Information processing device, information processing method and information processing program

Abstract

To provide an information processing device, an information processing method and an information processing program capable of reducing a calculation load while suppressing deterioration of spatial resolution in characteristic space adaptive phasing.SOLUTION: An information processing device determines an adaptive weight to be used in phasing processing of received signals received by a sensor array consisting of a plurality of sensors, and comprises: a preprocessing unit that performs a Fourier transform on an extracted received signal, which is a part of the received signal; a provisional calculation unit that calculates a provisional adaptive weight based on an eigenvalue of a covariance matrix of the Fourier-transformed extracted received signal; and a postprocessing unit that calculates an adaptive weight by performing an inverse Fourier transform on the provisional adaptive weight and performing zero-padding interpolation and a Fourier transform on the inverse Fourier-transformed provisional adaptive weight.SELECTED DRAWING: Figure 3

Description

[Technical Field]

[0001] The present invention relates to an information processing device, an information processing method, and an information processing program for determining adaptive weights used in phasing processing. [Background technology]

[0002] It is known that phasing processing is performed on received signals in the field of sonar and the like. Phasing processing is a technology that calculates weights in spatial directions (forming a filter) to maintain the original state of a signal arriving from a specific direction while suppressing ambient noise arriving from directions other than the specific direction. Phasing processing is mainly divided into fixed weight type and adaptive weight type, and adaptive weight type is a method that is particularly effective for suppressing noise with directional characteristics. Non-Patent Document 1 discloses eigenspace adaptive phasing, which is a type of adaptive weight type phasing processing. [Prior art documents] [Non-patent literature]

[0003] [Non-Patent Document 1] Stephen M. Kogon, “Experimental results for passive sonar arrays with eigenvector-based adaptive beamformers”, Systems and Computers, 2002 Summary of the Invention [Problem to be solved by the invention]

[0004] However, adaptive weight phasing requires a larger computational load than fixed weight phasing because adaptive weights are calculated according to the input. Furthermore, eigenspace adaptive phasing requires eigenvalue decomposition to calculate the adaptive weights, which results in a particularly large computational load. Furthermore, when using array sensors such as sonar, the amount of calculation can be reduced by reducing the number of sensors, but reducing the number of sensors results in a decrease in the spatial resolution of the phasing process.

[0005] The present invention has been made to solve the above-mentioned problems, and aims to provide an information processing device, an information processing method, and an information processing program that can reduce the calculation load while suppressing a decrease in spatial resolution in phasing processing. [Means for solving the problem]

[0006] The information processing device of the present invention is an information processing device that determines adaptive weights to be used in phasing processing of received signals received by a sensor array consisting of a plurality of sensors, and includes a pre-processing unit that performs a Fourier transform on an extracted received signal, which is a part of the received signal, a provisional calculation unit that calculates temporary adaptive weights based on the eigenvalues of the covariance matrix of the Fourier transformed extracted received signal, and a post-processing unit that calculates adaptive weights by performing an inverse Fourier transform on the temporary adaptive weights and performing zero-padding interpolation and a Fourier transform on the inverse Fourier transformed temporary adaptive weights.

[0007] The information processing method according to the present invention is an information processing method performed by an information processing device that determines adaptive weights to be used in phasing processing of received signals received by a sensor array consisting of a plurality of sensors, and includes a preprocessing step that performs a Fourier transform on an extracted received signal, which is a part of the received signal; a provisional calculation step that calculates temporary adaptive weights based on the eigenvalues of the covariance matrix of the Fourier transformed extracted received signal; and a calculation step that calculates adaptive weights by performing an inverse Fourier transform on the temporary adaptive weights and performing zero-padding interpolation and a Fourier transform on the inverse Fourier transformed temporary adaptive weights.

