JP6086541B2 - Signal processing system and signal processing method - Google Patents

Description

  The present invention relates to a signal processing system and signal for compressing a signal in an analog domain before performing AD (Analog-to-Digital) conversion on the signal, AD converting the compressed signal, and restoring the original signal by digital signal processing. It relates to the processing method.

  Conventionally, it is known that a network band can be effectively utilized by AD-converting, compressing, and transmitting a signal (data series). In this specification, a technique for AD-converting, compressing, and restoring a signal is referred to as a conventional digital compression sensing technique. Note that a technique of compressing a signal in the analog domain before performing AD conversion on the signal, AD converting and restoring the compressed signal is referred to as a conventional analog compressed sensing technique. Conventional digital compression sensing technology requires a high-speed AD converter to perform processing after AD conversion of the signal, but conventional analog compression sensing compresses the signal in the analog domain, so high-speed AD conversion is required. No converter is required.

  The concept of conventional digital compressed sensing (for example, see Non-Patent Documents 1 and 2) is shown in FIG. FIG. 8 assumes a sparse signal in the frequency domain or time domain. First, the input signal is sampled by an AD converter having a speed that is Nyquist rate of the input signal (twice the bandwidth of the input signal). Next, the sampled signal is divided for each number of columns of the random measurement matrix and multiplied with the random measurement matrix. For example, in FIG. 8, since a random measurement matrix composed of 40 × 100 is used, the sampled signal is divided into 100 pieces and multiplied with the random measurement matrix. As a result, 40 compressed data are obtained from 100 input signals. Next, the compressed data is restored using a random restoration matrix and a restoration algorithm such as an L1-minimization restoration algorithm.

  FIG. 9 shows a system configuration for performing conventional digital compressed sensing. In FIG. 9, an input signal is converted from an analog signal to a digital signal by the AD conversion unit 1002, and the conversion result is output to the compression sensing unit 1004. The random measurement matrix generation unit 1008 generates a random measurement matrix composed of random variables. The random restoration matrix generation unit 1010 generates a random restoration matrix by multiplying the random measurement matrix by the IFFT matrix. The compression sensing unit 1004 compresses the input digital signal using the random measurement matrix generated by the random measurement matrix generation unit 1008. The compression result is output to the signal restoration unit 1006. The signal restoration unit 1006 restores the input compressed data using a restoration algorithm such as a random restoration matrix and an L1-minimization restoration algorithm, and outputs the restored signal.

  The concept of conventional analog compressed sensing (see, for example, Non-Patent Document 3) is shown in FIG. FIG. 10 assumes a sparse signal in the frequency domain or time domain. First, an input signal is demultiplexed into a plurality of identical signals, and each demultiplexed signal is modulated using a modulator having a speed of the input signal Nyquist rate (twice the bandwidth of the input signal). This modulator has the characteristic that the sign changes randomly at the speed of the Nyquist rate. Further, in each modulator, the conversion pattern of positive and negative signs differs for each branch, but the period of each pattern is the same for each branch. For example, the example illustrated in FIG. 10 illustrates an example in which the number of branches is four and the cycle of positive / negative sign conversion is 100 for each branch.

  Next, the signal converted by the modulator of each branch is input to a low-pass filter unit (low pass filter). The output signal of the low-pass filter section for each branch is sampled using an AD converter having a lower speed than the Nyquist rate of the input signal. In the example shown in FIG. 10, analog signals corresponding to 100 digital signals in the conventional digital compression sensing are input, and four branches output 10 compressed data, so 40 signals from 100 signals. Pieces of compressed data are obtained. Here, the analog signal corresponding to 100 input signals means an analog signal that becomes 100 digital samples when AD conversion is performed by an AD converter having a Nyquist rate. Next, the compressed data is restored using a random restoration matrix calculated from the positive / negative conversion pattern for each branch and the response characteristics of the low-pass filter unit and a restoration algorithm such as an L1-minimization restoration algorithm.

  FIG. 11 shows a system configuration for performing conventional analog compressed sensing. In FIG. 11, an input signal is demultiplexed into N (N is a natural number) signals by a demultiplexing unit 2001. The demultiplexed signals are input to the modulation units 2002 of the first branch unit 201, the second branch unit 202, and the N-th branch unit 203, respectively. The positive / negative pattern generation unit 2008 generates a positive / negative pattern composed of random variables. The modulation unit 2002 modulates the signal input from the demultiplexing unit 2001 using the positive / negative conversion pattern generated by the positive / negative pattern generation unit 2008. In each branch, the output signal of the modulation unit 2002 is input to the low-pass filter unit 2003.

  Next, in the AD conversion unit 2004 of each of the first to N-th branch units 201 to 203, the output signal of the low-pass filter unit 2003 for each branch is converted from an analog signal to a digital signal, and the conversion result is sent to the coupling unit 2005. Output. The combining unit 2005 combines the output signals of the AD conversion units 2004 for the first to N-th branch units 201 to 203 and outputs the combined signals to the signal restoration unit 2006. The random restoration matrix generation unit 2010 multiplies the IFFT matrix by a random measurement matrix composed of the positive / negative conversion pattern generated by the positive / negative pattern generation unit 2008 and the response characteristic of the low-pass filter unit for each branch to generate a random restoration matrix. The signal restoration unit 2006 performs signal restoration on the compressed data input from the combining unit 2005 using a restoration algorithm such as a random restoration matrix and an L1-minimization restoration algorithm, and outputs the restored signal.

  Compared with FIG. 8 and FIG. 10, in the conventional digital compression sensing of FIG. 8, 100 signals are compressed to 40 data after AD conversion of the input signal, but the conventional analog of FIG. In compression sensing, the input signal is compressed in the analog domain before AD conversion. As a result, the compression of 100 signals to 40 compressed data is the same as conventional digital compression sensing. It is.

