JP2015050659A - Signal processing system and signal processing method - Google Patents

Abstract

PROBLEM TO BE SOLVED: To suppress noise in a restoration signal obtained by restoring compression data.SOLUTION: A measurement matrix generation parameter to be used for generating a random measurement matrix, a restoration matrix generation parameter to be used for generating a random restoration matrix, and a quantization threshold to be used for 2 bit quantization are generated as compression of a signal, the random measurement matrix is generated by using a measurement matrix generation parameter, the signal is compressed by using the random measurement matrix, and 2 bit quantization of the compression signal outputted by a compression sensing part is performed by using the quantization threshold. Also, as restoration of the compressed signal, a random restoration matrix is generated by using a restoration matrix generation parameter, the signal 2 bit quantized is restored, a restoration signal obtained by restoration is outputted, and a result of restoration is reported to a control unit.

Description

  The present invention relates to a signal processing system and a signal processing method.

In order to effectively use the bandwidth of a network, a signal (data series) is converted from analog to digital (AD conversion) and then compressed, and the compressed signal is transmitted.
In this specification, compressing an AD-converted signal and restoring the compressed signal is referred to as compressed sensing.
In this specification, 1-bit compressed sensing refers to compressing an AD converted signal, leaving only the code information of the compressed signal, and restoring the original signal using the remaining code information. Called.

In compressed sensing, compressed data is generated by multiplying input data by a random measurement matrix. The random measurement matrix is composed of random variables, and is composed of the number of rows such that the number of data after compression (data after compression) is smaller than the number of data before compression.
In 1-bit compression sensing, compressed data is generated by multiplying input data by a random measurement matrix and leaving only the sign information of the multiplication result. Since the compressed data is formed only by the code information, a compression effect equivalent to the number of quantization bits can be obtained.

The concept of compressed sensing will be described with reference to FIG.
The input data in the figure is a sparse signal in the frequency domain or time domain, and consists of 100 data. In the figure, an example is shown in which 40 pieces of compressed data are generated by converting 100 pieces of input data using a 40 × 100 random measurement matrix.
In compressed sensing, the compressed data is quantized. When the compressed data is quantized with K bits, the total number of bits of the compressed data is (40 × K) bits. For example, if 8-bit quantization (K = 8) is performed, the total number of bits of compressed data is 320 bits (see, for example, Non-Patent Document 1 and Non-Patent Document 2).

FIG. 15 illustrates a configuration example of a signal processing device that performs compression sensing.
In the signal processing apparatus shown in FIG. 1, an AD conversion unit 1001 converts an analog input signal into a digital signal.
The random measurement matrix generation unit 1002 generates a random measurement matrix composed of random variables.
The compression sensing unit 1003 compresses the digital signal input from the AD conversion unit 1001 using the random measurement matrix generated by the random measurement matrix generation unit 1002, and outputs compressed data.
The random restoration matrix generation unit 1004 generates a random restoration matrix by multiplying the random measurement matrix by an IFFT (Inverse fast Fourier transform) matrix.
The signal restoration unit 1005 restores the compressed data output from the compressed sensing unit 1003 using the random measurement matrix generated by the random measurement matrix generation unit 1002 and outputs a restored signal. When the signal restoration unit 1005 restores the signal, for example, a restoration algorithm such as L1-minimization is used.

The concept of 1-bit compressed sensing will be described with reference to FIG.
The input data in the figure is a sparse signal in the frequency domain or time domain, and consists of 100 data. In the figure, an example in which 60 pieces of compressed data are generated by converting 100 pieces of input data with a 60 × 100 random measurement matrix is shown.
In addition, in 1-bit compressed sensing, 1-bit compressed data is obtained by leaving only code information from the compressed data. In the figure, sign () indicates a function that leaves only code information.
As an example, sign (13, -2, 5, 2) becomes (1, -1, 1, 1). That is, in 1-bit compression sensing, 1-bit compressed data is generated by converting 0 or more code information into 1 and converting code information less than 0 into −1 (code determination). In 1-bit compressed sensing, the number of bits necessary for quantization of compressed data is 1 bit, and the total number of bits as 1-bit compressed data in this case is 60 bits (for example, Non-Patent Document 3). reference).

FIG. 17 shows a configuration example of a signal processing device that performs compression sensing. In this figure, the same parts as those in FIG.
In the signal processing device illustrated in FIG. 17, the AD conversion unit 1001 converts an analog input signal into a digital signal.
The random measurement matrix generation unit 1002 generates a random measurement matrix composed of random variables.
The compression sensing unit 1003 compresses the digital signal input from the AD conversion unit 1001 using the random measurement matrix generated by the random measurement matrix generation unit 1002, and outputs compressed data.
The random restoration matrix generation unit 1004 generates a random restoration matrix by multiplying the random measurement matrix by an IFFT (Inverse fast Fourier transform) matrix.
The code determination unit 1006 performs the code determination described with reference to FIG. 16 on the code of the compressed data output by the compression sensing unit 1003, and generates 1-bit compressed data by leaving only the code information.
The signal restoration unit 1005 restores the 1-bit compressed data output from the code determination unit 1006 using the random measurement matrix generated by the random measurement matrix generation unit 1002 and outputs a restored signal.

E. Candes et al., “An Introduction to Compressive Sampling,” IEEE Signal Processing Magazine, pp. 21-30, March, 2008. E. Candes et al., “Near-Optimal Signal Recovery from Random Projections: Universal Encoding Strategies?” IEEE Transaction on Information Theory, pp. 5406-5425, December, 2006. L. Jacques et al., “Robust 1-Bit Compressive Sensing via Binary Stable Embeddings of Sparse Vectors,” CoRR abs / 1104.3160, 2011.

In 1-bit compressed sensing, information on the absolute value of a signal is lost due to compression.
As an example, FIG. 18 shows the original signal 1 (10, 0, 20, -30, 0) and the original signal 2 (0.1, 0, 0.2, -0.3, 0) before compression. It is shown. The original signal 1 and the original signal 2 are signals in which the ratio of each coefficient in the frequency domain is the same and only the absolute value is different. When such original signal 1 and original signal 2 are respectively compressed and restored by 1-bit compression sensing, the absolute value information is lost in the restored signal of original signal 1 and the restored signal of original signal 2. As a result, the same signal is obtained.

  More specifically, as shown in FIG. 19, information on the absolute value of the original signal is stored until the compression process. However, the code determination by the code determination unit 1006 removes the component of the absolute value of the original signal from the compressed data and extracts only the code information. Thus, in 1-bit compressed sensing, information on the absolute value of the original signal is lost. Since the original signal 1 and the original signal 2 have the same 1-bit compressed data, the restored signal is the same signal as shown in FIG.

