JP7585870B2 - Adaptive phasing processing device, adaptive phasing system, adaptive phasing processing method, and program for adaptive phasing processing method - Google Patents

Description

This technology relates to an adaptive phasing processing device that performs phasing processing based on signals received by multiple receivers, an adaptive phasing system, an adaptive phasing processing method, and a program that causes a computer to perform this method. In particular, it aims to reduce the amount of processing.

When processing signals using vibration waves such as sound waves as signals, signal processing is performed by phasing the signals received by multiple receivers (antennas: hereafter, receivers or signals from receivers may be referred to as CH) that act as sensors to receive the signals. Phasing processing can be mainly divided into "sector type phasing" that uses signals received by fixed receivers for multiple phasing directions (standby directions), and "CH sliding type phasing" that performs processing by changing and sliding the receiver according to the phasing direction.

CH slide type phasing performs phasing processing using the signal from the receiver that is in the front direction for each phasing direction. With CH slide type phasing, the signal from the receiver used changes for each direction. By performing phasing processing using CH slide type phasing, it is possible to suppress deterioration of detection ability.

In recent years, adaptive phasing has become widely used in phasing processing. Adaptive phasing is a method of adaptively performing phasing processing based on information about the signal and noise detected by the receiver. By performing adaptive phasing, it is possible to suppress signals arriving from a direction different from the desired direction while maintaining the sensitivity of the signal arriving from the desired direction.

Examples of adaptive phasing include Dominant Mode Rejection (DMR), Minimum Variance Distortionless Response (MVDR), and Eigenvector/Beam Association and Excision (EBAE). All of these adaptive phasing methods are superior in suppressing interference sounds than conventional phasing methods.

The EBAE method uses eigenvalue decomposition of the covariance matrix between sensor outputs to associate eigenvectors in the signal subspace with steering vectors, and constrains the sensitivity to the orientation of the associated steering vector (see, for example, non-patent document 1).

Stephen M. Kogon, “Experimental results for passive sonar arrays with eigenvector-based adaptive beamformers”, Signal, Systems and Computers, 2002

If the EBAE adaptive phasing process is applied to a process in which the signals used are changed as in CH sliding phasing, the eigenvalue decomposition of the covariance matrix, which is computationally heavy, will be performed for each phasing direction. This will result in a problem of an increased amount of calculations and other processing. For this reason, although CH sliding phasing has higher detection capabilities, sector phasing is generally used when performing EBAE adaptive phasing.

Therefore, it has been desirable to realize an adaptive phasing processing device, an adaptive phasing system, an adaptive phasing processing method, and a program for the adaptive phasing processing method that can apply EBAE-type adaptive phasing processing to CH slide-type phasing without causing the problems described above.

For this reason, the adaptive phasing processing device according to the present disclosure includes a covariance matrix estimation unit that calculates a covariance matrix between frequency-divided signals received by a plurality of receivers, an eigenvalue decomposition unit that performs eigenvalue decomposition on the covariance matrix to calculate eigenvalues and eigenvectors, a matching unit that performs matching processing to match eigenvectors with steering vectors based on the eigenvalues and eigenvectors, a weight calculation unit that performs adaptive weight processing to calculate adaptive weights based on the matching between the eigenvectors and steering vectors, a normalization unit that normalizes the adaptive weights, a phasing unit that performs phasing processing of frequency-divided signals based on the normalized adaptive weights , and the weight calculation unit performs adaptive weighting, and the normalization weight creation unit creates normalization weights that are values for weighting signals for which phasing processing is performed, the normalization unit performs normalization processing. The covariance matrix estimation unit, the eigenvalue decomposition unit, the correspondence unit, and the weight calculation unit perform processing for all signals, and the normalization unit selects adaptive weights for the signals used by the phasing unit for the phasing processing based on the normalization weights, and performs normalization processing.

The adaptive phasing system disclosed herein also includes a frequency division unit that performs processing to frequency-divide signals received by multiple receivers, and the adaptive phasing processing device described above.