12 and 13 represent digital representations of the measurement matrix of the example of the conventional analog compressed sensing shown in FIG. First, FIG. 12 shows a digital representation of the measurement matrix at branch i (i is a natural number). Here, the patterns of the positive and negative signs of the branch i are represented as bi _{, 0} , bi _{, 1} , ..., bi _{, 99} , respectively. Each b _{i, j} is +1 if the pattern generated by the positive / negative pattern generation unit is positive, and is −1 if the pattern generated by the positive / negative pattern generation unit is negative. In each branch, an input signal is converted into positive and negative in an analog domain by using a positive / negative conversion pattern in an analog domain. When this is digitally expressed, bi _{, 0} , bi _{, 1} ,..., Bi _{, 99} can be expressed as a 100 × 100 matrix having diagonal components.

  Next, the response characteristics of the low-pass filter unit 2003 and the AD conversion unit 2004 in each branch can be expressed as a 10 × 100 matrix as shown in FIG. By multiplying this matrix by the above-mentioned 100 × 100 positive / negative pattern matrix, a digital representation of the measurement matrix of branch i representing the relationship between the input signal and output signal of branch i can be obtained. Considering that the output signals of the branches are combined by the combining unit 2005, the final digital representation of the measurement matrix can be obtained by combining the measurement matrices of the branches as shown in FIG. In FIG. 13, the measurement matrix size of each branch is 10 × 100, and the outputs of the four branches are combined, so the final measurement matrix size is 40 × 100.

FIG. 14 is a diagram showing the relationship between the modulator speed and the positive / negative pattern period of the conventional analog compression sensing, the Nyquist rate of the input signal, the pass band of the low-pass filter unit, and the speed of the AD converter. . In FIG. 14, the _Nyquist rate of the input signal and the positive / negative sign conversion speed of the modulation unit are expressed as f _Nyquist and T _modulation , respectively, and have a relationship of f _Nyquist = 1 / T _modulation . Incidentally, the pass band of the low pass filter, and the speed of the AD converter, the period of the sign pattern of the modulation unit, _{respectively,} and _{f LPF,} and _{f ADC,} expressed as _{T period, f} LPF ₌ f _ADC = 1 / T _period .

The speed (f _ADC ) of the AD converter of each branch, the number of branches (here, the number of branches is n (n is a natural number)), and the area where the actual signal is in the entire band (f _occupied ). The relationship needs to satisfy f _ADC × n ≧ λ × f _occupied . Here, it is known that λ is a value from about 3 to 4 (for example, Non-Patent Documents 1 and 2). Here, λ is set to 4 for simplicity. For example, when there is only a signal of 10 MHz in the entire band of 100 MHz, the multiplication (f _ADC × n) of the speed (f _ADC ) and the number of branches (n) of each AD converter needs to be 40 Msps. There is.

Next, the relationship between the speed (f _ADC ) of the AD converter in each branch and the number of branches (n) will be described. multiplication of f _ADC and n _are the 4 times _{f OCCUPIED, if f OCCUPIED} and _{f ADC} is given, the number of branches (n) becomes _{(4 × f occupied / f ADC} ). For example, in the case of the above example, since there is only a signal of 10 MHz in the entire band of 100 MHz, the data rate after compression needs to be 40 Msps. Here, for example, if the speed of the AD converter of each branch is 5 Msps, 8 branches are required (8 × 5 Msps = 40 Msps). Alternatively, if the AD converter speed of each branch is 10 Msps, four branches are required (4 × 10 Msps = 40 Msps). That is, the number of branches can be reduced by increasing the sampling rate of the AD converter for each branch.

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. M. Mishali et al., “From Theory to Practice: Sub-Nyquist Sampling of Sparse Wideband Analog Signals,” IEEE Journal of Selected Topics on Signal Processing, pp. 375-391, April 2010.

In the conventional analog compression sensing technology, the number of branches can be reduced by increasing the speed of the AD converter for each branch. By the way, as shown in the example of FIG. 14, the AD conversion speed (f _ADC ) for each branch has a relationship between the period (T _period ) and the reciprocal number (f _ADC = 1 / T _period ) of the positive / negative code pattern of the modulation unit. Therefore, when the speed of the AD conversion unit for each branch is increased, there is a problem that the cycle of the positive / negative code pattern of the modulation unit is shortened. For example, FIG. 15 shows an example in which f _ADC is doubled. By doubling the AD conversion speed (f ′ _ADC = 2 × f _ADC ) for each branch, the number of branches can be halved, but the period of the positive / negative code pattern of the modulator is halved. End up. When the cycle of the positive / negative code pattern is shortened, the randomness of the random measurement matrix is lost, which causes a problem that the restoration performance deteriorates.

  The present invention has been made in view of such circumstances, and an object of the present invention is to provide a signal processing system and a signal processing method that can reduce the number of branches without deteriorating the restoration performance in analog compression sensing.

  The present invention is a signal processing system that compresses and restores a signal having a sparse property in a frequency domain or a time domain, and includes a demultiplexing unit that demultiplexes an input signal into a plurality of identical signals, A positive / negative pattern generation unit that generates a positive / negative conversion pattern, a modulation unit that modulates a demultiplexed signal with the positive / negative conversion pattern, a low-pass filter unit that passes a low frequency component of the output of the modulation unit, and the low-pass filter unit The same number of branching parts as the demultiplexed signals composed of the AD converting part for AD converting the output of the band-pass filter part, the coupling part for combining the outputs of the AD converting parts, the positive / negative conversion pattern, and the low frequency band A random restoration matrix generating unit for generating a random restoration matrix from a random measurement matrix composed of response characteristics of a pass filter unit and an IFFT matrix; an output signal of the coupling unit; A signal restoration unit that restores the signal, and a signal detection unit that detects a signal component included in the entire band of the restoration signal output from the signal restoration unit and calculates a ratio between the total bandwidth and the signal bandwidth And the positive / negative pattern generation unit, the cycle of the positive / negative conversion pattern, the modulation rate of the modulation unit, and the AD conversion rate of the AD conversion unit based on the ratio between the total bandwidth and the signal bandwidth And a control unit.