As described above, due to the loss of information on the absolute value of the original signal, noise increases with respect to a sparse signal in the time domain (a signal divided into a portion where a signal component exists and a portion where no signal component exists on the time axis).
As a specific example, FIG. 20 shows an original signal that is sparse in the time domain, and a restored signal obtained by compressing and restoring the original signal by 1-bit compression sensing. The original signal has a non-signal portion T1 having no effective signal (effective signal) and a signal portion T2 having an effective signal on the time axis. Note that noise is superimposed on both the non-signal portion T1 and the signal portion T2.
When 1-bit compressed sensing and restoration are performed on such a signal, the absolute value information is lost as described above. As a result, in the restored signal, even if the SNR (Signal to Noise Ratio) of the original signal is high, the absolute value of the noise is amplified.

  An object of the present invention is to provide a technique for suppressing noise in a decompressed signal obtained by decompressing compressed data.

  One aspect of the present invention is to generate a measurement matrix generation parameter used for generating a random measurement matrix, a recovery matrix generation parameter used for generating a random recovery matrix, and a quantization threshold used for 2-bit quantization. A control unit that performs generation, a random measurement matrix generation unit that generates the random measurement matrix using the measurement matrix generation parameter, a compressed sensing unit that compresses a signal using the random measurement matrix, A 2-bit quantization unit that performs 2-bit quantization of the compressed signal output from the compressed sensing unit using the quantization threshold, and a random restoration matrix that generates a random restoration matrix using the restoration matrix generation parameter When generating a 2-bit quantized signal using the generator and the random recovery matrix, and outputting the recovered signal obtained by the recovery Moni, a signal processing system and a signal restoration unit for notifying the restored result to the control unit.

  One aspect of the present invention includes a remote station and a central station, and the remote station uses a measurement matrix generation parameter to generate a random measurement matrix, and uses the random measurement matrix. A compressed sensing unit that compresses a signal and a 2-bit quantum that performs 2-bit quantization of a signal output from the compressed sensing unit using a quantization threshold and transmits the 2-bit quantized signal to the central station The central station performs the generation of the measurement matrix generation parameter, the generation of the restoration matrix generation parameter, and the generation of the quantization threshold, the generated measurement matrix generation parameter, and the generated quantization A control unit that transmits a threshold to the remote station; a random restoration matrix generation unit that generates a random restoration matrix using the generated restoration matrix generation parameter; and A signal processing system comprising: a signal restoration unit that restores a 2-bit quantized signal using a restoration matrix, outputs a restoration signal obtained by the restoration, and notifies the control unit of the restoration result .

  One aspect of the present invention is the above-described signal processing system, in which the signal restoration unit has an absolute value of a signal output by the compressed sensing unit out of a 2-bit quantized signal that is equal to or greater than a quantization threshold. This is restored for the signal indicated by the first bit, and is not restored for the signal indicated by the first bit that the absolute value of the signal output from the compressed sensing unit is less than the quantization threshold.

  One aspect of the present invention is a measurement matrix generation parameter used for generation of a random measurement matrix, generation of a restoration matrix generation parameter used for generation of a random recovery matrix, and an absolute value determination threshold used for determination of the presence / absence of an effective signal A control unit that generates, a random measurement matrix generation unit that generates a random measurement matrix using the measurement matrix generation parameter, a compressed sensing unit that compresses a signal using the random measurement matrix, If the sum of the absolute values of the compressed signals output by the compression sensing unit is greater than or equal to the absolute value determination threshold, a signal with a signal determination value indicating that the compressed signal has an effective signal is output, and the compressed signal If the sum of the absolute values is less than the absolute value determination threshold, an absolute value determination that outputs a signal with a signal determination value indicating that the compressed signal does not have a valid signal is output. And the compressed signal in the signal output by the absolute value determination unit is converted to a sign determination value indicating a positive value if the compression signal is zero or more, and a negative value if less than zero A code determination unit that converts the compressed signal into a code determination value indicating that there is a signal, outputs a signal including the signal determination value and the code determination value, and generates a random recovery matrix using the recovery matrix generation parameter A random restoration matrix generation unit that performs the restoration using the random restoration matrix for the code determination value in the signal output from the code determination unit, outputs a restoration signal obtained by the restoration, and outputs the restoration result to the control unit It is a signal processing system provided with the signal restoration part to notify.

  One aspect of the present invention includes a remote station and a central station, and the remote station generates a random measurement matrix using a measurement matrix generation parameter, and a signal using the random measurement matrix. And a signal determination value indicating that the compressed signal has an effective signal if the sum of absolute values of the compressed signals output from the compressed sensing unit is equal to or greater than an absolute value determination threshold value. Absolute value judgment that outputs a signal with a signal judgment value indicating that the compressed signal does not have a valid signal if the sum of the absolute values of the compressed signals is less than an absolute value judgment threshold. And the compression signal in the signal output by the absolute value determination unit is converted to a sign determination value indicating a positive value if the compression signal is greater than or equal to zero, and a negative value if less than zero about A code determination unit that converts the compressed signal into a code determination value, and outputs a signal including the signal determination value and the code determination value, and the central station generates the measurement matrix generation parameter, and a restoration matrix generation A control unit that generates a parameter and the absolute value determination threshold; a random restoration matrix generation unit that generates a random restoration matrix using the restoration matrix generation parameter; and a signal output from the code determination unit The signal processing system includes a signal restoration unit that performs restoration using the random restoration matrix on the code determination value, outputs a restoration signal obtained by restoration, and notifies the control unit of the restoration result.

  One aspect of the present invention is the above-described signal processing system, wherein the signal restoration unit restores a signal to which a signal determination value indicating that it has an effective signal is given from among the signals output from the code determination unit. Processing is performed, and a signal to which a signal determination value indicating that no valid signal is provided is not restored.

  One aspect of the present invention is to generate a measurement matrix generation parameter used for generating a random measurement matrix, a recovery matrix generation parameter used for generating a random recovery matrix, and a quantization threshold used for 2-bit quantization. A control step for generating, a random measurement matrix generating step for generating the random measurement matrix using the measurement matrix generation parameter, a compressed sensing step for compressing a signal using the random measurement matrix, A 2-bit quantization step for performing 2-bit quantization of the compressed signal output from the compressed sensing step using the quantization threshold, and a random recovery matrix for generating a random recovery matrix using the recovery matrix generation parameter And generating a 2-bit quantized signal using the random recovery matrix and generating the recovered signal Therefore it outputs the resulting recovered signal is a signal processing method and a signal restoration step of notifying the restored result to the control step.

  One aspect of the present invention is a measurement matrix generation parameter used for generation of a random measurement matrix, generation of a restoration matrix generation parameter used for generation of a random recovery matrix, and an absolute value determination threshold used for determination of the presence / absence of an effective signal A control step for generating, a random measurement matrix generation step for generating a random measurement matrix using the measurement matrix generation parameter, a compressed sensing step for compressing a signal using the random measurement matrix, If the sum of absolute values of the compressed signals output by the compression sensing step is equal to or greater than the absolute value determination threshold, a signal with a signal determination value indicating that the compressed signal has an effective signal is output, and the compressed signal If the sum of absolute values is less than the absolute value determination threshold, a signal determination value indicating that the compressed signal does not have a valid signal is assigned. An absolute value determining step for outputting a signal, and if the compressed signal in the signal output by the absolute value determining step is zero or more, the compressed signal is converted to a sign determination value indicating a positive value and less than zero If so, a code determination step of converting the compressed signal into a code determination value indicating a negative value and outputting a signal including the signal determination value and the code determination value, and using the restoration matrix generation parameter Then, a random restoration matrix generation step for generating a random restoration matrix and a code decision value in the signal output by the code decision step are restored using the random restoration matrix, and a restoration signal obtained by the restoration is output. And a signal restoration step of notifying a restoration result to the control step.