[0008] The information processing program according to the present invention is an information processing program for determining adaptive weights to be used in phasing processing of received signals received by a sensor array consisting of a plurality of sensors, and causes a processor of an information processing device to execute the following steps: a preprocessing step of performing a Fourier transform on an extracted received signal, which is a part of the received signal; a provisional calculation step of calculating temporary adaptive weights based on the eigenvalues of the covariance matrix of the Fourier transformed extracted received signal; and a calculation step of calculating adaptive weights by performing an inverse Fourier transform on the temporary adaptive weights and performing zero-padding interpolation and a Fourier transform on the inverse Fourier transformed temporary adaptive weights. [Effects of the Invention]

[0009] According to the present invention, the number of frequencies in the covariance matrix is reduced and frequency interpolation is performed on the calculated temporary adaptive weights, which makes it possible to reduce the computational load of eigenvalue decomposition while suppressing a decrease in spatial resolution in eigenspace adaptive phasing. [Brief explanation of the drawings]

[0010] [Figure 1] FIG. 10 is a diagram for explaining a conventional phasing process. [Figure 2] FIG. 1 is a diagram for explaining the definition of a coordinate system. [Figure 3] 1 is a block diagram showing a phasing processing system according to a first embodiment. [Figure 4] FIG. 4 is a diagram for explaining the processing of a sample extraction unit according to the first embodiment. [Figure 5] FIG. 4 is a diagram for explaining the processing of an interpolation unit according to the first embodiment. [Figure 6] FIG. 4 is a diagram for explaining processing by an extraction processing unit according to the first embodiment. [Figure 7] 3 is a flowchart showing an information processing method according to the first embodiment. [Figure 8] FIG. 10 is a diagram for explaining the processing of a sample extraction unit according to the second embodiment. [Figure 9] 10 is a flowchart showing an information processing method according to the second embodiment. DETAILED DESCRIPTION OF THE INVENTION

[0011] Hereinafter, embodiments of the present invention will be described with reference to the drawings. The present invention is not limited to the following embodiments, and various modifications are possible without departing from the spirit of the present invention. Furthermore, the present invention includes all possible combinations of the configurations shown in the following embodiments. In addition, in each drawing, components with the same reference numerals are the same or equivalent, and this is common throughout the entire specification.

[0012] Embodiment 1 Before describing the first embodiment, an overview of conventional phasing processing, which is a prerequisite for the phasing processing according to the first embodiment, will be described with reference to FIG. 1. FIG. 1 is a diagram for explaining conventional phasing processing. The phasing processing is a technique for enhancing a signal arriving from a desired specific direction with respect to the output of a sensor array composed of multiple sensors that receive signals that become observation values in a sonar, radar, or the like. Hereinafter, a signal received by a sensor may be referred to as a "received signal." The phasing processing aims to suppress ambient noise arriving from directions other than the specific direction while maintaining the signal arriving from the specific direction in its original state by calculating a weight in the spatial direction (forming a filter) using a phase difference that occurs due to differences in the positions of the sensors.

[0013] The conventional phasing process will be explained in more detail. Suppose the number of sensors is M and the received signal vector χ = [χ1, χ2, , χ M ] T is transformed into the frequency domain by Fourier transform: X(k)=[X1(f),X2(f), ,X M (f)] T , the phasing weights are W(f)=[W1(f),W2(f), ,W M (f)] T Then, the phased output Y(f) is expressed by the following equation (1). H means the conjugate transpose operation, [·] T means the transposition operation.

[0014]

number

[0015] FIG. 2 is a diagram illustrating the definition of the coordinate system. In FIG. 2, the x-axis corresponds to the direction in which each array is arranged when the sensor array is a planar array. The y-axis corresponds to the front-to-back direction of the sensor array, with the y-axis + direction pointing forward. The z-axis corresponds to the vertical direction, with the z-axis + direction pointing upward. θ is the horizontal angle (the angle in the xy plane, i.e., the angle relative to the x-axis and y-axis). θ is a positive value in the clockwise region based on the x-axis and a negative value in the counterclockwise region. φ is the elevation angle (the angle relative to the xy plane and z-axis). φ is a positive value in the upward direction and a negative value in the downward direction. u is a direction vector in the xyz space. u is used for three-dimensional phasing. v is a direction vector in the xyz space. v is used for two-dimensional phasing.

[0016] In the case of fixed weight type beamforming, the beamforming weight W is a steering vector v that contains information on the phase difference between sensors. The angle of the azimuth to be beamformed is expressed as the horizontal angle θ b and the elevation angle φ b Then, the steering vector v is expressed by the following equation (2).