  The present invention is a signal processing system composed of a central station and a remote station for compressing and restoring a signal having a sparse property in a frequency domain or a time domain, and the remote station receives a plurality of input signals. A demultiplexing unit for demultiplexing into the same signal, a positive / negative pattern generation unit for generating a random positive / negative conversion pattern, a modulation unit for modulating the demultiplexed signal with the positive / negative conversion pattern, and a low output of the modulation unit Combines the same number of branching parts as the demultiplexed signal consisting of a low-pass filter part that passes frequency components and an AD converter part that AD-converts the output of the low-pass filter, and the outputs of the AD converter parts And a coupling unit for transmitting to the central station, wherein the central station performs random recovery from the random measurement matrix and the IFFT matrix comprising the positive / negative conversion pattern and the response characteristics of the low-pass filter unit. A random restoration matrix generation unit for generating a matrix; a signal restoration unit for restoring a signal using the signal received from the remote station and the random restoration matrix; and within the entire band of the restoration signal output from the signal restoration unit A signal detection unit that detects a signal component included in the signal and calculates a ratio between a total bandwidth and a signal bandwidth, based on a ratio between the total bandwidth and the signal bandwidth, the positive / negative pattern generation unit, And a control unit that controls a cycle of the positive / negative conversion pattern, a modulation speed of the modulation unit, and an AD conversion speed of the AD conversion unit.

  In the present invention, the positive / negative pattern generation unit can change the cycle of the positive / negative conversion pattern and the number of positive / negative conversion patterns included in one cycle, and the modulation unit can change the modulation speed to the Nyquist rate of the input signal. The low-pass filter unit can change the passband, the AD conversion unit can change the AD conversion speed, and the control unit can match the cycle of the positive / negative conversion pattern. The number of positive and negative patterns included in a period, the modulation speed, the passband, and the AD conversion speed are determined.

In the present invention, the control unit includes a period of the positive / negative conversion pattern (T _period ), a number of positive / negative patterns included in one period (T _period / T _modulation ), and the modulation speed (1 / T _modulation ). When the passband (f _LPF ) and the AD conversion speed (f _ADC ) are determined, the number of demultiplexing is N (N is a natural number), and the modulation speed is the Nyquist rate of the input signal. Set to K (K is a natural number) times, λ is a coefficient for determining the amount of information that satisfies the requirements of the restoration performance from the detected signal bandwidth (f _occupied ), and a random code length L that can maintain randomness _{If, (N / L) × (} f Nyquist / f occupied) = a _{N × (f ADC / f occupied} ) ≧ λ, 1 / T modulation And K × _{f Nyquist,} and _{f ADC = f LPF = 1 /} T period, and the _{T period} / _{T modulation} ≧ L, characterized in that it as true, respectively.

  The present invention relates to a signal processing method performed by a signal processing system that compresses and decompresses a signal having a sparse property in a frequency domain or a time domain, and demultiplexes an input signal into a plurality of identical signals. A positive / negative pattern generation step for generating a random positive / negative conversion pattern, a modulation step for modulating the demultiplexed signal with the positive / negative conversion pattern, and a low-pass filtering for passing a low frequency component of the output of the modulation step A step of performing a branching step consisting of a step and an AD conversion step for AD converting the output of the low-pass filtering step in parallel by the number of demultiplexes, and a combining step of combining the outputs of the AD conversion steps And a random characteristic consisting of response characteristics of the positive / negative conversion pattern and the low-pass filtering step A random restoration matrix generation step for generating a random restoration matrix from a constant matrix and an IFFT matrix, a signal restoration step for restoring a signal using the output signal of the combination step and the random restoration matrix, and the signal restoration step. A signal detection step for detecting a signal component included in the entire band of the restored signal and calculating a ratio between the total bandwidth and the signal bandwidth, and based on the ratio between the total bandwidth and the signal bandwidth, The method includes a positive / negative pattern generation step, a cycle of the positive / negative conversion pattern, a modulation speed of the modulation step, and a control step of controlling an AD conversion speed of the AD conversion step.

  The present invention is a signal processing method performed by a signal processing system composed of a central station and a remote station that compresses and decompresses a signal having a sparse property in the frequency domain or in the time domain. A demultiplexing step for demultiplexing the input signal into a plurality of identical signals, a positive / negative pattern generation step for the remote station to generate a random positive / negative conversion pattern, and a modulation for modulating the demultiplexed signal with the positive / negative conversion pattern A branch processing step consisting of a step, a low-pass filtering step for passing a low-frequency component of the output of the modulation step, and an AD conversion step for AD-converting the output of the low-pass filtering step by the number of the demultiplexing The step performed in parallel and the remote station combine the outputs of the AD conversion steps and send them to the central station. A combining step, a random restoration matrix generating step for generating a random restoration matrix from a random measurement matrix and an IFFT matrix consisting of response characteristics of the positive / negative conversion pattern and the low-pass filtering step, and the central station comprising: A signal restoration step for restoring a signal using the signal received from the remote station and the random restoration matrix, and a signal component included in the entire band of the restoration signal output by the central station in the signal restoration step A signal detection step for calculating a ratio between the total bandwidth and the signal bandwidth, a positive / negative pattern generation step based on a ratio between the total bandwidth and the signal bandwidth, and a cycle of the positive / negative conversion pattern And a control step for controlling the modulation speed of the modulation step and the AD conversion speed of the AD conversion step. It is characterized in.