The adaptive phasing method according to the present disclosure includes a covariance matrix estimation step of calculating a covariance matrix between frequency-divided signals received by a plurality of receivers, an eigenvalue decomposition step of performing eigenvalue decomposition on the covariance matrix to calculate eigenvalues and eigenvectors, a matching step of performing matching processing to match eigenvectors with steering vectors based on the eigenvalues and eigenvectors, a weight calculation step of performing adaptive weight processing to calculate adaptive weights based on the matching between eigenvectors and steering vectors, a normalization step of performing normalization processing to normalize the adaptive weights, and a phasing step of performing phasing processing of frequency-divided signals based on the normalized adaptive weights. the weighting process includes a correspondence weight creation step of creating correspondence weights which are weighting values calculated when performing correspondence by correspondence processing in the correspondence step, a steering weight creation step of creating steering weights which are weighting values for each steering vector which are calculated when performing adaptive weight processing in the weight calculation step, and a normalization weight creation step of creating normalization weights which are values used for weighting signals for which phasing processing is performed which are calculated when performing normalization processing in the normalization step, and the covariance matrix estimation step, eigenvalue decomposition step, correspondence step, and weight calculation step process all signals, and the normalization step selects adaptive weights related to the signals used for phasing processing in the phasing step based on the normalization weights , and performs normalization processing.

The program for the adaptive phasing processing method according to the present disclosure includes a covariance matrix estimation step of calculating a covariance matrix between frequency-divided signals received by a plurality of receivers, an eigenvalue decomposition step of calculating eigenvalues and eigenvectors by performing eigenvalue decomposition on the covariance matrix, a matching step of performing a matching process to match eigenvectors with steering vectors based on the eigenvalues and eigenvectors, a weight calculation step of performing adaptive weight processing to calculate adaptive weights based on the matching between eigenvectors and steering vectors, a normalization step of performing a normalization process to normalize the adaptive weights, and a phasing step of performing phasing processing of frequency-divided signals based on the normalized adaptive weights, and further includes a correspondence calculation step before performing each of the correspondence step, the weight calculation step, and the normalization step. the weighting process includes a correspondence weight creation step of creating correspondence weights which are weighting values calculated when performing correspondence by the correspondence process in the correspondence step, a steering weight creation step of creating steering weights which are weighting values for each steering vector which are calculated when performing adaptive weighting in the weight calculation step, and a normalization weight creation step of creating normalization weights which are values used for weighting signals for which phasing processing is performed which are calculated when performing normalization processing in the normalization step, and the covariance matrix estimation step, eigenvalue decomposition step, correspondence step, and weight calculation step process all signals, and the normalization step causes the computer to select adaptive weights related to the signals used for phasing processing in the phasing step based on the normalization weights , and perform normalization processing.

According to this disclosure, the covariance matrix estimation unit, eigenvalue decomposition unit, matching unit, and weight calculation unit process all signals, and the normalization unit selects and processes the signal used by the phasing unit for phasing processing. Therefore, it is possible to improve performance by performing processing using the signal selected by the normalization unit for the phasing direction while suppressing an increase in the amount of processing caused by changing the signal used for phasing processing. Therefore, it is possible to perform phasing processing using adaptive phasing processing of the EBAE method.

1 is a diagram showing a configuration of an adaptive phasing system 100 including an adaptive phasing processing device 40 according to a first embodiment. 3A to 3C are diagrams showing examples of types of a receiver array 200 according to the first embodiment. FIG. 2 is a diagram illustrating an example of a hardware configuration of the adaptive phasing system 100 according to the first embodiment. 10 is a diagram illustrating the processing of an association weight generation unit 26 according to the first embodiment. FIG. 5 is a diagram illustrating the process of a normalization weight generating unit 28 according to the first embodiment. FIG. 3 is a diagram for explaining an example of the procedure of an adaptive phasing processing method performed by the adaptive phasing system 100 according to the first embodiment. FIG.

The following describes embodiments of the adaptive phasing processing device with reference to the drawings. The figures in the drawings in the following description are schematic, and may omit or simplify the configuration as appropriate for the sake of convenience. In addition, in the drawings, the same reference numerals are used to denote the same or equivalent parts, and are common throughout the embodiments described below. The forms of the components shown in the entire specification are merely examples and are not limited to these descriptions. In particular, the combination of components is not limited to the combination in the embodiment, and components described in other embodiments may be applied to other embodiments as appropriate. In addition, when multiple similar devices are distinguished by subscripts, etc., and there is no need to distinguish or identify them, the subscripts may be omitted.

Embodiment 1.
<Outline of the adaptive phasing system in the first embodiment>
1 is a diagram showing the configuration of an adaptive phasing system 100 including an adaptive phasing processing device 40 according to embodiment 1. The adaptive phasing system 100 in embodiment 1 performs processing based on signals from a plurality of receivers 201-1 to 201-M (M is an integer equal to or greater than 2) that constitute a receiver array 200.

Figure 2 shows an example of the type of receiver array 200 according to the first embodiment. Here, the receiver array 200 that sends signals to the adaptive phasing system 100 will be described. Figure 2(a) shows a cylindrical cylinder array (CA: Cylindrical Array). Also, Figure 2(b) shows a conformal array (CfA: ConFormal Array) having a curved surface shape. In the first embodiment, these receiver arrays 200 are used to perform signal processing for a full-circle search for a target sound. Here, the receiver array 200 in the first embodiment is assumed to be a cylinder array or a conformal array.