  According to the present invention, it is possible to suppress noise in a decompressed signal obtained by decompressing compressed data.

It is a figure which shows the structural example of the signal processing system in 1st Embodiment. It is a figure which shows an example of 2 bit quantization which the signal processing system in 1st Embodiment performs. It is a flowchart which shows the example of a process sequence which the signal processing system in 1st Embodiment performs. It is a figure which shows the structural example of the signal processing system in 3rd Embodiment. It is a figure which shows an example of the absolute value determination and sign determination which the signal processing system in 3rd Embodiment performs. It is a flowchart which shows the example of a process sequence which the signal processing system in 3rd Embodiment performs. It is a figure which shows the example of whole structure of the communication system with which the signal processing system in 5th-8th embodiment is applied. It is a figure which shows the structural example of the remote station in the signal processing system of 5th Embodiment, and a central station. It is a flowchart which shows the example of a process sequence which the remote station in the signal processing system of 5th Embodiment performs. It is a flowchart which shows the example of a process sequence which the central station in the signal processing system of 5th Embodiment performs. It is a figure which shows the structural example of the remote station and central station in the signal processing system of 7th Embodiment. It is a flowchart which shows the example of a process sequence which the remote station in the signal processing system of 7th Embodiment performs. It is a flowchart which shows the example of a process sequence which the central station in the signal processing system of 7th Embodiment performs. It is a figure which shows the concept of compression sensing. It is a figure which shows the structural example of the signal processing apparatus which performs compression sensing. It is a figure which shows the concept of 1 bit compression sensing. It is a figure which shows the structural example of the signal processing apparatus which performs 1 bit compression sensing. It is a figure which shows typically that the information of the absolute value of an original signal is lost from a decompression | restoration signal by 1 bit compression sensing and decoding. It is a figure which shows one specific example about the information of the absolute value of an original signal being lost from a decompression | restoration signal by 1 bit compression sensing and decoding. It is a figure which compares and shows the original signal and decompression | restoration signal in 1 bit compression sensing.

<First Embodiment>
FIG. 1 shows a configuration example of a signal processing system according to the first embodiment of the present invention.
The signal processing system shown in the figure includes an AD conversion unit 101, a random measurement matrix generation unit 102, a compressed sensing unit 103, a 2-bit quantization unit 104, a random restoration matrix generation unit 105, a signal restoration unit 106, and a control unit 107. .

The AD conversion unit 101 converts an analog input signal into digital. The signal input by the AD conversion unit 101 is a signal received through communication, for example.
The random measurement matrix generation unit 102 generates a random measurement matrix using a parameter (measurement matrix generation parameter) used for generating a random measurement matrix. The random measurement matrix generation unit 102 receives measurement matrix generation parameters from the control unit 107. The random measurement matrix is, for example, the number of rows and columns of the random measurement matrix, the type of random variable that forms the random measurement matrix, and the like.

The compression sensing unit 103 compresses a signal using a random measurement matrix.
For this purpose, the compressed sensing unit 103 performs a matrix operation (multiplication) between the digital signal input from the AD conversion unit 101 and the random measurement matrix generated by the random measurement matrix generation unit 102.
At this time, the compressed sensing unit 103 divides the signal input from the AD conversion unit 101 for each number of columns of the random measurement matrix and sequentially performs matrix calculation. For example, when 1000 pieces of data per second are output from the AD conversion unit 101, if the number of rows of the random measurement matrix is 40 and the number of columns is 100 (random matrix of 40 rows × 100 columns), the AD conversion unit 101 The signal to be output is sequentially divided by 100 in correspondence with the number of columns of the random measurement matrix, and the random measurement matrix and the matrix operation are performed. In this case, 1000 signals output per second from the AD converter 101 are divided into 10 100 rows × 1 column vectors, and 40 rows × 100 columns random measurement matrix and matrix calculation are performed. Ten vectors of 40 rows × 1 column are generated. A vector of 40 rows × 1 column generated in this way is compressed data.

  The 2-bit quantization unit 104 performs 2-bit quantization of the compressed signal (compressed data) output from the compressed sensing unit 103 using the quantization threshold input from the control unit 107.

FIG. 2 shows a specific example of 2-bit quantization performed by the 2-bit quantization unit 104.
In the figure, compressed data compressed by the compression sensing unit 103 and 2-bit quantized data obtained by performing 2-bit quantization on the compressed data with a quantization threshold = 10 are shown.
The 2-bit quantization unit 104 sets the first bit of the 2-bit quantized data to “1” if the absolute value of the compressed data is equal to or greater than the quantization threshold (10 or more). On the other hand, if the absolute value of the compressed data is less than the quantization threshold, the first bit of the 2-bit quantized data is set to “0”.
The 2-bit quantization unit 104 sets the second bit of the 2-bit quantized data to “1” if the compressed data is 0 or more, and 2-bit quantization if the compressed data is less than 0 (negative value). The second bit of data is set to “0”.
Thus, the 2-bit quantization unit 104 performs 2-bit quantization of the compressed data.

Returning to FIG.
The random restoration matrix generation unit 105 generates a random restoration matrix using a restoration matrix generation parameter.
Here, the restoration matrix generation parameter may be the same as the measurement row example generation parameter, for example. The random restoration matrix generation unit 105 generates a random measurement matrix using a restoration matrix generation parameter, and generates a random restoration matrix by multiplying the right side of the generated random measurement matrix by an inverse base matrix of a signal basis. To do.

As an example, when the random measurement matrix is a random matrix with 40 rows × 100 columns, the right side of the random measurement matrix is multiplied by an IFFT (Inverse fast Fourier transform) matrix of 100 rows × 100 columns to obtain 40 rows × 100 columns. Generate a random restoration matrix.
In this case, when the signal is restored on another basis, the inverse transform of that basis may be used. For example, when restoring the coefficients in the DCT (Discrete Cosine Transform) domain instead of restoring the coefficients in the frequency domain, a random restoration matrix is generated by multiplying the IDCT (Inverse Discrete Cosine Transform) matrix instead of the IFFT matrix To do.
The random restoration matrix generation unit 105 inputs parameters and base types used for generation of the random measurement matrix from the control unit 107.

The signal restoration unit 106 restores the signal that has been 2-bit quantized by the 2-bit quantization unit 104, outputs a restoration signal obtained by the restoration, and notifies the control unit 107 of the restoration result.
The signal restoration unit 106 uses the random restoration matrix generated by the random restoration matrix generation unit 105 and the signal (2-bit quantized data) output from the 2-bit quantization unit 104 when decoding the signal. Further, the signal restoration unit 106 can restore the signal by using a restoration algorithm such as BIHT (Binary Iterative Hard Thresholding).