  In the present invention, the positive / negative pattern generation step can change the cycle of the positive / negative conversion pattern and the number of positive / negative conversion patterns included in one cycle, and the modulation step can change the modulation speed to the Nyquist rate of the input signal. The low-pass filtering step can change the passband, the AD conversion step can change the AD conversion speed, and the control step can be set to the cycle of the positive / negative conversion pattern. The number of positive and negative patterns included in a period, the modulation speed, the passband, and the AD conversion speed are determined.

In the present invention, the control step includes the period of the positive / negative conversion pattern (T _period ), the number of positive / negative patterns included in one period (T _period / T _modulation ), and the modulation speed (1 / T _modulation ). When the passband (f _LPF ) and the AD conversion speed (f _ADC ) are determined, the number of demultiplexing is N (N is a natural number), and the modulation speed is the Nyquist rate of the input signal. Set to K (K is a natural number) times, λ is a coefficient for determining the amount of information that satisfies the requirements of the restoration performance from the detected signal bandwidth (f _occupied ), and a random code length L that can maintain randomness (N / L) × (f _Nyquist / f _occupied ) = N × (f _ADC / f _occupied ) ≧ λ, and 1 / T _{modulati It is characterized in that on} = K × f _Nyquist , f _ADC = f _LPF = 1 / T _period , and T _period / T _modulation ≧ L.

  According to the present invention, the restoration performance is improved in analog compressed sensing by increasing the AD conversion speed and modulating the modulation speed of the modulation section by a constant multiple (for example, 2 times, 3 times, etc.) faster than Nyquist of the input signal. The effect that the number of branches can be reduced without degrading the is obtained.

It is a block diagram which shows the structure of the 1st Embodiment of this invention. It is a flowchart which shows operation | movement of the signal processing system shown in FIG. It is a block diagram which shows the structure of the 2nd Embodiment of this invention. FIG. 4 is a block diagram showing configurations of a remote station 1 and a central station 10 shown in FIG. 3. 5 is a flowchart showing the operation of the remote station shown in FIG. 5 is a flowchart showing the operation of the central station shown in FIG. It is a figure which shows the relationship of each parameter of this invention. It is a figure which shows the concept of the conventional digital compression sensing. It is a figure which shows the system configuration | structure which performs the conventional digital compression sensing. It is a figure which shows the concept of the conventional analog compression sensing. It is a figure which shows the system configuration | structure which performs the conventional analog compression sensing. It is a figure which shows the digital representation of the measurement matrix in the branch i of the example of the conventional analog compression sensing. It is a figure which shows the digital representation of the whole measurement matrix of the example of the conventional analog compression sensing. It is a figure which shows the relationship of each parameter of the conventional analog compression sensing. It is a figure which shows the relationship of each parameter of the conventional analog compression sensing at the time of double AD conversion speed from FIG.

<First Embodiment>
Hereinafter, a signal processing system according to a first embodiment of the present invention will be described with reference to the drawings. FIG. 1 is a block diagram showing the configuration of the embodiment. In this figure, the same parts as those of the conventional signal processing system shown in FIG. The signal processing system shown in this figure is different from the conventional signal processing system in that a signal detection unit 2011 and a control unit 2012 are provided.

  In FIG. 1, a demultiplexing unit 2001 receives a received signal, demultiplexes it into N (N is a natural number) signals having the same signal characteristics as the input signal, and each of these N signals is a first signal. To the modulation unit 2002 of the N-th branch units 201 to 203.

  The positive / negative pattern generation unit 2008 of the first to N-th branch units 201 to 203 uses a positive / negative pattern cycle input from the control unit 2012 and a random number generation seed to generate a random code having the number of cycles of the positive / negative pattern. Is generated. Here, the random code is a code having a value of +1 or -1. The generated positive / negative pattern is output to the modulation unit 2002 of the first to Nth branch units 201 to 203.

  The modulation unit 2002 sets the modulation rate of the modulator using the modulation rate setting parameter input from the control unit 2012. The modulation speed has a feature that is a constant multiple of Nyquist rate of the input signal of the demultiplexing unit 2001 (for example, 1 time, 2 times, 3 times, etc. of Nyquist rate). The modulation unit 2002 modulates the signal output from the demultiplexing unit 2001 using the positive / negative pattern generated by the positive / negative pattern generation unit 2008, and the modulated signal passes through the low-pass of the first to Nth branch units 201-203. The data is output to the filter unit 2003.

  In the modulation unit 2002 of the first to N-th branch units 201 to 203, modulation means that an input signal to each modulation unit 2002 is output as it is if the sign of the positive / negative pattern is +1, and the sign of the positive / negative pattern is − If it is 1, it indicates an operation of converting the phase by 180 degrees and outputting it.

  Each low-pass filter unit 2003 has a variable passband setting, and sets a passband using a passband setting parameter input from the control unit 2012. Each low-pass filter unit 2003 receives the output signal of each modulation unit 2002, performs low-pass filtering, and outputs the output to each AD conversion unit 2004.

  Each AD conversion unit 2004 sets an AD conversion speed using the AD conversion speed setting parameter input from the control unit 2012. In the AD conversion unit 2004, each low-pass filter unit 2003 receives the output signal, performs analog-digital conversion on the input signal, and outputs the result to the combining unit 2005.

  The combining unit 2005 combines the output signal of each AD conversion unit 2004 for each number of compressed data for each branch using the number of compressed data for each branch input from the control unit 2012, and performs signal restoration on the combined data. Output to the unit 2006. For example, when the number of branches is 4 (N = 4) and the number of compressed data per branch is 10, the combining unit 2005 divides the output of each AD conversion unit 2004 into 10 units and combines them. For each combination, 40 pieces of compressed data are output to the signal restoration unit 2006.