Here, the receiver array 200 shown in FIG. 2 has a mount to which the array is fixed. In such a receiver array 200, a rear baffle is often installed and a cardioid receiver 201 is used in order to suppress noise coming from the direction opposite to the phasing direction. For this reason, in the receiver array 200, the receiver 201 installed in the direction opposite to the direction from which the target sound arrives is designed not to receive the target sound. If phasing processing is performed assuming that all receivers 201 receive the target sound, there are problems such as deterioration of the amplitude and the collapse of the phase relationship between each receiver 201. For this reason, if sector-type phasing processing is performed using signals received by all receivers 201 of the receiver array 200, the gain of the phasing processing may be worse than expected, which may lead to a deterioration of the detection ability. In response to this, CH slide-type phasing processing can be used to perform phasing processing using the signal from the receiver 201 in the front direction for each phasing direction, thereby suppressing the deterioration of detection capabilities.

The adaptive phasing system 100 is a system that performs frequency division processing and phasing processing on the signals output by each receiver 201 in the receiver array 200. The adaptive phasing system 100 has an input terminal k, a frequency division device 10, an adaptive phasing processing device 40, and an output terminal Rout. Signals sent from each receiver 201 in the receiver array 200 are input to the input terminal k. Therefore, the input terminal k has M terminals, input terminal k-1 to input terminal k-M. Furthermore, the output terminal Rout outputs a signal that has been phased by the adaptive phasing processing device 40.

Furthermore, the frequency division device 10 has a frequency division unit 11. The frequency division unit 11 performs frequency division processing on M signals input from input terminals k-1 to k-M. The adaptive phasing processing device 40 in the first embodiment performs an adaptive phasing processing method using the EBAE method. Here, the adaptive phasing processing device 40 has an EBAE processing unit 20 and a phasing unit 30. The EBAE processing unit 20 has a covariance matrix estimation unit 21, an eigenvalue decomposition unit 22, a matching unit 23, a weight calculation unit 24, a normalization unit 25, a matching weight creation unit 26, a steering weight creation unit 27, and a normalization weight creation unit 28.

FIG. 3 is a diagram showing an example of a hardware configuration of the adaptive phasing system 100 according to the first embodiment. The adaptive phasing system 100 is, for example, an information processing device including a computer. The adaptive phasing system 100 has a storage unit 60 and a control unit 50. The storage unit 60 stores the results of the arithmetic processing performed by the control unit 50. The storage unit 60 is, for example, a recording device such as a hard disk drive (HDD) or a solid state drive (SSD). The control unit 50 has a control processing device 51 such as a central processing unit (CPU) that executes processing according to a program. The control unit 50 also has a memory 52 such as a read-only memory or random access memory that stores a program. The control unit 50 realizes the processing performed by the frequency division device 10 and the adaptive phasing processing device 40.

Here, some or all of the processing functions performed by the frequency division device 10 and the adaptive phasing processing device 40 may be configured with dedicated circuits. The dedicated circuits may be single circuits, composite circuits, programmed processors, ASICs (Application Specific Integrated Circuits), FPGAs (Field-Programmable Gate Arrays), etc. The dedicated circuits may also be combinations of these circuits. By having the dedicated circuits perform some or all of the signal processing, it is possible to speed up the signal processing. Also, the functions of each device and each part of the adaptive phasing system 100 may be shared and processed by multiple information processing devices.

As described above, the frequency division unit 11 performs frequency division processing on the M signals to divide the signal components by frequency. The frequency division unit 11 performs frequency division processing on each of the M signals. The frequency division unit 11 outputs data related to the signals that have undergone frequency division processing to the phasing unit 30 and the covariance matrix estimation unit 21 of the adaptive phasing processing device 40.

In the EBAE processing unit 20 of the adaptive phasing processing device 40, the covariance matrix estimating unit 21 performs processing to calculate the covariance matrix R between a plurality of signals based on the data from the frequency dividing unit 11. The covariance matrix R can be expressed by equation (1). In equation (1), T is the integral time of the covariance matrix R. Furthermore, x _i is a value for each signal, and x _i = [x ₁ , x ₂ , ..., x _M ] ^T. The covariance matrix estimating unit 21 outputs the covariance matrix R between the signals as data to the eigenvalue decomposition unit 22.