The signal restoration unit 106 performs signal restoration using only the second bit of the two bits of the 2-bit quantized data output from the 2-bit quantization unit 104.
For example, when the output of the 2-bit quantizing unit is five 2-bit quantized data (10, 01, 10, 00, 11), the signal restoration unit 106 outputs the second bit of these 2-bit quantized data. (0, 1, 0, 0, 1) and a random restoration matrix are used for restoration.

On the other hand, the first bit of the 2-bit quantized data is used for determining the presence or absence of a signal.
Specifically, when the 2-bit quantized data is (00, 01, 00, 00, 01), the signal restoration unit 106 corresponds to the 2-bit quantized data because the first bits are all “0”. The signal portion to be determined is a no-signal portion that does not have an effective signal (effective signal) only with noise.
On the other hand, if the first bit in the 2 bits as the 2-bit quantized data is “1”, the restored signal restored using the 2-bit quantized data has a valid signal. The signal part is determined.
The signal restoration unit 106 outputs a restoration signal obtained by the restoration. Further, the signal restoration unit 106 notifies the control unit 107 of a restoration result based on a determination result on the restoration signal, the signaled portion, and the non-signal portion.

The control unit 107 generates a measurement matrix generation parameter, a restoration matrix generation parameter, and a quantization threshold. As described above, the measurement matrix generation parameter is used by the random measurement matrix generation unit 102 to generate a random measurement matrix. The restoration matrix generation parameter is used by the random restoration matrix generation unit 105 to generate a random restoration matrix. The quantization threshold is used by the 2-bit quantization unit 104 to perform 2-bit quantization.
In addition, the control unit 107 updates the measurement matrix generation parameter, the recovery matrix generation parameter, and the quantization threshold using the information on the recovery result output from the signal recovery unit 106.

The flowchart of FIG. 3 shows an example of a processing procedure executed by the signal processing system shown in FIG.
The AD conversion unit 101 converts an analog input signal into digital (step S101).
The control unit 107 generates a measurement matrix generation parameter, a restoration matrix generation parameter, and a quantization threshold (Step S102). Further, in the same step S102, the control unit 107 updates the measurement matrix generation parameter, the recovery matrix generation parameter, and the quantization threshold based on the recovery result output from the signal recovery unit 106.

  The random measurement matrix generation unit 102 receives the measurement matrix generation parameter generated by the control unit 107 in step S102, and generates a random measurement matrix using the input measurement matrix generation parameter (step S103).

The compressed sensing unit 103 uses the random measurement matrix generated by the random measurement matrix generation unit 102 in step S103, compresses the signal output from the AD conversion unit 101, and outputs compressed data (step S104).
The 2-bit quantization unit 104 receives the quantization threshold generated by the control unit 107 in step S102, and performs 2-bit quantization of the compressed data output from the compression sensing unit 103 using the input quantization threshold ( Step S105).

The random restoration matrix generation unit 105 receives the restoration matrix generation parameter from the control unit 107, and generates a random restoration matrix using the input restoration matrix generation parameter (step S106).
The signal restoration unit 106 restores the 2-bit quantized data using the random restoration matrix generated by the random restoration matrix generation unit 105, outputs a restoration signal obtained by the restoration, and outputs the restoration result to the control unit 107. Output (step S107).

Second Embodiment
Next, the second embodiment will be described. A configuration example of the signal processing system in the second embodiment may be the same as that in FIG.
The signal restoration unit 106 in the previous first embodiment performs restoration regardless of the presence or absence of a valid signal.
In contrast, the signal restoration unit 106 in the second embodiment performs restoration only for 2-bit quantized data having an effective signal, and does not perform restoration for 2-bit quantized data having no effective signal. Composed.

That is, the signal restoration unit 106 performs restoration on the 2-bit quantized data in which the first bit is “1” among the signals (2-bit quantized data) quantized by the 2-bit quantizing unit 104. The 2-bit quantized data whose first bit is “0” is not restored.
For example, when the 2-bit quantized data is (10, 01, 10, 00, 11) and (00, 01, 00, 00, 01), it is as follows. In this case, the 2-bit quantized data (10, 01, 10, 00, 11) includes a signal whose first bit is 1, and the 2-bit quantized data (00, 01, 00, 00, 01). Does not have a signal whose first bit is 1. This indicates that the 2-bit quantized data (10, 01, 10, 00, 11) has a valid signal, and the 2-bit quantized data (00, 01, 00, 00, 01) has a valid signal. It has no noise and only noise components.

Therefore, the signal restoration unit 106 restores the 2-bit quantized data (10, 01, 10, 00, 11), but the 2-bit quantized data (00, 01, 00, 00, 01). Do not restore.
As described above, in the second embodiment, the signal restoration unit 106 restores only the signaled portion, whereby the signal restoration using only the noise component can be prevented. As a result, useless processing is omitted, so that the processing load can be reduced.

<Third Embodiment>
Subsequently, the third embodiment will be described.
FIG. 4 shows a configuration example of the signal processing system in the third embodiment. In the figure, the same parts as those in FIG. 1 are denoted by the same reference numerals, and the description of the same configuration as in the first embodiment is omitted.

The signal processing system shown in FIG. 4 includes an AD conversion unit 101, a random measurement matrix generation unit 102, a compressed sensing unit 103, a random restoration matrix generation unit 105, a signal restoration unit 106, a control unit 107, an absolute value determination unit 111, and a code determination. The unit 112 is provided.
In FIG. 4, the AD conversion unit 101, the random measurement matrix generation unit 102, the compressed sensing unit 103, and the random restoration matrix generation unit 105 may perform the same process as in FIG. 1.

The absolute value determination unit 111 performs absolute value determination on the compressed data (compressed signal) output from the compressed sensing unit 103.
Specifically, the absolute value determination unit 111 indicates that the compression signal has a valid signal if the sum of the absolute values of the compressed data (compressed signal) output from the compression sensing unit 103 is equal to or greater than the absolute value determination threshold. A signal to which “1” is added as a determination bit (an example of a signal determination value) is output. Further, the absolute value determination unit 111 receives a signal to which “0” is added as a determination bit indicating that the compressed signal does not have a valid signal if the sum of the absolute values of the compressed signals is less than the absolute value determination threshold. Output.
That is, the absolute value determination unit 111 determines whether or not the absolute value of the compressed data is a certain value or more as absolute value determination.
The absolute value determination threshold is generated by the control unit 107. The absolute value determination unit 111 inputs the absolute value determination threshold generated by the control unit 107.

The code determination unit 112 performs code determination for the compressed data.
Specifically, the sign determination unit 112, if the compressed signal in the signal output from the absolute value determination unit 111 is greater than or equal to zero, sign determination data “1” (an example of a code determination value indicating a positive value) ) To convert the compressed signal. On the other hand, if the compressed signal is less than zero, the sign determination unit 112 converts the compressed signal into “0” sign determination data (a sign determination value indicating a negative value). Then, the code determination unit 112 outputs a signal (determination bit-added code determination data) including a determination bit and code determination data.