  The random restoration matrix generation unit 2010 uses the period of the positive / negative pattern input from the control unit 2012, the random number generation seed, the response characteristic of each low-pass filter unit 2003, and the response characteristic of each AD conversion unit 2004. As shown in the examples of FIGS. 12 and 13, a random measurement matrix is generated. Here, the positive and negative pattern period and the random number generation seed use the same values as those used by the positive and negative pattern generation units 2008. The random restoration matrix generation unit 2010 multiplies the generated random measurement matrix and the IFFT matrix to generate a random restoration matrix. Using the generated random restoration matrix and a restoration algorithm such as the L1-minimization restoration algorithm, the output signal of the combining unit 2005 is restored.

  The signal detection unit 2011 detects a signal from the restoration waveform input from the signal restoration unit 2006, obtains a total ratio of the total bandwidth and the detected signal bandwidth, and outputs it to the control unit 2012. As a signal detection method, for example, Fourier transform is performed on the reconstructed waveform, power spectral density is obtained, and each frequency component is compared with a threshold Th, so that the band of the signal included in the entire band from the frequency component exceeding the threshold Th is obtained. To detect. Further, after integration for each band, the threshold value Th may be compared. Furthermore, the signal bandwidth may be acquired by detecting and identifying the signal by a matched filter, periodic steadiness analysis, or the like and referring to the database.

  The control unit 2012 determines each parameter based on the ratio between the total bandwidth input from the signal detection unit 2011 and the detected signal bandwidth. The control unit 2012 determines the cycle of the positive / negative pattern and the random number generation seed for each branch, and outputs it to the positive / negative pattern generation unit 2008 for each branch. Here, the cycle of the positive / negative pattern is the same for each branch, but since the random number generation seed is different for each branch, each positive / negative pattern generation unit 2008 generates a different positive / negative pattern with the same cycle. In addition, the control unit 2012 determines a modulation speed setting parameter and outputs it to each modulation unit 2002. The control unit 2012 also determines a passband setting parameter and outputs it to each low-pass filter unit 2003. The control unit 2012 also determines an AD conversion speed setting parameter and outputs it to each AD conversion unit 2004. Further, the control unit 2012 determines the number of compressed data for each branch and outputs it to the combining unit 2005. Further, the control unit 2012 outputs the cycle of the positive / negative pattern and the random number generation seed for each branch to the random restoration matrix generation unit 2010.

Here, the modulation speed setting parameter (1 / T _modulation ), the passband setting parameter (f _LPF ), the AD conversion speed setting parameter (f _ADC ), and the cycle of the positive / negative pattern (T _period ) are as described above. It is determined in consideration of the _{Nyquist rate} (f _Nyquist ), the band (f _occuped ) of the area where the actual signal is present in the target band of the input signal, and the number of branches (N). For example, when the modulation speed is K times the Nyquist rate of the input signal, 1 / T _modulation = K × f _Nquiquist , N × f _ADC = λ × f _occupied , and f _ADC = f _LPF = 1 / T _period Each parameter is determined so that each holds. Here, the number of positive / negative codes in the cycle of the positive / negative pattern is T _period / T _modulation , and if the random code length that can maintain randomness is L, it is necessary to satisfy the relational expression of T _period / T _modulation ≧ L. There is. λ is a coefficient that determines the amount of information necessary for restoring the band of the region where the signal actually exists in the target band of the input signal with respect to the Nyquist rate of the input signal.

  Next, the operation of the signal processing system shown in FIG. 1 will be described with reference to FIG. FIG. 2 is a flowchart showing the operation of the signal processing system shown in FIG. First, the demultiplexing unit 2001 receives the received signal, demultiplexes the received signal into N signals having the same signal characteristics as the input signal, and the N signals are first to N-th branch units 201 to 201, respectively. It outputs to the modulation | alteration part 2002 of 203 (step S11).

  Next, the control unit 2012 determines the cycle of the positive / negative pattern and the random number generation seed for each branch, and outputs it to each positive / negative pattern generation unit 2008. Here, the cycle of the positive / negative pattern is the same for each branch, but since the random number generation seed is different for each branch, the positive / negative pattern generation unit 2008 of the first to Nth branch units 201 to 203 has the same cycle. A positive / negative pattern is generated. In addition, the control unit 2012 determines a modulation speed setting parameter and outputs it to each modulation unit 2002. Further, a passband setting parameter is determined and output to the low pass filter unit 2003 of each branch. Also, an AD conversion speed setting parameter is determined and output to the AD conversion unit 2004 of each branch. In addition, the number of compressed data for each branch is determined and output to the combining unit 2005. Further, the cycle of the positive / negative pattern and the random number generation seed for each branch are output to the random restoration matrix generation unit 2010 (step S12).

  Next, each positive / negative pattern generation unit 2008 generates a random code having the number of periods of the positive / negative pattern by using the positive / negative pattern cycle input from the control unit 2012 and the random number generation seed. Here, the random code is a code having a value of +1 or -1. The generated positive / negative pattern is output to each modulation unit 2002 (step S13).

  Next, each modulation unit 2002 sets the modulation rate of the modulator using the modulation rate setting parameter input from the control unit 2012. The modulation speed has a feature that is a constant multiple of Nyquist rate of the input signal of the demultiplexing unit 2001 (for example, 1 time, 2 times, 3 times, etc. of Nyquist rate). The modulation unit 2002 modulates the signal output from the demultiplexing unit 2001 using the positive / negative pattern generated by each positive / negative pattern generation unit 2008, and outputs the modulated signal to each low-pass filter unit 2003 (step S14). ).

  Next, each low-pass filter unit 2003 has a variable pass band setting, and sets the pass band using the pass band setting parameter input from the control unit 2012. The low-pass filter unit 2003 receives the output signal of each modulation unit 2002, performs low-pass filtering, and outputs the output to each AD conversion unit 2004 (step S15).