The eigenvalue decomposition unit 22 performs eigenvalue decomposition processing to perform eigenvalue decomposition based on the covariance matrix R from the covariance matrix estimation unit 21. Here, the eigenvalue matrix Λ, the matrix E of the eigenvector e, and the eigenvector e calculated by the eigenvalue decomposition unit 22 can be defined by equations (2) to (4).

For the sake of convenience, in the following explanation, we will transform equation (2) into equation (5). Specifically, we will transform the eigenvalue matrix Λ to define the eigenvalue vector λ as in equation (5).

In equation (5), diag means obtaining diagonal elements for a matrix. In addition to rearranging the eigenvalues, the eigenvalue decomposition unit 22 also rearranges the eigenvectors e corresponding to the eigenvalues. The eigenvalue decomposition unit 22 calculates the eigenvalues and eigenvectors e, thereby decomposing the M signals into signal components and noise components. The eigenvalue decomposition unit 22 outputs the calculated eigenvalues and eigenvectors e as data to the association unit 23.

FIG. 4 is a diagram for explaining the processing of the correspondence weight creation unit 26 according to the first embodiment. FIG. 4 is a diagram showing an image of the correspondence weight when the receiver 201 having cardioid directivity is used. FIG. 4(a) shows a case where the receiver array 200 is a cylinder array. Also, FIG. 4(b) shows a case where the receiver array 200 is a conformal array. The correspondence weight creation unit 26 performs a process of calculating and creating a value to be weighted for each correspondence as a correspondence weight when the correspondence unit 23 in the first embodiment described later performs the correspondence processing. The correspondence weight creation unit 26 in the first embodiment performs a process of calculating the value of the directivity coefficient a _b for each receiver 201 according to the phasing direction and creating a directivity coefficient matrix A having the directivity coefficient a _b as an element. Here, the directivity coefficient matrix A can be expressed by Equation (6) and Equation (7). In Equation (6), B is the number of waiting beams. The correspondence weight creating unit 26 further creates a directivity coefficient matrix A in which the directivity coefficients a _b normalized so that the size of the entire matrix becomes 1 are used as correspondence weights, and outputs the directivity coefficient matrix A to the correspondence unit 23 .

The correspondence unit 23 performs a correspondence process for associating the eigenvector e with the steering vector v. The steering vector v means a vector having a phase difference according to the expected signal arrival direction. Here, in the first embodiment, when the correspondence unit 23 performs the correspondence process, a process of multiplying the normalized directivity coefficients a 1 _b created by the correspondence weight creation unit 26 as correspondence weights is performed prior to the correspondence. The weighting by the directivity coefficients a 1 _b by the correspondence unit 23 is indirectly reflected in the selection of data in the process performed by the normalization unit 25 described later. In the correspondence process, the correspondence unit 23 calculates the inner product ES of the eigenvector e and the steering vector v for the number of signals N* determined as the signal subspace based on equations (8) to (10). Here, the steering vector matrix V is defined by equation (8). Also, the steering vector v is defined by equation (9). In equation (8), B is the number of waiting beams. Also, in equation (9), b=1, 2, ..., B.

From equations (8) and (9), for the number of signals N* determined as the signal subspace, the inner product ES of the eigenvector e and the steering vector ν is expressed by the following equation (10).

The correspondence unit 23 outputs the peak direction of the inner product ES of the eigenvector e and the steering vector ν for the number of signals N* in the signal subspace, calculated using equation (10), to the weight calculation unit 24 as data on the correspondence direction between the eigenvector e and the steering vector ν.

The steering weight creation unit 27 creates steering weights to be multiplied with the steering vector v in order for the weight calculation unit 24, which will be described later, to weight the steering vector v used for adaptive weights. The steering weight creation unit 27 in the first embodiment performs a process of calculating the values of directivity coefficients a _b and creating a directivity coefficient matrix A having the directivity coefficients a _b as elements, in the same manner as the association weight creation unit 26. Then, the steering weight creation unit 27 outputs data of the directivity coefficient matrix A to the weight calculation unit 24.

The weight calculation unit 24 performs adaptive weight processing to calculate an adaptive weight matrix W having adaptive weights as elements based on the corresponding direction from the correspondence unit 23. Here, in the first embodiment, when the weight calculation unit 24 performs adaptive weight processing, a process of multiplying the directivity coefficients a _b created by the steering weight creation unit 27 as corresponding weights is performed. The weighting by the weight calculation unit 24 using the directivity coefficients a _b is indirectly reflected in the selection of data in the process performed by the normalization unit 25, which will be described later. The adaptive weight matrix W is expressed by the following equations (11) and (12).