[0017]

number

[0018] where τ m (θ b ,φ b ) is the delay given to the m-th sensor. The position vector of the sensor is r m =[x m ,y m ,z m ], the speed of sound is c, and the direction vector is u(θ b ,φ b )=[cosθ b cosφ b, sinθ b cosφ b , sinφ b ]T, then τ m (θ b ,φ b )=(r m u(θ b ,φ b )) / c.

[0019] In the case of adaptive weighted phasing (hereinafter sometimes referred to as "adaptive phasing"), the phasing weight W is determined depending on the input X. Here, as shown in the following equation (3), a constrained minimization problem is solved to find W such that the sensitivity in the direction to be phasing is maintained at 1 and the power of the phasing output is minimized. Note that E[·] represents the ensemble averaging operation.

[0020]

number

[0021] The result of solving equation (3) for W is as shown in the following equation (4).

[0022]

number

[0023] Here, R is the covariance matrix of the received signal vector, and is expressed by the following equation (5).

[0024]

number

[0025] In adaptive phasing, W opt (f,θ b ,φ b ) is applied to equation (1), which improves the lateral resolution compared to fixed-weight type phasing. Furthermore, as shown in Figure 1, in eigenspace type adaptive phasing, eigenvalue decomposition is performed on the covariance matrix R as shown in the following equation (6).

[0026] [Number]

[0027] Here, λ m is an eigenvalue, and e m is the eigenvector corresponding to λ m . Since the covariance matrix is a Hermitian matrix, it can be expressed as follows using the eigenvalues and eigenvectors from the spectral decomposition of the Hermitian matrix.

[0028] [Number] At this time, assuming the number of signals is N (N < M), by rearranging the eigenvalues in descending order of magnitude, the eigenvalues λ1, λ2, ···, λ N caused by the signal power and the eigenvalues λ N+1 , λ N+2 , ···, λ M caused by the noise power can be classified. The noise power α is estimated as follows using λ N+1 , λ N+2 , ···, λ M .

[0029] [Number]

[0030] Then, from equations (4), (7), and (8), the adaptive weights of the eigen-space type adaptive phase synchronization are obtained as in the following equation (9).

[0031] [Number]

[0032] In equation (9), the product of the first numerator term and the input X is the same as in fixed-weight phasing, but the product of the second numerator term and the input X functions as interference noise removal. Therefore, if the steering vector and the eigenvector caused by the signal perfectly match, the signal level can be maintained, but if they do not perfectly match, it is considered interference noise and the signal level drops. This can cause effects such as errors in the detected sensor position. For this reason, the eigenvector e obtained by eigenvalue decomposition of the covariance matrix m The constraints in equation (3) are relaxed in the correspondence between the eigenvector e corresponding to the eigenvalues resulting from the signal, as shown in the following equation (10): m The dot product of and the steering vector v is taken, and the direction in which the dot product is maximum is regarded as the direction in which the signal is arriving.

[0033]

number

[0034] where θ B =[θ1,θ2, ,θ B ], B is the phase-regulating azimuth number. The inner product is maximized (θ max ,φ b ), the calculation is performed excluding the second numerator term in equation (9), which results in processing equivalent to conventional phasing, and makes it possible to avoid a decrease in signal level.

[0035] (Issues with conventional phasing processing) First, as mentioned above, eigenspace adaptive phasing requires eigenvalue decomposition of the covariance matrix. If the number of sensors is M, the amount of calculation per frequency for eigenvalue decomposition is O(M 3 ) Furthermore, since this is a phasing process in the frequency domain, there are as many covariance matrices for eigenvalue decomposition as there are frequencies. Therefore, if the number of frequencies is K, the amount of calculation for eigenvalue decomposition is O(KM 3) As such, eigenvalue decomposition has a large computational load, and the amount of required calculations is the largest among all the processes in eigenspace adaptive phasing. Therefore, the large amount of calculation required for eigenvalue decomposition is a major factor that prevents the realization of real-time processing of eigenspace adaptive phasing. As such, eigenspace adaptive phasing has a large computational load, and it is particularly desirable to reduce the amount of calculation required for eigenvalue decomposition. Note that the computational load is O(KM 3 ), the most effective way to reduce the amount of calculation is to reduce the number of sensors. However, reducing the number of sensors leads to a decrease in the spatial resolution of the phasing process.