  Next, each AD conversion unit 2004 sets the AD conversion speed using the AD conversion speed setting parameter input from the control unit 2012. The AD conversion unit 2004 receives the output signal of each low-pass filter unit 2003, performs analog-digital conversion of the input signal, and outputs the result to the combining unit 2005 (step S16).

  Next, the combining unit 2005 combines the output signal of each AD conversion unit 2004 for each number of compressed data for each branch using the number of compressed data for each branch input from the control unit 2012, and outputs the combined data as a signal. The data is output to the restoration unit 2006. For example, when the number of branches is 4 and the number of compressed data per branch is 10, the combining unit 2005 divides the output of each AD conversion unit into 10 units, combines them, and performs one combination. Then, 40 pieces of compressed data are output to the signal restoration unit 2006 (step S17).

  Next, the random restoration matrix generation unit 2010 obtains the cycle of the positive / negative pattern input from the control unit 2012, the random number generation seed, the response characteristics of each low-pass filter unit 2003, and the response characteristics of each AD conversion unit 2004. Used to generate a random measurement matrix as shown in the examples of FIGS. Here, the positive and negative pattern period and the random number generation seed use the same values as those used by the positive and negative pattern generation units 2008. The random restoration matrix generation unit 2010 generates a random restoration matrix by multiplying the generated random measurement matrix and the IFFT matrix (step S18).

  Next, the signal restoration unit 2006 uses the generated random restoration matrix and a restoration algorithm such as an L1-minimization restoration algorithm to restore the output signal of the combining unit 2005 (step S19).

  Next, the signal detection unit 2011 detects a signal from the restoration waveform input from the signal restoration unit 2006, obtains the total ratio of the total bandwidth and the detected signal bandwidth, and outputs it to the control unit 2012 (step S20). . As a signal detection method, for example, Fourier transform is performed on the reconstructed waveform, power spectral density is obtained, and each frequency component is compared with a threshold Th, so that the band of the signal included in the entire band from the frequency component exceeding the threshold Th is obtained. To detect. Further, after integration for each band, the threshold value Th may be compared. Furthermore, the signal bandwidth may be acquired by detecting and identifying the signal by a matched filter, periodic steadiness analysis, or the like and referring to the database.

<Second Embodiment>
Next, a signal processing system according to a second embodiment of the present invention will be described. FIG. 3 is a block diagram showing the configuration of the embodiment. The communication system shown in FIG. 3 includes one or more remote stations 1 to 3 and a central station 10, and the remote stations 1 to 3 and the central station 10 are connected by a transmission path. The signal processing system according to the second embodiment is obtained by applying the signal processing system of the present invention to the communication system shown in FIG.

  4 is a block diagram showing the configuration of the remote station 1 and the central station 10 shown in FIG. The control unit 3012 determines the cycle of the positive / negative pattern and the random number generation seed for each branch, and sends it to the positive / negative pattern generation unit 3008 of each branch via the transmission path. Here, the cycle of the positive / negative pattern is the same for each branch, but since the random number generation seed is different for each branch, each positive / negative pattern generation unit 3008 generates different positive / negative patterns with the same cycle.

  In addition, the control unit 3012 determines a modulation speed setting parameter and outputs it to each modulation unit 3002. In addition, the control unit 3012 determines a passband setting parameter and outputs it to each low-pass filter unit 3003. Also, the control unit 3012 determines an AD conversion speed setting parameter and outputs it to each AD conversion unit 3004. Also, the control unit 3012 determines the number of compressed data for each branch, and outputs it to the combining unit 3005. Further, the control unit 3012 outputs the cycle of the positive / negative pattern and the random number generation seed for each branch to the random restoration matrix generation unit 3010.

  The demultiplexing unit 3001 receives the received signal, demultiplexes the received signal into N signals having the same signal characteristics as the input signal, and the N signals are first to Nth branch units 301 to 303, respectively. Are output to the modulation unit 3002.

  Each positive / negative pattern generation unit 3008 generates a random code having the number of positive / negative pattern periods by using the positive / negative pattern period input from the control unit 3012 via the transmission path and the random number generation seed. Here, the random code is a code having a value of +1 or -1. The positive / negative pattern generation unit 3008 outputs the generated positive / negative pattern to each modulation unit 3002.

  Each modulation unit 3002 sets the modulation rate of the modulator using the modulation rate setting parameter input from the control unit 3012 via the transmission path. The modulation speed has a feature that is a constant multiple of the Nyquist rate of the input signal of the demultiplexing unit 3001 (for example, 1 time, 2 times, 3 times, etc.). Modulating section 3002 modulates the signal output from demultiplexing section 3001 using the positive / negative pattern generated by each positive / negative pattern generating section 3008 and outputs the modulated signal to each low-pass filter section 3003.

  In each modulation section 3002, modulation means that an input signal to each modulation section 3002 is output as it is if the sign of the positive / negative pattern is +1, and the phase is converted by 180 degrees if the sign of the positive / negative pattern is -1. Output.

  Each low-pass filter unit 3003 has a variable passband setting, and sets a passband using a passband setting parameter input from the band control unit 3012 via the transmission path. The low-pass filter unit 3003 receives the output signal of each modulation unit 3002, performs low-pass filtering, and outputs the output to each AD conversion unit 3004.

  Each AD conversion unit 3004 sets an AD conversion speed using an AD conversion speed setting parameter input from the control unit 3012 via the transmission path. The AD conversion unit 3004 receives the output signal of each low-pass filter unit 3003 as input, performs analog-digital conversion on the input signal, and outputs the result to the combining unit 3005.

  The combining unit 3005 combines the output signal of each AD conversion unit 3004 for each number of compressed data for each branch, using the number of compressed data for each branch input from the control unit 3012 via the transmission path. Data is output to the signal restoration unit 3006 via the transmission line.