The adaptive weight vector w, which is a column vector in the adaptive weight matrix W, is expressed by the following formula (13 ₎ . _v'b is a steering vector _v'b = vb _ab multiplied by the directivity coefficient _ab . The weight calculation unit 24 does not multiply the normalized directivity coefficient _ab as in the association unit 23, but multiplies the steering vector v by the directivity coefficient _ab itself, which indicates the sensitivity of the directivity, to perform calculations related to the adaptive weight. Here, β is a coefficient obtained from the eigenvalue and is expressed by formulas (14) and (15). DL in formula (14) is a diagonal loading term. Also, δ in formula (15) is an exclusion coefficient. The exclusion coefficient δ is a coefficient that works so that δ=0 in the direction associated in the signal processing performed by the association unit 23 and δ=1 in other directions. The weight calculation unit 24 outputs the calculated adaptive weight matrix W to the normalization unit 25 as data.

The normalization weight generation unit 28 performs a process of calculating and generating a normalization weight vector cb that allows data related to the signal of the receiver 201 to be used to be used in the calculation depending on the phasing direction when the normalization unit 25 described later performs a normalization process. Here, _the normalization weight vector _cb is _cb = [ _c1 , _c2 , ... _cM ] ^T .

FIG. 5 is a diagram for explaining the processing of the normalization weight creation unit 28 according to the first embodiment. Each element of the normalization weight vector c _b has a value between 0 and 1. The normalization weight is set so that signals in a range that is not affected by sensitivity degradation due to cardioid directivity or baffles in the phasing direction are selected. For example, when selecting signals in a range that is not affected by the mount in the receiver array 200, signals in a range of ±90 degrees or ±60 degrees in the phasing direction are used. FIG. 5(a) shows the range of the receivers 201 used when the phasing direction is ±90 degrees. As shown in FIG. 5(b), the normalization weight is set so that the weighting of the signal of the receiver 201 to be used is 1, and the weighting of the signal of the receiver 201 not to be used is 0. Then, the signals of the receivers 201 in between are weighted based on a function having a continuous connection between 0 and 1, such as the trigonometric function cos, to smoothly connect the weight values.

The normalization unit 25 weights each element of the adaptive weight matrix W from the weight calculation unit 24 based on the normalization weight vector _cb from the normalization weight creation unit 28. The weighting increases the difference between the element values related to the signals of the receivers 201 used in the phasing process and the element values related to the signals of the receivers 201 not used. This makes it easier for the data related to the signals of the receivers 201 used in the phasing process to be directly reflected in the selection and processing. Then, the normalization unit 25 performs normalization based on the weighted data of the adaptive weight matrix W. The normalization unit 25 performs normalization using equation (16). The normalization unit 25 outputs data of each component obtained by normalizing the adaptive weight matrix W to the phasing unit 30.

The phasing unit 30 performs phasing processing based on data related to the output from the frequency division unit 11 in the frequency division device 10 and the output from the normalization unit 25. By performing phasing processing using data obtained by normalizing the adaptive weight matrix W, it is possible to suppress signal components arriving from a direction different from the original direction. If the frequency-divided 1-bin sensor output signal is x, the phasing output y output from the phasing unit 30 is expressed by the following equation (17). As a result, a sharp peak is formed in the standby beam output from the output terminal Rout for signals arriving from the original direction.

Figure 6 is a diagram for explaining an example of the procedure of the adaptive phasing processing method performed by the adaptive phasing system 100 according to the first embodiment. An overview of the adaptive phasing processing method according to the first embodiment described above will be explained with reference to Figure 6. When signals are input from the multiple receivers 201 of the receiver array 200, the frequency division unit 11 of the frequency division device 10 performs a frequency division process (step S101).

The covariance matrix estimation unit 21 of the adaptive phasing processing device 40 performs a covariance matrix estimation step to calculate the covariance matrix R between the frequency-divided output signals (step S102). The eigenvalue decomposition unit 22 performs an eigenvalue decomposition step to perform eigenvalue decomposition on the calculated covariance matrix R to calculate the eigenvalues and eigenvectors e (step S103).

The correspondence weight creation unit 26 performs a correspondence weight creation step of creating correspondence weights (step S104). Then, the correspondence unit 23 performs a correspondence step of matching the eigenvector e with the steering vector v (step S105). Here, when performing the correspondence, the correspondence unit 23 performs weighting using the correspondence weights.

The steering weight creation unit 27 performs a steering weight creation step to create steering weights (step S106). Then, the weight calculation unit 24 performs a weight calculation step to calculate an adaptive weight matrix W based on the correspondence direction related to the correspondence between the eigenvalue and the steering vector ν (step S107). Here, when calculating the adaptive weight matrix W, the weight calculation unit 24 performs weighting by multiplying the steering weight by the steering vector ν.