[0036] Second, in phasing processing in the frequency domain, if the time length of the Fourier transform does not match the periodic length of the frequency components contained in the received signal, discontinuities will occur in the time waveform converted back to the time domain by inverse Fourier transform.

[0037] (Phasing processing system 1) A phasing processing system 1 according to the first embodiment will be described. FIG. 3 is a block diagram showing the phasing processing system 1 according to the first embodiment. The phasing processing system 1 receives signals that are observation values in a sonar, radar, or the like, and performs eigenspace adaptive phasing on the received signals. As shown in FIG. 3, the phasing processing system 1 includes a sensor array 2 composed of a plurality of sensors that receive signals, and an information processing device 3 that performs phasing processing. The sensors are, for example, microphones.

[0038] The information processing device 3 includes, as functional units, a vector conversion unit 11, a pre-processing unit 12, a provisional calculation unit 13, a post-processing unit 14, a phasing processing unit 15, and an extraction unit 16. Such information processing device 3 is configured by an arithmetic unit such as a microcomputer that realizes each functional unit by a processor reading and executing a program stored in a memory, or hardware such as a circuit device corresponding to each functional unit. The following description of each function of the information processing device 3 will focus on differences from the conventional phasing processing described above, and will omit a description of common or corresponding points.

[0039] The vector conversion unit 11 converts the received signal vector χ received by the sensor array 2 into the frequency domain by Fourier transform. The received signal vector after Fourier transform is represented by X. The number of samples (also called the "FFT order") in the Fourier transform by the vector conversion unit 11 is K. The vector conversion unit 11 outputs the received signal vector X to the phasing processing unit 15. The received signal vector corresponds to the "received signal" of the present invention.

[0040] The pre-processing unit 12 is a functional unit that performs additional processing in addition to the conventional processing described above in order to reduce the number of frequencies. The pre-processing unit 12 performs a Fourier transform on a portion of the received signal vector χ received by the sensor array 2. The pre-processing unit 12 includes a sample extraction unit 21 and a signal conversion unit 22.

[0041] FIG. 4 is a diagram illustrating the processing of the sample extraction unit 21 according to the first embodiment. As shown in FIG. 4, the sample extraction unit 21 extracts the first K / D samples from a received signal vector χ having K samples, and discards the remaining data. D is a sample reduction constant. The received signal vector extracted by the sample extraction unit 21 and from which the number of samples has been reduced is referred to as the extracted received signal vector χ'. The sample extraction unit 21 outputs the extracted received signal vector χ' to the signal conversion unit 22. The extracted received signal vector corresponds to the "extracted received signal" of the present invention.

[0042] The signal conversion unit 22 performs a Fourier transform on the extracted received signal vector χ'. The signal conversion unit 22 outputs the extracted received signal vector X' in the frequency domain after the Fourier transform to the provisional calculation unit 13. Assuming that the extracted received signal vector χ' is a real number, if the number of samples is not reduced and is K as in the conventional phasing processing described above, the (K / 2)+2-th to K-th samples are the complex conjugates of the K / 2-th to 2-nd samples, and the maximum number of frequencies is (K / 2)+1. In contrast, if the extracted received signal vector χ' with the number of samples reduced to K / D is converted to the frequency domain by a Fourier transform, the maximum number of frequencies is (K / D / 2)+1, and the number of frequencies required for calculating the covariance matrix (described later) can be reduced.

[0043] The provisional calculation unit 13 performs the processes of calculating a covariance matrix, eigenvalue decomposition, calculating adaptive weights, and performing signal correspondence in the conventional eigenspace-type adaptive phasing described above. However, the adaptive weights calculated by the provisional calculation unit 13 are those in which the number of samples has been reduced by the pre-processing unit 12. For this reason, the adaptive weights calculated by the provisional calculation unit 13 are referred to as provisional adaptive weights. The provisional calculation unit 13 includes a covariance matrix calculation unit 31, a decomposition unit 32, an adaptive weight calculation unit 33, and a correspondence unit 34.