  The random restoration matrix generation unit 3010 uses the period of the positive / negative pattern input from the control unit 3012, the random number generation seed, the response characteristic of each low-pass filter unit 3003, and the response characteristic of each AD conversion unit 3004, A random measurement matrix is generated as in the examples of FIGS. Here, the positive and negative pattern period and random number generation seed use the same values as those used by the positive and negative pattern generation units 3008. The random restoration matrix generation unit 3010 generates a random restoration matrix by multiplying the generated random measurement matrix and the IFFT matrix. Using the generated random recovery matrix and a recovery algorithm such as the L1-minimization recovery algorithm, the signal input from the coupling unit 3005 via the transmission path is recovered.

  The signal detection unit 3011 detects a signal from the restoration waveform input from the signal restoration unit 3006, obtains the total ratio of the total bandwidth and the detected signal bandwidth, and outputs it to the control unit 3012. As a signal detection method, for example, Fourier transform is performed on the reconstructed waveform, power spectral density is obtained, and each frequency component is compared with a threshold Th, so that the band of the signal included in the entire band from the frequency component exceeding the threshold Th is obtained. To detect. Further, after integration for each band, the threshold value Th may be compared. Furthermore, the signal bandwidth may be acquired by detecting and identifying the signal by a matched filter, periodic steadiness analysis, or the like and referring to the database.

  Next, the operation of the signal processing system shown in FIG. 4 will be described with reference to FIGS. 5 and 6 are flowcharts showing the operation of the signal processing system shown in FIG. 4, FIG. 5 shows the operation of the remote station 1, and FIG. 6 shows the operation of the central station 10. First, the operation of the remote station 1 will be described with reference to FIG.

  First, the modulation unit 3002 AD-converts the modulation speed setting parameter, the positive / negative pattern generation unit 3008 converts the cycle of the positive / negative pattern and the random number generation seed for each branch, and the low-pass filter unit 3003 converts the passband setting parameter. The unit 3004 receives the AD conversion speed setting parameter, and the combining unit 3005 receives the number of compressed data for each branch from the control unit 3012 of the central station via the transmission path (step S211).

  Next, the demultiplexing unit 3001 receives the received signal, demultiplexes the received signal into N signals having the same signal characteristics as the input signal, and the N signals are first to Nth branch units 301, respectively. To 303 modulation section 3002 (step S212).

  Next, each positive / negative pattern generation unit 3008 generates a random code having the number of positive / negative pattern periods by using the positive / negative pattern period input from the control unit 3012 via the transmission path and the random number generation seed. . Here, the random code is a code having a value of +1 or -1. Each positive / negative pattern generation unit 3008 outputs the generated positive / negative pattern to each modulation unit 3002 (step S213).

  Next, each modulation unit 3002 sets the modulation rate of the modulator using the modulation rate setting parameter input from the control unit 3012 via the transmission path. The modulation speed has a characteristic that is a constant multiple of the Nyquist rate of the input signal of the demultiplexing unit 3001 (for example, 1 time, 2 times, 3 times, etc. of the Nyquist rate). The modulation unit 3002 modulates the signal output from the demultiplexing unit 3001 using the positive / negative pattern generated by each positive / negative pattern generation unit 3008, and outputs the modulated signal to each low-pass filter unit 3003 (step S214). ).

  Next, each low-pass filter unit 3003 has a variable passband setting, and sets a passband using a passband setting parameter input from the band control unit 3012 via the transmission path. The low-pass filter unit 3003 receives the output signal of each modulation unit 3002, performs low-pass filtering, and outputs the output to each AD conversion unit 3004 (step S215).

  Next, each AD conversion unit 3004 sets the AD conversion speed using the AD conversion speed setting parameter input from the control unit 3012 via the transmission path. In the AD conversion unit 3004, each low-pass filter unit 3003 receives the output signal, performs analog-digital conversion on the input signal, and outputs the result to the combining unit 3005 (step S216).

  Next, the combining unit 3005 combines the output signal of each AD conversion unit 3004 for each compressed data number for each branch, using the compressed data number for each branch input from the control unit 3012 via the transmission path. (Step S217), the combined data is transmitted to the restoration unit 3006 via the transmission path (Step S218).

  Next, the operation of the central office 10 will be described with reference to FIG. First, the control unit 3012 determines the cycle of the positive / negative pattern and the random number generation seed for each branch, and sends it to the positive / negative pattern generation unit 3008 via the transmission path. In addition, the control unit 3012 determines a modulation speed setting parameter, determines the passband setting parameter for each modulation unit 3002 via the transmission line, and passes it to each low-pass filter via the transmission line. In the unit 3003, an AD conversion speed setting parameter is determined, and the compressed data number for each branch is determined in each AD conversion unit 3004 via the transmission line, and the compressed data number for each branch is determined in the coupling unit 3005 via the transmission line. The positive and negative pattern periods and the random number generation seed for each branch are sent to the random restoration matrix generation unit 3010 (steps S221 and S222).

  Next, the signal restoration unit 3006 receives data transmitted by the coupling unit 3005 of the remote station 1 via the transmission path (step 223).

  Next, the random restoration matrix generation unit 3010 calculates the cycle of the positive / negative pattern input from the control unit 3012, the random number generation seed, the response characteristics of each low-pass filter unit 3003, and the response characteristics of each AD conversion unit 3004. Used to generate a random measurement matrix as in the example of FIGS. Here, the positive and negative pattern period and random number generation seed use the same values as those used by the positive and negative pattern generation units 3008. The random restoration matrix generation unit 3010 generates a random restoration matrix by multiplying the generated random measurement matrix and the IFFT matrix (step S224).

  Next, the signal restoration unit 3006 restores the output signal of the combining unit 3005 using the generated random restoration matrix and a restoration algorithm such as an L1-minimization restoration algorithm (step S225).