FIG. 5 shows an example of absolute value determination by the absolute value determination unit 111 and code determination by the code determination unit 112.
The absolute value determination unit 111 divides the compressed data output from the compressed sensing unit 103 corresponding to the number of rows of the random measurement matrix. FIG. 5 shows a case where the number of rows of the random measurement matrix is nine.

  The absolute value determination unit 111 inputs the absolute value determination threshold generated by the control unit 107. In the figure, an example in which the absolute value determination threshold is 100 is given. The absolute value determination unit 111 calculates the sum of the absolute values of the data in the compressed data for one row, and determines whether or not the calculated sum of the absolute values is 100 (absolute value determination threshold) or more.

Then, the absolute value determination unit 111 assigns a determination bit of 1 to the compressed data if the sum of absolute values is 100 or more, and assigns a determination bit of 0 to the compressed data if it is less than 100.
Specifically, since the first compressed data is (100, −110, 250, −310, −500, 180, 50, −320, −50), the sum of the absolute values of the compressed data is 1870. is there. In this case, since the sum of absolute values is 100 or more, 1 determination bit is assigned to the first compressed data.
The second compressed data is (1, -1.1, 2.5, -3.1, -5, 1.8, 0.5, -3.2, -0.5). Therefore, the sum of the absolute values of the compressed data is 18.7. Since the sum of the absolute values in this case is less than 100, a determination bit of 0 is assigned to the second compressed data.

Next, the code determination unit 112 performs code determination on the compressed data (determination bit-added data) to which the determination bit is added by the absolute value determination unit 111 as follows.
That is, the code determination unit 112 converts the compressed data to “1” if the value of the compressed data, which is a portion obtained by removing the determination bit from the determination bit addition data, is 0 or more. On the other hand, if the value of the compressed data excluding the determination bit is less than 0, the compressed data is converted to “0”.

Specifically, in the example of FIG. 5, the first (100, −110, 250, −310, −500, 180, 50, −320, −50) compressed data is (1, 0, 1, 0, 0, 1, 1, 0, 0). The second (1, -1.1, 2.5, -3.1, -5, 1.8, 0.5, -3.2, -0.5) is equivalent to (1, 0 , 1, 0, 0, 1, 1, 0, 0).
At this time, the code determination unit 112 does not discard the determination bit added to the compressed data, but directly discards the signal (determination bit) attached to the data converted according to the code determination result (code determination data). (Assignment code determination data) is output. Thereby, in this embodiment, the absolute value information is maintained without loss.

Returning to FIG.
The signal restoration unit 106 restores the code determination data in the signal output from the code determination unit 112 by using a random restoration matrix, outputs a restoration signal obtained by the restoration, and notifies the control unit 107 of the restoration result. .
When the signal restoration unit 106 restores a signal, a restoration algorithm such as BIHT may be applied as in the first embodiment.

At this time, the signal restoration unit 106 performs restoration using the code determination data obtained by removing the determination bit from the signal (determination bit-added code determination data) output from the code determination unit 112.
Specifically, in the case of FIG. 5, the signals output from the code determination unit 112 are (1, 1, 0, 1, 0, 0, 1, 1, 0, 0) and (0, 1, 0). 1, 0, 0, 1, 1, 0, 0), but the first value of each is a determination bit. The signal restoration unit 106 excludes the determination bits, (1, 0, 1, 0, 0, 1, 1, 0, 0) and (1, 0, 1, 0, 0, 1, 1, 0, 0). The signal is restored by using each code determination data.

The determination bit given to the code determination data is used for determining whether there is a valid signal. That is, if the determination bit is “1”, the signal restoration unit 106 determines that the code determination data corresponds to a signal portion having a valid signal, and if the determination bit is “0”, the corresponding code determination data. Is determined to be a non-signal portion having no valid signal.
For example, the signal restoration unit 106 notifies the control unit 107 of a restoration signal and a determination result indicating the presence or absence of the signal.

  The control unit 107 generates a measurement matrix generation parameter, a restoration matrix generation parameter, and an absolute value determination threshold.

The flowchart of FIG. 6 shows an example of a processing procedure executed by the signal processing system of the third embodiment shown in FIG.
The AD conversion unit 101 converts an analog input signal into digital (step S201).
The control unit 107 generates a measurement matrix generation parameter, a restoration matrix generation parameter, and an absolute value determination threshold (Step S202). Further, in the same step S202, the control unit 107 updates the measurement matrix generation parameter, the recovery matrix generation parameter, and the absolute value determination threshold based on the recovery result output from the signal recovery unit 106.

  The random measurement matrix generation unit 102 receives the measurement matrix generation parameter generated by the control unit 107 in step S201, and generates a random measurement matrix using the input measurement matrix generation parameter (step S203).

  The compressed sensing unit 103 uses the random measurement matrix generated by the random measurement matrix generation unit 102 in step S203 to compress the signal output from the AD conversion unit 101, and outputs compressed data (step S204).

  The absolute value determination unit 111 executes absolute value determination (step S205). That is, the absolute value determination unit 111 compares the sum of the absolute values of the compressed data output from the compression sensing unit 103 with the absolute value determination threshold input from the control unit 107, as described in FIG. In addition, a determination bit of “1” or “0” is added to the compressed data.

  The code determination unit 112 performs code determination (step S206). That is, the code determination unit 112 converts the compressed data value in the compressed data (determination bit-attached data) to which the determination bit is added by the absolute value determination unit 111 to 0 or more, and converts it to “1” code determination data. If it is less than 0, it is converted into code determination data of “0”. The code determination unit 112 outputs a signal (determination bit-added code determination data) obtained by adding a determination bit to the code determination data.

The random restoration matrix generation unit 105 receives the restoration matrix generation parameter from the control unit 107, and generates a random restoration matrix using the input restoration matrix generation parameter (step S207).
The signal restoration unit 106 uses the random restoration matrix generated by the random restoration matrix generation unit 105 to restore the code determination data in the signal output by the code determination unit 112, and outputs a restoration signal obtained by the restoration. The restoration result is output to the control unit 107 (step S208).

<Fourth embodiment>
Subsequently, a fourth embodiment will be described. The configuration example of the signal processing system in the fourth embodiment may be the same as that in FIG. 4 in the third embodiment. However, the restoration processing by the signal restoration unit 106 is different from the third embodiment as described below.
The signal restoration unit 106 according to the fourth embodiment performs restoration processing on the code determination data to which the determination bit “1” is assigned from the code determination data output from the code determination unit 112, and determines the determination bit “0”. The code determination data to which is added is not restored.
That is, the signal restoration unit 106 according to the fourth embodiment performs a restoration process on a signal provided with a determination bit indicating that it has a valid signal among the signals output from the code determination unit 112 and does not have a valid signal. The signal to which the determination bit indicating this is added is not restored.

  In this way, the signal restoration unit 106 restores only the code determination data having a valid signal and does not restore the code decision data having only a noise component that does not have a valid signal. Can be reduced.