[0044] The covariance matrix calculation unit 31 calculates the covariance matrix R in the above-mentioned equation (5) based on the extracted received signal vector X′ output from the signal conversion unit 22. The covariance matrix calculation unit 31 outputs the calculated covariance matrix R to the decomposition unit 32.

[0045] The decomposition unit 32 calculates the eigenvalue λ based on the covariance matrix R output from the covariance matrix calculation unit 31 in the above-mentioned equation (7). m , and the eigenvalue λ m The eigenvector e corresponding to m The decomposition unit 32 calculates the calculated eigenvalue λ m and eigenvector e m to the adaptive weight calculation unit 33. The decomposition unit 32 outputs the calculated eigenvector e m is output to the association unit 34.

[0046] The adaptive weight calculation unit 33 calculates the eigenvector e output from the decomposition unit 32 in the above-mentioned equation (9). m Based on this, the temporary adaptive weights W in the frequency domain are calculated. pre Calculate W in Equation (9). opt is W pre The adaptive weight calculation unit 33 calculates the temporary adaptive weight W pre is output to the post-processing unit 14.

[0047] The matching unit 34 calculates the eigenvector e obtained by eigenvalue decomposition of the covariance matrix. mSpecifically, in the above-mentioned equation (10), the eigenvector e m and the steering vector v, the signal arrival direction θ max The correlation unit 34 calculates the calculated signal arrival direction θ max is output to the phasing processor 15.

[0048] The post-processing unit 14 is a functional unit that performs additional processing in addition to the conventional processing described above in order to interpolate the number of frequencies. The post-processing unit 14 calculates the temporary adaptive weights W pre is inverse Fourier transformed, and the temporary adaptive weights w pre By performing zero-padding interpolation and Fourier transform on the adaptive weight W opt The post-processing unit 14 includes a weight inverse conversion unit 41, an interpolation unit 42, and a weight conversion unit 43.

[0049] The weight inverse transform unit 41 calculates the temporary adaptive weights W pre , and calculate the temporary adaptive weights w pre The weight inverse transform unit 41 calculates the calculated temporary adaptive weight w pre is output to the interpolation unit 42.

[0050] 5 is a diagram illustrating the processing of the interpolation unit according to the first embodiment. As shown in FIG. 5, the interpolation unit 42 calculates temporary adaptive weights w pre In the zero-filling interpolation, the temporary adaptive weight w pre Interpolation is performed on the temporary adaptive weight w' so that the number of samples changes from K / D shown in FIG. 5(a) to K shown in FIG. 5(b). In other words, the number of samples to be interpolated is K-(K / D). The temporary adaptive weight w', whose central portion has been zero-padded and interpolated, is called the temporary adaptive weight w'. pre The interpolation unit calculates the temporary adaptive weight w' pre is output to the weight conversion unit 43.

[0051] The weight conversion unit 43 converts the temporary adaptive weight w' pre , and the adaptive weights Wopt The weight conversion unit 43 calculates the adaptive weight W opt is output to the phasing processor 15. The interpolator 42 performs interpolation to obtain the temporary adaptive weight w' pre The adaptive weight W after Fourier transform of opt The number of frequencies can be set to the same number as the number of frequencies when the pre-processing unit 12 does not reduce the number of samples (FFT order).

[0052] The phasing processor 15 uses the χ output from the vector converter 11 and the adaptive weight W output from the weight converter 43 in the above-mentioned equation (1). opt , and signal arrival direction θ max The phased output Y is calculated based on the above.

[0053] The extraction unit 16 is a functional unit that performs processing added to the conventional processing described above in order to extract the central portion of the output data after phasing processing. The extraction unit 16 performs an inverse Fourier transform on the phasing processing result based on the adaptive weight calculated by the post-processing unit 14, and extracts the central portion of the data of the inverse Fourier transformed phasing processing result. The extraction unit 16 has a phasing inverse transform unit 61 and an extraction processing unit 62.

[0054] The phase-matching inverse transform unit 61 performs an inverse Fourier transform on the phase-matching output Y to obtain the phase-matching output y pre The phase-matching inverse transform unit 61 calculates the phase-matching output y pre is output to the extraction processing unit 62.

