from the m-th transmitting station. For example, the weighting factor can be adjusted at intervals equal to the shortest interval over which significant fading is expected to occur. The optimum W(k, m) can be calculated by applying the well-known least mean square method over an interval of this length, for example.

[0060] The signal estimator 22 in FIG. 3, and the adder 23 in FIG. 2 that subtracts P(k, m) from E(k, m−1), can be viewed as a device for transferring signal power from the residual signal E(k, m−1) to the estimated symbol value D(k, m). As the received signal r(n) travels through the interference canceler 13, interference is steadily removed and the estimated symbol values should become increasingly accurate. So should the estimated interference, but as noted above, accurate estimation of the interference P(K, m) is difficult. The adverse effects of inaccurate estimation of interference are mollified, however, by the weighting factors W(k, m). These weighting factors accordingly improve the accuracy of the final estimated symbol values D(K, m) which are delivered to the data decoder 14, thereby reducing the number of errors made in decoding the data.

[0061] FIG. 4 illustrates a variation of the first embodiment in which all of the weighting-factor calculators 31 in the first stage receive the same residual signal Es(1, M), which is the final residual signal output from that stage. The weighting-factor calculators 31 in the individual signal estimators 22 therefore work in concert to reduce the power of Es(1, M), instead of working separately to reduce the power of Es(1, 1), Es(1, 2), . . . , Es(1, M). Similarly, for values of k greater than one, the weighting-factor calculators 31 in the k-th stage all receive Es(k, M), and work in concert to reduce the power of this residual signal.

[0062] The weighting in this variation also serves the purpose of reducing the adverse effects of inaccurate estimates of interference. Since all weighting-factor calculators 31 in the same stage cooperate in reducing the power of the same residual signal, however, the power of this signal tends to be reduced more effectively.

[0063] FIG. 5 illustrates another variation of the first embodiment, in which the estimated symbol value D(k, m) output by the m-th signal estimator 22 in the k-th stage is not weighted by W(k, m). That is, D(k, m) is equal to the hard-decision value output by the hard-decision unit 29, multiplied by the estimated channel distortion z′(k, m). The weighting factor W(k, m) is applied only in calculating the estimated interference P(K, m). This variation is also expected to reduce the adverse effects of inaccurately estimated interference.

[0064] Referring to FIG. 6, the interference canceler 13 in the second embodiment comprises K stages 32, each with M signal estimators 22 coupled in parallel. The letter “p” is used to identify the signals output by these signal estimators 22, to distinguish these signals from the corresponding signals in the first embodiment.

[0065] The received signal r(n) is furnished to all of the signal estimators 22 in the first stage. An adder 33 subtracts the estimated interference signals Pp(1, 1), Pp(1, 2) . . . , Pp(1, M) output by all of the signal estimators 22 in the first stage 31 from the received signal r(n), thereby producing a first residual signal Ep(1), which is furnished to all of the signal estimators 22 in the second stage. The estimated interference signals Pp(2, m) output by the signal estimators 22 in the second stage are all subtracted from Ep(1) to produce a second residual signal Ep(2). This process continues until a final residual signal Ep(K) is obtained.

[0066] The internal structure and operation of the signal estimators 22 in the second embodiment is the same as in the first embodiment. Each signal estimator 22 receives the estimated symbol value Dp(k−1, m) from the corresponding signal estimator 22 in the preceding stage, and the residual signal Ep(k−1). These signals were denoted D(k−1, m) and E(k, m−1) in FIG. 2. The weighting-factor calculator 31 in each signal estimator 22 receives the residual signal Ep(k), which was denoted E(k, m) in FIG. 3.

[0067] In the second embodiment, the weighting of the estimated interference signals reduces the adverse effect of inaccurate interference estimation to an even greater degree than in the first embodiment. This is partly because the adverse effect was greater to begin with, due to the inaccurate interference estimation inherent in the parallel configuration of the signal estimators 22, but another factor is that all of the weighting-factor calculators 31 in the k-th stage cooperate to minimize the power of the same residual signal Ep(k).

[0068] Referring to FIG. 7, the third embodiment is a Rake receiver that combines the received signal over, in this case, three paths from each transmitting station. The configuration of the signal estimators 22 is accordingly altered as shown in FIG. 7.

[0069] Specifically, the spreading code PN_moutput by the spreading-code generator 24 is given three different delays by three delay units 35, corresponding to the delays on three presumed signal transmission paths from the m-th station. The three delayed spreading codes are furnished to three despreaders 25 and three respreaders 30, which are coupled to three respective channel estimators 27, three conjugators 28, and three weighting-factor calculators 31. The three despreaders 25 produce three estimated signals V1, V2, and V3, which are provided to a Rake correlator 36. The Rake correlator 36 combines these three signals V1, V2, and V3 in an optimal way by the well-known Rake method, and furnishes the combined signal Vh to the hard-decision unit 29.

[0070] The three respreaders 30 produce three estimated interference signals, one for each of the three paths. A path combiner 37 combines (adds) these three signal to produce a single estimated interference signal P(k, m) representing the total interference contributed by the m-th station on all three paths.