<Fifth Embodiment>
Subsequently, a fifth embodiment will be described.
FIG. 7 shows an example of the overall configuration of a communication system including the signal processing system according to the fifth embodiment. The signal processing system according to the fifth embodiment is applied to a communication system in which one central station 300 is communicably connected to a plurality of remote stations 200 via a transmission line 400.

FIG. 8 shows a configuration example of the remote station 200 and the central station 300. First, the configuration of the remote station 200 will be described.
The remote station 200 shown in the figure includes an AD conversion unit 201, a random measurement matrix generation unit 202, a compressed sensing unit 203, and a 2-bit quantization unit 204.
The AD conversion unit 201, the random measurement matrix generation unit 202, the compression sensing unit 203, and the 2-bit quantization unit 204 are respectively the AD conversion unit 101, the random measurement matrix generation unit 102, and the compression in the first embodiment shown in FIG. A configuration similar to that of the sensing unit 103 and the 2-bit quantization unit 104 may be used.

However, the random measurement matrix generation unit 102 in the fifth embodiment receives the measurement matrix generation parameter transmitted from the central station 300, and uses the input measurement matrix generation parameter for the measurement of the random measurement matrix. Similarly, the 2-bit quantization unit 204 receives the quantization threshold transmitted from the central station 300, and uses the input quantization threshold for 2-bit quantization.
Further, the 2-bit quantization unit 104 in the remote station 200 transmits the 2-bit quantized data obtained by the 2-bit quantization to the central station 300 via the transmission path 400.

Next, the configuration of the central station 300 will be described.
The central station 300 includes a random restoration matrix generation unit 301, a signal restoration unit 302, and a control unit 303.
Similar to the random restoration matrix generation unit 105 in FIG. 1, the random restoration matrix generation unit 301 generates a random restoration matrix using a restoration matrix generation parameter. In the fifth embodiment, the random restoration matrix generation unit 301 inputs restoration matrix generation parameters from the control unit 303.

The signal restoration unit 302 receives the 2-bit quantized data transmitted from the remote station 200 via the transmission path 400 and restores the input 2-bit quantized data. Note that the signal restoration unit 302 may perform restoration using the same restoration algorithm as the signal restoration unit 106 of FIG. 1 corresponding to the first embodiment.
The signal restoration unit 302 outputs a restoration signal obtained by the restoration and notifies the control unit 303 of the restoration result.

  The control unit 303 generates a measurement matrix generation parameter, a restoration matrix generation parameter, and a quantization threshold. In addition, the control unit 303 appropriately updates the measurement matrix generation parameter, the recovery matrix generation parameter, and the quantization threshold using the recovery result output from the signal recovery unit 106.

  In addition, the control unit 303 in the fifth embodiment transmits the measurement matrix generation parameter and the quantization threshold to the remote station 200 via the transmission path 400. A random measurement matrix generation unit 202 in the remote station 200 inputs measurement matrix generation parameters received by the remote station 200. In addition, the 2-bit quantization unit 204 inputs the quantization threshold received by the remote station 200.

The flowchart of FIG. 9 shows an example of a processing procedure executed by the remote station 200 in the fifth embodiment.
In the remote station 200, the random measurement matrix generation unit 202 inputs the measurement matrix generation parameter transmitted from the central station 300 and received by the remote station 200 (step S301).
In addition, the 2-bit quantization unit 204 receives the quantization threshold value transmitted from the central station 300 and received by the remote station 200 (step S302).
The AD conversion unit 201 converts an analog input signal into digital (step S303).

  The random measurement matrix generation unit 202 generates a random measurement matrix using the measurement matrix generation parameter input in step S301 (step S304).

  The compressed sensing unit 203 uses the random measurement matrix generated by the random measurement matrix generation unit 102 in step S304, compresses the signal output from the AD conversion unit 101, and outputs compressed data (step S305).

The 2-bit quantization unit 204 performs 2-bit quantization of the compressed data output from the compressed sensing unit 203 using the quantization threshold input in step S302 (step S306).
The 2-bit quantization unit 204 transmits the 2-bit quantized data obtained by the 2-bit quantization in step S306 to the central station 300 (step S307).

The flowchart of FIG. 10 shows an example of a processing procedure executed by the central station 300 in the fifth embodiment.
In the central station 300, the control unit 303 generates a measurement matrix generation parameter, a restoration matrix generation parameter, and a quantization threshold (step S401). In step S401, the control unit 303 updates the measurement matrix generation parameter, the recovery matrix generation parameter, and the quantization threshold based on the recovery result output from the signal recovery unit 106.
The control unit 303 transmits the measurement matrix generation parameter and the quantization threshold among the measurement matrix generation parameter, the restoration matrix generation parameter, and the quantization threshold generated in step S401 to the remote station 200 (step S402).

The signal restoration unit 302 receives the 2-bit quantized data transmitted from the remote station 200 and received by the central station 300 (step S403).
The random restoration matrix generation unit 301 receives the restoration matrix generation parameter generated by the control unit 303 in step S401, and generates a random restoration matrix using the input restoration matrix generation parameter (step S404).
The signal restoration unit 302 uses the random restoration matrix generated by the random restoration matrix generation unit 301 in step S404 to restore the 2-bit quantized data input in step S403. The signal restoration unit 302 outputs the restoration signal obtained by the restoration, and outputs the restoration result to the control unit 303 (step S405).

<Sixth Embodiment>
Subsequently, a sixth embodiment will be described. A configuration example of the signal processing system in the sixth embodiment may be the same as that in FIG. 8 in the fifth embodiment.
In addition, as in the second embodiment, the signal restoration unit 302 in the sixth embodiment restores only 2-bit quantized data having a valid signal, and restores 2-bit quantized data not having a valid signal. It is configured not to perform.
That is, the signal restoration unit 106 restores the signal having the first bit “1” in the 2-bit quantized data output from the 2-bit quantization unit 104, and the signal having the first bit “0”. Do not restore. Thereby, the load of the restoration process in the central station 300 is reduced.

<Seventh embodiment>
Subsequently, a seventh embodiment will be described.
The signal processing system according to the seventh embodiment is included in the communication system having the configuration shown in FIG. 7 as in the fifth and sixth embodiments. That is, the signal processing system according to the seventh embodiment has a configuration in which the remote station 200 and the central station 300 are communicably connected via the transmission line 400.

FIG. 11 shows a configuration example of the remote station 200 and the central station 300 in the seventh embodiment.
First, the remote station 200 will be described. In the figure, the same parts as those in FIG. 8 corresponding to the fifth embodiment are denoted by the same reference numerals, and the description thereof is omitted as appropriate.

The remote station 200 shown in the figure includes an AD conversion unit 201, a random measurement matrix generation unit 202, a compressed sensing unit 203, an absolute value determination unit 211, and a code determination unit 212.
The AD conversion unit 201, the random measurement matrix generation unit 202, the compressed sensing unit 203, the absolute value determination unit 211, and the sign determination unit 212 are respectively the AD conversion unit 101 and the random measurement matrix generation of FIG. 4 corresponding to the third embodiment. The same processing as that of the unit 102, the compression sensing unit 103, the absolute value determination unit 111, and the code determination unit 112 may be executed.