[0055] 6 is a diagram for explaining the processing of the extraction processing unit 62 according to the first embodiment. As shown in FIG. 6, the extraction processing unit 62 extracts the phased output y pre The number of samples to be extracted is K / D. The extracted phased output sample is called phased output y. The extraction processing unit 62 outputs the calculated phased output y to a device that detects an object in a specific direction based on the phase processing result.

[0056] The series of operations performed by the interpolation unit 42, weighting unit 43, phasing processing unit 15, phasing inverse transformation unit 61, and extraction processing unit 62 executes a process called the overlap-save method. The interpolation unit 42 performs zero padding in the center of the data, and the phasing output y pre By extracting the central data from the other data, it is possible to avoid occurrence of discontinuities in the time waveform.

[0057] (Information processing method) Next, an information processing method by the information processing device 3 according to the first embodiment will be described with reference to FIG. 7. FIG. 7 is a flowchart showing the information processing method according to the first embodiment. First, the vector conversion unit 11 performs a Fourier transform on the received signal vector χ received by the sensor array 2 (step S1). The order of the Fourier transform here is K. Next, the sample extraction unit 21 extracts the first K / D samples from the received signal vector χ, which has K samples, as an extracted received signal vector χ', and discards the remaining data (step S2). Furthermore, the signal conversion unit 22 performs a Fourier transform on the extracted received signal vector χ' to calculate an extracted received signal vector X' (step S3). Then, the covariance matrix calculation unit 31 calculates the covariance matrix R (step S4).

[0058] Here, for each frequency, eigenvalue decomposition of the covariance matrix R, calculation of temporary adaptive weights, and signal correspondence are performed. Specifically, it is determined whether or not loops for the number of frequencies (maximum (K / D / 2)+1 times) have been completed (step S5), and if loops for the number of frequencies have not been completed (step S5: NO), the processes of steps S6 to S8 are performed until loops for the number of frequencies are completed. First, the decomposition unit 32 performs eigenvalue decomposition of the covariance matrix R, and calculates the eigenvalues λ m and eigenvector e m Next, the adaptive weight calculation unit 33 calculates the temporary adaptive weight W in the frequency domain. pre (Step S7). The association unit 34 also calculates the eigenvector e m and the steering vector v, and the signal arrival direction θ max(Step S8) The order of executing the process of step S7 and the process of step S8 may be reversed.

[0059] When the loop is completed for the number of frequencies (step S5: YES), interpolation from the temporary adaptive weight to the adaptive weight is performed by the processing of steps S9 to S11. First, the weight inverse conversion unit 41 converts the temporary adaptive weight W pre , and calculate the temporary adaptive weights w pre Next, the interpolation unit calculates the temporary adaptive weights w in the time domain (step S9). pre The temporary adaptive weight w' is calculated by zero-filling the center of pre (Step S10). Then, the weight conversion unit 43 calculates the temporary adaptive weight w' pre , and the adaptive weights W opt is calculated (step S11).

[0060] After the adaptive weights are calculated, the phasing processor 15 performs phasing processing to calculate the phasing output Y (step S12). Then, the phasing inverse transformer 61 performs an inverse Fourier transform on the phasing output Y to obtain the phasing output y pre is calculated (step S13), the extraction processing unit 62 extracts the phased output y pre A central sample is extracted from the signal, and a phased output y is calculated (step S14). Note that the processing of step S1 does not have to be performed in the above order as long as it is performed before the processing of step S12. For example, step S1 may be performed immediately before step S12.

[0061] As described above, according to the first embodiment, the number of frequencies in the covariance matrix is reduced, and frequency interpolation is performed on the calculated temporary adaptive weights. Therefore, in the phasing process, it is possible to reduce the calculation load while suppressing a decrease in spatial resolution. Specifically, by reducing the FFT order, the number of frequencies K in the covariance matrix is reduced to K / D. As a result, the amount of calculation for eigenvalue decomposition is reduced to O((K / D)M 3 ) and the calculated temporary adaptive weight Wpre Then, frequency interpolation is performed by performing an inverse Fourier transform on the inverse Fourier transformed temporary adaptive weights and zero-padding interpolation and a Fourier transform on them. Therefore, even if the number of frequencies in the covariance matrix is reduced, adaptive weights can be calculated for the same number of frequencies as the adaptive weights calculated without reducing the number of frequencies. Therefore, it is possible to suppress a decrease in the spatial resolution of the phasing process.