The normalization weight creation unit 28 performs a normalization weight creation step to create normalization weights (step S108). Then, the normalization unit 25 performs a normalization step to normalize the adaptive weight matrix W (step S109). When performing normalization, the adaptive weight matrix W is multiplied by the normalization weights to perform weighting and normalization. The phasing unit 30 performs a phasing process to perform phasing processing on the frequency-divided output signal using the normalized adaptive weights (step S110).

As described above, according to the adaptive phasing system 100 in the first embodiment, in the adaptive phasing processing device 40, the covariance matrix estimator 21, eigenvalue decomposition unit 22, matching unit 23, and weight calculation unit 24 perform processing based on signals from all receivers 201. Then, in the normalization unit 25, normalization processing is performed based on each phasing direction. Therefore, for the processing up to the weight calculation unit 24, it is not necessary to perform processing that changes the signal for each phasing direction. For this reason, it is not necessary to perform eigenvalue decomposition and the like based on signals that change for each direction, and even if EBAE-type adaptive phasing processing is applied to CH slide-type phasing, an increase in the amount of processing can be suppressed. Therefore, real-time processing can be performed.

For example, in the processing of the covariance matrix estimation unit 21, the number of receivers 201 is M. In addition, if the number of receivers 201 used in the CH slide type phasing processing is M/2 and the number of beams is B, the amount of calculation of the covariance matrix R in the processing of the CH slide type phasing is M/2×M/2×B=(M ² ·B)/4. On the other hand, in the adaptive phasing system 100 in the first embodiment, the amount of calculation of the covariance matrix R is M×M (M ² ). Comparing the amount of calculation of both, when the number of beams B is more than 4, the adaptive phasing system 100 in the first embodiment has an advantage in the amount of calculation. Generally, the number of beams B in the case of full-circle monitoring is more than 4 in most cases. Therefore, by using the adaptive phasing system 100 in the first embodiment, it is possible to expect a reduction in the amount of calculation in many cases.

However, when the receiver 201 has cardioid directivity, for example, and the covariance matrix estimation process and eigenvalue decomposition process are performed on the signals of all the receivers 201, the output from the matching unit 23 that matches the eigenvector e with the steering vector ν deteriorates. This deteriorates the ability to remove interference sounds and the like contained in the received signal. For example, when CH slide type phasing is performed, the number of CHs and the aperture area differ slightly between adjacent beams, causing the beam output to become discontinuous in the azimuth direction. When the beam output becomes discontinuous, this leads to deterioration in the accuracy of the interpolation process and the azimuth accuracy in the subsequent signal processing. If the problem is the amount of processing, there is also a method of dividing the receiver array 200 into multiple subarrays, determining the viewing angle for each subarray, and performing sector type phasing. However, the performance deteriorates compared to CH slide type phasing. Also, discontinuity in the azimuth direction occurs between the subarrays.

Therefore, the adaptive phasing system 100 in the first embodiment has a correspondence weight creation unit 26 and a steering weight creation unit 27. The correspondence weight creation unit 26 calculates a directivity coefficient a _b according to the phasing direction. The steering vector v, which is associated with the eigenvector e in the association unit 23, can be treated as a transfer function of the arriving signal. Therefore, the correspondence weight created by the correspondence weight creation unit 26 is multiplied by the directivity coefficient a _b according to the phasing direction, and the influence of the baffle and cardioid processing is set for the steering vector v, which is the transfer function. At this time, the correspondence weight creation unit 26 uses the directivity coefficient a _b normalized so that the overall magnitude is 1 as the correspondence weight, thereby suppressing the deterioration of the data output by the association unit 23.

On the other hand, in order to balance the normal steering vector v term (first term in the numerator and denominator) and the deletion term (second term in the numerator and denominator) in equation (11), the weight calculation unit 24 does not normalize the magnitude to 1, and uses the directivity coefficients a and _b directly from the steering weight creation unit 27. For this reason, the weight calculation unit 24 can perform robust calculations even when performing adaptive weight calculations that incorporate the directivity coefficients a and _b .

The normalization unit 25 can perform CH slide-type phasing while suppressing performance degradation due to rear baffle and cardioid processing, while also ensuring that there is no distortion in the input/output levels due to the limiting of the receivers 201 used by the normalization unit 25 for processing. In addition, the normalization weight creation unit 28 determines weight values that provide a transition width for the weight values related to the signals of the receivers 201 used in the phasing process and the weight values related to the signals of the receivers 201 not used, using trigonometric functions such as cos, and connects them continuously to ensure a smooth transition. This ensures azimuth continuity for the beam output that is the result of the phasing process.