[0071] The path-diversity configuration shown in FIG. 7 can be employed with either the serial interconnection scheme of the first embodiment or the parallel scheme of the second embodiment. FIG. 8 shows an example of the effect when the Rake configuration of FIG. 7 is combined with the parallel interconnection scheme of FIG. 6. The number of transmitting stations (M) is shown on the horizontal axis, and the average bit error rate on the vertical axis.

[0072] The data in FIG. 8 were obtained by simulation of the operation of the invented receiver and a conventional receiver. The number of stages (K) was three in both cases. The conventional receiver was identical to the invented receiver, except that it did not use any weighting factors W(k, m). All received signals were assumed to be affected by Rayleigh fading with a maximum Doppler frequency of 40 Hz and equal levels on both waves. Curve 38 is for the conventional receiver; curve 39 is for the invented receiver. Weighting produced an improvement of 37.7% in the bit error rate with eight transmitting stations, and an improvement of 43.2% with twelve transmitting stations.

[0073] Referring to FIG. 9, the fourth embodiment multiplies the estimated channel distortion z′(k, m) by the weighting factors, instead of supplying weighting factors to the respreader 30.

[0074] In this embodiment, instead of a single weighting factor being generated by a single weighting-factor calculator, a set of N weighting factors is generated by a plurality of weighting-factor calculators 40, where N is an arbitrary positive integer. The weighting factors W₁, W₂, . . . , W_Nare supplied to respective multipliers 41, which are coupled in series with delay circuits 42. The estimated channel effect z′(k, m) is fed to the first multiplier. The outputs of the multipliers are added by an adder 43 to produce a weighted estimate of the distortion z″(k, m), which is supplied to the respreader 30. The multipliers 41, delay circuits 42, and adder 43 thus operate as a digital filter that convolves the estimated distortion values z′(k, m) over a certain period of time with a certain number of weighting factors.

[0075] The weighting factors W₁, W₂, . . . , W_Ncan be calculated by a least mean square method as in the preceding embodiments. Least-squares methods of estimating digital filter coefficients are well known.

[0076] The fourth embodiment has substantially the same effect as the first embodiment. In the special case in which N is equal to one, the fourth embodiment has exactly the same effect as the basic variation of the first embodiment, shown in FIG. 3. In this special case (N=1), there is only one multiplier 41, and no delay circuits 42 or adder 43.

[0077] Referring to FIG. 10, the fifth embodiment is similar to the fourth embodiment, except that the weighted estimate of the distortion z″(k, m) is also supplied to the conjugator 28, which furnishes the corresponding complex conjugate z″*(k, m) to the despreader 25. Use of the weighted estimate of the distortion, instead of the unweighted estimate z′(k, m), may improve the accuracy of the estimated symbol value V.

[0078] Referring to FIG. 11, the sixth embodiment differs from the first embodiment in that the weighting-factor calculator 31 estimates the weighting factor W(k, m) from the difference between the estimated symbol value V and the symbol value output by the hard-decision unit 29. This difference is obtained by an adder 45 that subtracts V from the hard decision result. If the estimated symbol values V differ from the hard decision results in a consistent manner over a certain interval, the weighting-factor calculator 31 adjusts the weighting factor W(k, m) according to the difference.

[0079] The sixth embodiment also improves the accuracy of the estimated interference values P(k, m) and reduces the power of the residual signal E(k, m), without requiring feedback of the residual signal to the weighting-factor calculator 31.

[0080] All of the embodiments and variations described above employ the same basic concept of observing the residual signals (or the difference between the estimated symbol values before and after the hard decision), estimating weighting factors that tend to minimize the power of the residual signal, and weighting the estimated symbol values, or weighting the estimated channel distortions and therefore also the estimated symbol values, with the estimated weighting factors. This weighting lessens the inaccuracy of the estimated interference values P(k, m). The overall effect is a reduced bit error rate, as illustrated in FIG. 8.

[0081] Since the weighting factors can compensate for fading, the present invention is particularly suited for use in mobile communication systems, where fading is an ever-present problems. The invented receiver can be employed in both base-station equipment and mobile-station equipment in such mobile communication systems. The usefulness of the invention is not limited to mobile communication systems, however; the invention can be practiced in any CDMA system in which co-channel interference occurs.

[0082] The invention is not limited to the circuit configurations and interconnections shown in the embodiments above. For example, the number of assumed paths in the third embodiment is not limited to three. In the fourth and fifth embodiments, various digital filter configurations are possible other than the configuration illustrated in FIGS. 9 and 10.

[0083] Since P(k, m) is a function of the same time variable (n) as the received signal r(n), the weighting factor W(k, m) in the first embodiment does not have to be a single numerical value; W(k, m) can be a set of digital filter coefficients, making the weighting operation into a digital filtering operation, similar to the operation illustrated in the fourth and fifth embodiments. In the fourth and fifth embodiments, a plurality of weighting factors were convolved with the estimated distortion over a certain interval of time; in the first embodiment, a plurality of weighting factors can be convolved with the estimated symbol values over a certain period of time.

[0084] Those skilled in the art will recognize that further modifications are possible within the scope claimed below.