[0062] Furthermore, according to embodiment 1, by introducing the overlap-save method, it is possible to avoid the occurrence of discontinuities in the time waveform that occur when the time length of the Fourier transform does not match the period length of the frequency components contained in the received signal.

[0063] Embodiment 2 The second embodiment differs from the first embodiment in that samples are also extracted from the received signal vector χ, which was discarded in the first embodiment. In the second embodiment, the same parts as those in the first embodiment are denoted by the same reference numerals and their explanations are omitted, and the explanation will focus on the differences from the first embodiment.

[0064] 8 is a diagram illustrating the processing of the sample extraction unit 21 according to the second embodiment. As shown in FIG. 8, the sample extraction unit 21 extracts K / D samples from the input received signal vector χ. Sample extraction is performed K / D samples at a time until (K / D)*d>K (d is the number of loops, d=1, 2, . . . , D), and if the number of remaining samples after the extraction operation is a fraction less than K / D, these are discarded. In this way, the sample extraction unit 21 also extracts samples from the time waveform that was discarded in the first embodiment. The sample extraction unit 21 outputs a plurality of extracted received signal vectors χ'd (d=1, 2, . . . , D) extracted from the received signal vector χ to the signal conversion unit 22.

[0065] The signal conversion unit 22 performs a Fourier transform on each of the maximum D extracted received signal vectors χ'd extracted by the sample extraction unit 21. The signal conversion unit 22 outputs each of the multiple extracted received signal vectors X'd after the Fourier transform to the covariance matrix calculation unit 31.

[0066] The covariance matrix calculation unit 31 calculates a covariance matrix for each of the plurality of extracted received signal vectors X'd after the Fourier transform. The covariance matrix calculated for each K / D sample is defined as the covariance matrix Rd. In this way, in the second embodiment, data is extracted for each K / D sample from among the K samples and used to calculate the covariance matrix.

[0067] Furthermore, the covariance matrix calculation unit 31 integrates the covariance matrix Rd calculated for each K / D sample to calculate the covariance matrix R. That is, the covariance matrix calculation unit 31 integrates up to D calculated covariance matrices Rd to obtain one covariance matrix R. The integration is performed using the following equation (11).

[0068]

number

[0069] Although the average is used for the integral in equation (11), an exponential integral may be used. The covariance matrix calculation unit 31 outputs the covariance matrix R to the decomposition unit 32.

[0070] (Information processing method) Next, an information processing method by the information processing device 3 according to the second embodiment will be described with reference to Fig. 9. Fig. 9 is a flowchart showing the information processing method according to the second embodiment. Here, only differences from the information processing method according to the first embodiment will be described. In the second embodiment, the sample extraction unit 21 determines whether or not (K / D)*d>K (d is the loop number, d=1, 2, . . . , D) is satisfied for the extracted sample (step S21). If (K / D)*d≦K for the extracted sample (step S21: NO), the processes of steps S2 to S4 are repeatedly executed. In the processes of steps S2 to S4, multiple covariance matrices Rd are calculated for each K / D sample.

[0071] When (K / D)*d>K for the extracted sample (step S21: YES), it is considered that sample extraction is complete, and the covariance matrix calculation unit 31 integrates the multiple covariance matrices Rd calculated for each of the K / D samples to calculate the covariance matrix R (step S22). The subsequent processing is the same as in embodiment 1 except that the covariance matrix R is the result of integrating the multiple covariance matrices Rd.

[0072] In the second embodiment, data on received signal vectors, which were discarded in the first embodiment, are also used to calculate the covariance matrix. The covariance matrix is a matrix that estimates the correlation between sensors, and the more data used in the calculation, the higher the estimation accuracy. Therefore, according to the second embodiment, it is possible to apply weight calculation that is more suitable for suppressing ambient noise than in the first embodiment.