For these reasons, the adaptive phasing processing device 40 of the adaptive phasing system 100 in embodiment 1 can realize processing using CH slide type phasing using the EBAE method.

Other embodiments.
In the adaptive phasing system 100 in the above-mentioned first embodiment, the directivity coefficients a and _b are used as the correspondence weights created by the correspondence weight creation unit 26, but the present invention is not limited to this. The correspondence weights can be coefficients that can reduce the influence of the signal of the receiver 201 in the opposite direction to the arrival direction of the target sound and can correctly associate the eigenvector e with the steering vector v, such as weighting the signal to be subjected to phasing so as to be larger.

In the adaptive phasing system 100 in the first embodiment described above, the directivity coefficients a and _b are used as the steering weights created by the steering weight creation unit 27, but the present invention is not limited to this. As with the correspondence weights, the steering weights can be weighted so that the signal to be subjected to phasing processing is increased, and the steering weights can be coefficients that can reduce the influence of the signal of the receiver 201 in the opposite direction to the arrival direction of the target sound and can correctly perform adaptive weighting.

In addition, in the adaptive phasing system 100 in the first embodiment described above, for the normalization weights created by the normalization weight creation unit 28, the weight values between the weight values related to the signals used in the phasing process and the weight values related to the signals not used are given a transient width based on a trigonometric function. However, this is not limited to this. The transient width can be given by values of other coefficients or functions, etc.

In addition, in the adaptive phasing system 100 in the first embodiment described above, the normalization weights created by the normalization weight creation unit 28 are configured to have a transient width between the weight values related to the signals used in the phasing process and the weight values related to the signals not used, but this is not limited to this. If discontinuity in the azimuth direction does not pose a problem in the results of the phasing process, the normalization weights may have no transient width.

In addition, in the adaptive phasing system 100 in the first embodiment described above, weighting is performed using the normalization weights created by the normalization weight creation unit 28, and normalization processing is performed based on the signals used by the normalization unit 25 for phasing processing, but this is not limited to this. For example, if discontinuity in the azimuth direction is not a problem in the results of the phasing processing, the normalization unit 25 may extract and select the signals to be used for phasing processing without using normalization weights, and perform normalization processing.

In the adaptive phasing system 100 in the above-described first embodiment, the normalization weight creation unit 28 creates normalization weights that weight the signal of the receiver 201 in use by setting the weight to 1 and weight the signal of the receiver 201 not in use by setting the weight to 0. However, this is not limited to this. For example, in the normalization weights created by the normalization weight creation unit 28, normalization weights based on the directivity coefficients a _b may be created as created by the association weight creation unit 26 and the steering weight creation unit 27, and the normalization unit 25 may perform weighting.

The adaptive phasing system 100 in the first embodiment described above is applied to a case where processing related to CH slide type phasing is performed, but is not limited to this. For example, it can also be applied to a phasing method in which different signals are used for each frequency.

In the above-mentioned first embodiment, the receiver array 200 is described as a cylindrical array or a conformal array, but this is not limited to this. For example, the present invention can also be applied to a receiver array 200 such as a spherical array (SA).

REFERENCE SIGNS LIST 10 Frequency division device 11 Frequency division section 20 EBAE processing section 21 Covariance matrix estimation section 22 Eigenvalue decomposition section 23 Correspondence section 24 Weight calculation section 25 Normalization section 26 Correspondence weight creation section 27 Steering weight creation section 28 Normalization weight creation section 30 Phasing section 40 Adaptive phasing processing device 50 Control section 51 Control processing device 52 Memory 60 Storage section 100 Adaptive phasing system 200 Receiver array 201, 201-1 to 201-M Receivers k-1 to k-M Input terminal Rout Output terminal

Claims (10)