However, the random measurement matrix generation unit 202 in the seventh embodiment receives the measurement matrix generation parameter transmitted from the central station 300 and uses it for measurement of the random measurement matrix. Similarly, the absolute value determination unit 211 receives the absolute value determination threshold value transmitted from the central office 300 and uses it for absolute value determination.
Also, the code determination unit 212 in the remote station 200 sends a signal (determination bit-added code determination data) obtained by adding a determination bit to the code determination data obtained by performing the code determination to the central station 300 via the transmission path 400. Send.

Next, the configuration of the central station 300 will be described.
The central station 300 includes a random restoration matrix generation unit 301, a signal restoration unit 302, and a control unit 303.
The random restoration matrix generation unit 301 generates a random restoration matrix using a restoration matrix generation parameter. Further, the random restoration matrix generation unit 301 inputs restoration matrix generation parameters from the control unit 303.

The signal restoration unit 302 receives a signal (determination bit-added code decision data) transmitted from the remote station 200 via the transmission line 400 and restores the code decision data in the input signal. Note that the signal restoration unit 302 may perform restoration using the same restoration algorithm as the signal restoration unit 106 of FIG. 4 corresponding to the second embodiment.
The signal restoration unit 302 outputs a restoration signal obtained by the restoration and notifies the control unit 303 of the restoration result.

  The control unit 303 generates a measurement matrix generation parameter, a restoration matrix generation parameter, and an absolute value determination threshold. Further, the control unit 303 appropriately updates the measurement matrix generation parameter, the recovery matrix generation parameter, and the absolute value determination threshold using the recovery result output from the signal recovery unit 106.

  In addition, the control unit 303 in the seventh embodiment transmits the measurement matrix generation parameter and the absolute value determination threshold to the remote station 200 via the transmission path 400. The random measurement matrix generation unit 202 in the remote station 200 receives the measurement matrix generation parameter received by the remote station 200 and uses the input measurement matrix generation parameter for generation of the random measurement matrix. Further, the absolute value determination unit 211 inputs the absolute value determination threshold received by the remote station 200 and uses the input absolute value determination threshold for absolute value determination.

The flowchart of FIG. 12 shows an example of a processing procedure executed by the remote station 200 in the seventh embodiment.
In the remote station 200, the random measurement matrix generation unit 202 inputs the measurement matrix generation parameter transmitted from the central station 300 and received by the remote station 200 (step S501).
Moreover, the absolute value determination unit 211 inputs an absolute value determination threshold value transmitted from the central station 300 and received by the remote station 200 (step S502).
The AD conversion unit 201 converts an analog input signal into digital (step S503).

  The random measurement matrix generation unit 202 generates a random measurement matrix using the measurement matrix generation parameter input in step S501 (step S504).

  The compressed sensing unit 203 compresses the signal output from the AD conversion unit 101 using the random measurement matrix generated by the random measurement matrix generation unit 102 in step S504, and outputs compressed data (step S505).

The absolute value determination unit 211 performs absolute value determination on the compressed data output by the compression sensing unit 203 using the absolute value determination threshold input in step S502 (step S506).
The code determination unit 112 performs code determination on the compressed data (determination bit-attached data) to which the determination bit is added by the absolute value determination unit 211 (step S507). The code determination unit 112 transmits a signal (determination bit-added code determination data) obtained by adding a determination bit to the code determination data obtained by the code determination to the central station 300 (step S508).

The flowchart of FIG. 13 shows an example of a processing procedure executed by the central station 300 in the seventh embodiment.
In the central station 300, the control unit 303 generates a measurement matrix generation parameter, a restoration matrix generation parameter, and an absolute value determination threshold (step S601). Further, in the same step S601, the control unit 303 also executes a process of updating the measurement matrix generation parameter, the recovery matrix generation parameter, and the absolute value determination threshold based on the recovery result output from the signal recovery unit 106.
The control unit 303 transmits the measurement matrix generation parameter and the absolute value determination threshold among the measurement matrix generation parameter, the restoration matrix generation parameter, and the absolute value determination threshold generated in step S601 to the remote station 200 (step S602).

The signal restoration unit 302 inputs a signal (determination bit-added code determination data) transmitted from the remote station 200 and received by the central station 300 (step S603).
The random restoration matrix generation unit 301 receives the restoration matrix generation parameter generated by the control unit 303 in step S601, and generates a random restoration matrix using the input restoration matrix generation parameter (step S604).
The signal restoration unit 302 uses the random restoration matrix generated by the random restoration matrix generation unit 301 in step S604 to restore the code determination data in the signal input in step S603. Then, the signal restoration unit 302 outputs the restoration signal obtained by the restoration, and outputs the restoration result to the control unit 303 (step S605).

<Eighth Embodiment>
Subsequently, an eighth embodiment will be described. The configuration example of the signal processing system in the eighth embodiment may be the same as that of the corresponding FIG. 11 in the seventh embodiment.
In addition, the signal restoration unit 302 in the eighth embodiment executes the same restoration process as in the fourth embodiment. That is, the signal restoration unit 302 performs restoration processing on the code determination data to which the determination bit of “1” is added from the code determination data output from the code determination unit 212, and is given the determination bit of “0”. The code determination data is not restored.
Thereby, the processing load in the central station 300 is reduced.

  As described above, in this embodiment, it is possible to include absolute value information in the compressed data by performing 2-bit quantization or performing absolute value determination and code determination in the compression stage. This makes it possible to discriminate between the signaled portion and the non-signal portion, so that it is possible to prevent the noise of the non-signal portion from being amplified and to suppress the noise. In addition, compression efficiency can be improved by suppressing noise. Furthermore, by decoding only the data corresponding to the signaled portion and not decoding the data corresponding to the non-signal portion, the processing load in the restoration can be reduced.

  You may make it implement | achieve the system or each apparatus in embodiment mentioned above with a computer. In that case, a program for realizing this function may be recorded on a computer-readable recording medium, and the program recorded on this recording medium may be read into a computer system and executed. Here, the “computer system” includes an OS and hardware such as peripheral devices. The “computer-readable recording medium” refers to a storage device such as a flexible medium, a magneto-optical disk, a portable medium such as a ROM and a CD-ROM, and a hard disk incorporated in a computer system. Furthermore, the “computer-readable recording medium” dynamically holds a program for a short time like a communication line when transmitting a program via a network such as the Internet or a communication line such as a telephone line. In this case, a volatile memory inside a computer system serving as a server or a client in that case may be included and a program held for a certain period of time. Further, the program may be a program for realizing a part of the above-described functions, and may be a program capable of realizing the functions described above in combination with a program already recorded in a computer system. You may implement | achieve using programmable logic devices, such as FPGA (Field Programmable Gate Array).