[0073] In the second embodiment, the number of times that the signal conversion unit 22 performs the Fourier transform increases by D−1 compared to the first embodiment. However, because the covariance matrix calculation unit 31 integrates up to D covariance matrices to obtain one covariance matrix, the number of eigenvalue decomposition calculations is the same as in the first embodiment. The computational complexity of the Fourier transform, which has increased by D−1 times, can be said to be minor compared to the computational complexity of the eigenvalue decomposition. Therefore, the computational complexity of the entire processing of the eigenspace adaptive phasing described in the second embodiment is equivalent to the computational complexity of the entire processing of the eigenspace adaptive phasing described in the first embodiment.

[0074] Although the first and second embodiments of the present invention have been described above, the present invention is not limited to the above-described first and second embodiments, and various modifications and applications are possible without departing from the spirit and scope of the present invention. For example, the method of reducing the amount of processing by reducing the number of covariance matrices described in the embodiments can also be used for QR decomposition, which is another method for solving the eigenvalue problem of a covariance matrix, instead of eigenvalue decomposition. In addition, adaptive phasing can also be used in a method of calculating the inverse matrix of the covariance matrix from the above equation (4) to find adaptive weights, and this method is also effective in reducing the amount of calculation required for inverse matrix calculation in such a method. MVDR (Minimum Variance Distortionless Response) is known as a representative adaptive phasing method that includes an inverse matrix calculation. [Explanation of symbols]

[0075] 1 phasing processing system, 2 sensor array, 3 information processing device, 11 vector conversion unit, 12 pre-processing unit, 13 provisional calculation unit, 14 post-processing unit, 15 phasing processing unit, 16 extraction unit, 21 sample extraction unit, 22 signal conversion unit, 31 covariance matrix calculation unit, 32 decomposition unit, 33 adaptive weight calculation unit, 34 matching unit, 41 weight inverse conversion unit, 42 interpolation unit, 43 weight conversion unit, 61 phasing inverse conversion unit, 62 extraction processing unit.

Claims

1. An information processing device that determines adaptive weights used in phasing processing of received signals received by a sensor array consisting of a plurality of sensors, a pre-processing unit that performs a Fourier transform on an extracted received signal that is a part of the received signal; a provisional calculation unit that calculates provisional adaptive weights based on eigenvalues of a covariance matrix of the extracted received signal that has been Fourier transformed; a post-processing unit that performs an inverse Fourier transform on the temporary adaptive weights and performs zero-padding interpolation and a Fourier transform on the temporary adaptive weights to calculate the adaptive weights. Information processing device.

2. an extracting unit that performs an inverse Fourier transform on the phasing processing result based on the adaptive weight calculated by the post-processing unit, and extracts a central portion of the data of the inverse Fourier transformed phasing processing result; The information processing device according to claim 1 .

3. The preprocessing unit extracts a plurality of the extracted received signals from the received signals, The provisional calculation unit obtains one covariance matrix by integrating a plurality of covariance matrices corresponding to the plurality of extracted received signals.

3. The information processing device according to claim 1 or 2.

4. An information processing method performed by an information processing device that determines adaptive weights to be used in phasing processing of received signals received by a sensor array consisting of a plurality of sensors, comprising: a pre-processing step of performing a Fourier transform on an extracted received signal that is a part of the received signal; a provisional calculation step of calculating provisional adaptive weights based on eigenvalues of a covariance matrix of the extracted received signals that have been Fourier transformed; a calculation step of calculating the adaptive weights by performing an inverse Fourier transform on the temporary adaptive weights and performing zero-padding interpolation and a Fourier transform on the temporary adaptive weights that have been inverse Fourier transformed. Information processing methods.

5. An information processing program for determining adaptive weights to be used in phasing processing of received signals received by a sensor array consisting of a plurality of sensors, a pre-processing step of performing a Fourier transform on an extracted received signal that is a part of the received signal; a provisional calculation step of calculating provisional adaptive weights based on eigenvalues of a covariance matrix of the extracted received signals that have been Fourier transformed; a calculation step of calculating the adaptive weight by performing an inverse Fourier transform on the temporary adaptive weight and performing zero-padding interpolation and a Fourier transform on the temporary adaptive weight after the inverse Fourier transform, Information processing program.

Images