A covariance matrix estimator that calculates a covariance matrix between signals received by a plurality of receivers and divided into frequencies;
an eigenvalue decomposition unit that performs eigenvalue decomposition on the covariance matrix to calculate eigenvalues and eigenvectors;
a matching unit that performs a matching process for matching the eigenvector with a steering vector based on the eigenvalue and the eigenvector;
a weight calculation unit that performs adaptive weight processing to calculate adaptive weights based on the correspondence between the eigenvectors and the steering vectors;
a normalization unit that performs a normalization process to normalize the adaptive weights;
a phasing unit that performs phasing processing of the frequency-divided signal based on the normalized adaptive weights ;
an association weight creation unit that creates association weights, which are weighting values calculated by the association unit when performing association by the association process;
a steering weight generation unit that generates a steering weight, which is a weighting value for each steering vector, calculated by the weight calculation unit when the adaptive weight processing is performed;
a normalization weight generation unit that generates a normalization weight, which is a value that is used to weight the signal that is subjected to the phasing process and is calculated when the normalization unit performs the normalization process;
Equipped with
The covariance matrix estimator, the eigenvalue decomposition unit, the matching unit, and the weight calculation unit perform processing for all of the signals, and the normalization unit selects the adaptive weight for the signal used by the phasing unit for the phasing processing based on the normalization weight, and performs the normalization processing. The association weight generation unit
The adaptive phasing processing device according to claim 1 , wherein the corresponding weights are generated so as to weight the signals used in the phasing processing so that the values related to the signals are large. The association weight generation unit
3. The adaptive phasing processing device according to claim 1 , further comprising: determining a directivity coefficient for each signal in accordance with a phasing direction; and normalizing the directivity coefficient to obtain the corresponding weight. The steering weight generation unit is
The adaptive phasing processing device according to claim 1 , wherein the steering weight is generated to weight the signal used in the phasing processing so that a value related to the signal is large. The steering weight generation unit is
The adaptive phasing processing device according to any one of claims 1 to 4 , wherein a directivity coefficient for each signal is determined according to a phasing direction and used as the steering weight. The normalization unit is
The adaptive phasing processing device according to any one of claims 1 to 5 , wherein the normalization weight is generated to weight the signal used in the phasing processing so as to increase the signal. The normalization unit is
The adaptive phasing processing device according to any one of claims 1 to 6 , wherein a directivity coefficient for each signal is determined according to a phasing direction and used as the normalization weight. A frequency division unit that performs processing to divide the frequency of signals received by a plurality of receivers;
An adaptive phasing system comprising the adaptive phasing processing device according to any one of claims 1 to 7 . A covariance matrix estimation step of calculating a covariance matrix between signals received by a plurality of receivers and divided into frequencies;
an eigenvalue decomposition step of performing eigenvalue decomposition on the covariance matrix to calculate eigenvalues and eigenvectors;
a matching process for matching the eigenvector with a steering vector based on the eigenvalue and the eigenvector;
a weight calculation step of performing an adaptive weight processing to calculate adaptive weights based on the correspondence between the eigenvectors and the steering vectors;
a normalization step of performing a normalization process on the adaptive weights;
and a phasing step of performing phasing processing of the frequency-divided signal based on the normalized adaptive weights,
Also, before performing the association step, the weight calculation step, and the normalization step,
a correspondence weight creation step of creating a correspondence weight which is a weighting value calculated when performing correspondence by the correspondence process in the correspondence step;
a steering weight creation step of creating a steering weight, which is a weighting value for each steering vector, to be calculated when performing the adaptive weighting process in the weight calculation step;
a normalization weight creation step of creating a normalization weight which is a value for weighting the signal subjected to the phasing process, the normalization weight being calculated when the normalization process is performed in the normalization step;
having
an adaptive phasing processing method in which the covariance matrix estimation step, the eigenvalue decomposition step, the matching step, and the weight calculation step perform processing for all of the signals, and the normalization step selects the adaptive weights related to the signals used in the phasing processing by the phasing step based on the normalization weights, and performs the normalization processing. A covariance matrix estimation step of calculating a covariance matrix between signals received by a plurality of receivers and divided into frequencies;
an eigenvalue decomposition step of performing eigenvalue decomposition on the covariance matrix to calculate eigenvalues and eigenvectors;
a matching process for matching the eigenvector with a steering vector based on the eigenvalue and the eigenvector;
a weight calculation step of performing an adaptive weight processing to calculate adaptive weights based on the correspondence between the eigenvectors and the steering vectors;
a normalization step of performing a normalization process on the adaptive weights;
and a phasing step of performing phasing processing of the frequency-divided signal based on the normalized adaptive weights,
Furthermore, before performing the association step, the weight calculation step, and the normalization step,
a correspondence weight creation step of creating a correspondence weight which is a weighting value calculated when performing correspondence by the correspondence process in the correspondence step;
a steering weight creation step of creating a steering weight, which is a weighting value for each steering vector, to be calculated when performing the adaptive weighting process in the weight calculation step;
a normalization weight creation step of creating a normalization weight which is a value for weighting the signal subjected to the phasing process, the normalization weight being calculated when the normalization process is performed in the normalization step;
having
a program for an adaptive phasing processing method, the program causing a computer to perform the covariance matrix estimation step, the eigenvalue decomposition step, the matching step, and the weight calculation step for processing all of the signals, and the normalization step for selecting the adaptive weights for the signals used in the phasing processing by the phasing step based on the normalization weights, and performing the normalization processing.

Images