  DESCRIPTION OF SYMBOLS 101 ... Conversion part, 102 ... Random measurement matrix production | generation part, 103 ... Compression sensing part, 104 ... 2-bit quantization part, 105 ... Random reconstruction matrix production | generation part, 106 ... Signal restoration part, 107 ... Control part, 111 ... Absolute value Judgment unit, 112 ... code judgment unit

Claims (8)

A control unit that generates a measurement matrix generation parameter used to generate a random measurement matrix, generates a recovery matrix generation parameter used to generate a random recovery matrix, and generates a quantization threshold used for 2-bit quantization; ,
A random measurement matrix generation unit that generates the random measurement matrix using the measurement matrix generation parameter;
A compressed sensing unit that compresses a signal using the random measurement matrix;
A 2-bit quantization unit that performs 2-bit quantization of the compressed signal output by the compressed sensing unit using the quantization threshold;
A random restoration matrix generation unit that generates a random restoration matrix using the restoration matrix generation parameter;
A signal processing system comprising: a signal restoration unit that performs restoration of a 2-bit quantized signal using the random restoration matrix, outputs a restoration signal obtained by the restoration, and notifies the control unit of the restoration result . With remote and central stations,
The remote station is
A random measurement matrix generation unit that generates a random measurement matrix using a measurement matrix generation parameter;
A compressed sensing unit that compresses a signal using the random measurement matrix;
A 2-bit quantization unit that performs 2-bit quantization of a signal output from the compressed sensing unit using a quantization threshold and transmits the 2-bit quantized signal to the central station;
The central office is
Generating the measurement matrix generation parameter, generating the restoration matrix generation parameter, and generating the quantization threshold, and generating the generated measurement matrix generation parameter and the generated quantization threshold to the remote station; ,
A random restoration matrix generation unit that generates a random restoration matrix using the generated restoration matrix generation parameter;
A signal processing system comprising: a signal restoration unit that restores a 2-bit quantized signal using the random restoration matrix, outputs a restoration signal obtained by the restoration, and notifies the control unit of the restoration result . The signal restoration unit
Among the 2-bit quantized signals, the signal indicated by the first bit indicating that the absolute value of the signal output by the compressed sensing unit is equal to or greater than the quantization threshold is restored, and the compressed sensing unit outputs The signal processing system according to claim 1 or 2, wherein the signal whose first bit indicates that the absolute value of the signal is less than a quantization threshold is not restored. Control unit for generating a measurement matrix generation parameter used for generating a random measurement matrix, generating a recovery matrix generation parameter used for generating a random recovery matrix, and generating an absolute value determination threshold used for determining the presence / absence of an effective signal When,
A random measurement matrix generation unit that generates a random measurement matrix using the measurement matrix generation parameter;
A compressed sensing unit that compresses a signal using the random measurement matrix;
If the sum of absolute values of the compressed signals output by the compression sensing unit is equal to or greater than the absolute value determination threshold, a signal with a signal determination value indicating that the compressed signal has an effective signal is output, and the compressed signal An absolute value determination unit that outputs a signal provided with a signal determination value indicating that the compressed signal does not have an effective signal if the sum of absolute values thereof is less than the absolute value determination threshold;
If the compressed signal in the signal output by the absolute value determination unit is zero or more, the compressed signal is converted to a sign determination value indicating a positive value, and if it is less than zero, it is a negative value. A code determination unit that converts the compressed signal into a code determination value, and outputs a signal including the signal determination value and the code determination value;
A random restoration matrix generation unit that generates a random restoration matrix using the restoration matrix generation parameter;
A signal restoration unit that performs restoration using the random restoration matrix for the code decision value in the signal output by the code decision unit, outputs a restoration signal obtained by restoration, and notifies the control unit of the restoration result; A signal processing system provided. With remote and central stations,
The remote station is
A random measurement matrix generation unit that generates a random measurement matrix using a measurement matrix generation parameter;
A compressed sensing unit that compresses the signal using a random measurement matrix; and
If the sum of absolute values of the compressed signals output by the compressed sensing unit is equal to or greater than an absolute value determination threshold, a signal with a signal determination value indicating that the compressed signal has an effective signal is output, and the compressed signal An absolute value determination unit that outputs a signal provided with a signal determination value indicating that the compressed signal does not have an effective signal if the sum of absolute values is less than an absolute value determination threshold;
If the compressed signal in the signal output by the absolute value determination unit is zero or more, the compressed signal is converted to a sign determination value indicating a positive value, and if it is less than zero, it indicates a negative value. A code determination unit that converts the compressed signal into a code determination value and outputs a signal including the signal determination value and the code determination value;
The central office is
A control unit that performs generation of the measurement matrix generation parameter, generation of a recovery matrix generation parameter, and generation of the absolute value determination threshold;
A random restoration matrix generation unit that generates a random restoration matrix using the restoration matrix generation parameter;
A signal restoration unit that performs restoration using the random restoration matrix for the code judgment value in the signal output by the code judgment unit, outputs a restoration signal obtained by restoration, and notifies the control unit of the restoration result; A signal processing system. The signal restoration unit
Among the signals output from the code determination unit, a restoration process is performed on a signal to which a signal determination value indicating that it has an effective signal is applied, and a signal to which a signal determination value indicating that there is no effective signal is applied The signal processing system according to claim 4 or 5, wherein no restoration is performed. A control step for generating a measurement matrix generation parameter used for generating a random measurement matrix, generating a recovery matrix generation parameter used for generating a random recovery matrix, and generating a quantization threshold used for 2-bit quantization; ,
A random measurement matrix generating step for generating the random measurement matrix using the measurement matrix generation parameter;
A compression sensing step for compressing the signal using the random measurement matrix;
A 2-bit quantization step for performing 2-bit quantization of the compressed signal output by the compressed sensing step using the quantization threshold;
A random restoration matrix generation step of generating a random restoration matrix using the restoration matrix generation parameter;
A signal processing method comprising: a signal restoration step of restoring a signal quantized by 2 bits using the random restoration matrix, outputting a restoration signal obtained by the restoration, and notifying the control step of the restoration result . A control step for generating a measurement matrix generation parameter used for generation of a random measurement matrix, generation of a recovery matrix generation parameter used for generation of a random recovery matrix, and generation of an absolute value determination threshold used for determination of presence / absence of an effective signal When,
A random measurement matrix generating step for generating a random measurement matrix using the measurement matrix generation parameter;
A compression sensing step for compressing the signal using the random measurement matrix;
If the sum of the absolute values of the compressed signals output by the compression sensing step is equal to or greater than the absolute value determination threshold, a signal with a signal determination value indicating that the compressed signal has an effective signal is output, and the compressed signal An absolute value determination step of outputting a signal provided with a signal determination value indicating that the compressed signal does not have an effective signal if the sum of absolute values thereof is less than the absolute value determination threshold;
If the compressed signal in the signal output by the absolute value determining step is zero or more, the compressed signal is converted to a sign determination value indicating a positive value, and if the compressed signal is less than zero, it is a negative value. A code determination step of converting the compressed signal into a code determination value, and outputting a signal including the signal determination value and the code determination value;
A random restoration matrix generation step of generating a random restoration matrix using the restoration matrix generation parameter;
A signal restoration step for performing restoration using the random restoration matrix on the code judgment value in the signal output by the code judgment step, outputting a restoration signal obtained by restoration, and notifying the restoration result to the control step; A signal processing method provided.

Images