  Next, the signal detection unit 3011 detects a signal from the reconstructed waveform input from the signal reconstructing unit 3006, obtains the total ratio of the total bandwidth and the detected signal bandwidth, and outputs it to the control unit 3012 (step S226). . As a signal detection method, for example, Fourier transform is performed on the reconstructed waveform, power spectral density is obtained, and each frequency component is compared with a threshold Th, so that the band of the signal included in the entire band from the frequency component exceeding the threshold Th is obtained. To detect. Further, after integration for each band, the threshold value Th may be compared. Furthermore, the signal bandwidth may be acquired by detecting and identifying the signal by a matched filter, periodic steadiness analysis, or the like and referring to the database.

  Thus, in analog compression sensing, by setting the modulation speed of the modulation unit to a constant multiple of the input Nyquist rate, it is possible to reduce the scale of the circuit while avoiding deterioration of the restoration performance.

  As described above, in the analog compressed sensing technology, in order to reduce the number of branches without degrading the restoration performance, the AD conversion speed is increased and the modulation speed of the modulation unit is increased by a constant multiple of the Nyquist rate of the input signal. It is possible to provide a signal processing system that does not reduce the number of positive and negative codes for each period of one positive and negative code pattern of the modulation unit. FIG. 7 is a diagram illustrating an example in which the modulation speed of the modulation unit is doubled from FIGS. 14 and 15. Along with the modulation speed of the modulation section, the positive / negative pattern generation speed in the positive / negative pattern generation section also doubles. In this example, the cycle of the positive / negative pattern is halved by doubling the AD converter speed and reducing the number of branches compared to FIG. 14, but the positive / negative pattern generation speed in the positive / negative pattern generation unit is doubled. Therefore, the number of positive and negative signs in one cycle is the same as in the example of FIG. That is, since the randomness of the random measurement matrix is not lost, the restoration performance is not deteriorated. Thus, even when the number of branches is reduced and the circuit scale is reduced, the number of positive / negative codes in the cycle of the positive / negative pattern is not shortened, and signal processing can be performed without deteriorating the restoration performance.

  The processing units shown in FIGS. 1 and 4 may be realized by 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 for realizing a part of the functions described above, and may be a program capable of realizing the functions described above in combination with a program already recorded in the computer system. It may be realized using hardware such as PLD (Programmable Logic Device) or FPGA (Field Programmable Gate Array).

  As mentioned above, although embodiment of this invention has been described with reference to drawings, the said embodiment is only the illustration of this invention, and it is clear that this invention is not limited to the said embodiment. is there. Therefore, additions, omissions, substitutions, and other modifications of the components may be made without departing from the technical idea and scope of the present invention.

  In analog compression sensing that performs AD conversion after compressing the signal in the analog domain, the circuit scale can be reduced without shortening the length of the random pattern, thus solving the degradation problem of restoration performance and the circuit. It can be applied to signal processing applications where it is essential to reduce the scale.

  2001, 3001 ... demultiplexing unit, 2002, 3002 ... modulation unit, 2003, 3003 ... low-pass filter unit, 2004, 3004 ... AD conversion unit, 2005, 3005 ... coupling unit, 2006, 3006 ... Signal restoration unit, 2008, 3008 ... Positive / negative pattern generation unit, 2010, 3010 ... Random restoration matrix generation unit, 2011, 3011 ... Signal detection unit, 2012, 3012 ... Control , 201, 301... 1st branch section, 202, 302... 2nd branch section, 203, 303... Nth branch section, 1, 2, 3. Central office

Claims (8)

A signal processing system that compresses and decompresses signals having a sparse property in the frequency domain or in the time domain,
A demultiplexing unit that demultiplexes the input signal into a plurality of identical signals;
A positive / negative pattern generation unit that generates a random positive / negative conversion pattern, a modulation unit that modulates a demultiplexed signal with the positive / negative conversion pattern, a low-pass filter unit that passes a low frequency component of the output of the modulation unit, The same number of branching units as the demultiplexed signal, and an AD conversion unit that AD converts the output of the low-pass filter unit;
A coupling unit coupling the outputs of the AD conversion units;
A random restoration matrix generation unit that generates a random restoration matrix from a random measurement matrix and an IFFT matrix composed of response characteristics of the positive / negative conversion pattern and the low-pass filter unit;
A signal restoration unit that performs signal restoration using the output signal of the combining unit and the random restoration matrix;
A signal detection unit that detects a signal component included in the entire band of the restored signal output from the signal restoration unit, and calculates a ratio between the total bandwidth and the signal bandwidth;
Based on the ratio between the total bandwidth and the signal bandwidth, the positive / negative pattern generation unit is controlled so as to generate the positive / negative conversion patterns having the same period for each branch unit, and the modulation speed of the modulation unit a signal processing system characterized by comprising said control unit for controlling the passband of the low-pass filter and the AD conversion section of the AD conversion speed and. A signal processing system composed of a central station and a remote station that compresses and decompresses signals having a sparse property in the frequency domain or in the time domain,
The remote station is
A demultiplexing unit that demultiplexes the input signal into a plurality of identical signals;
A positive / negative pattern generation unit that generates a random positive / negative conversion pattern, a modulation unit that modulates a demultiplexed signal with the positive / negative conversion pattern, a low-pass filter unit that passes a low frequency component of the output of the modulation unit, The same number of branching units as the demultiplexed signal, and an AD conversion unit that AD converts the output of the low-pass filter unit ;
A combination unit for combining the outputs of the AD conversion units and transmitting the combined output to the central office;
The central office is
A random restoration matrix generation unit that generates a random restoration matrix from a random measurement matrix and an IFFT matrix composed of response characteristics of the positive / negative conversion pattern and the low-pass filter unit;
A signal restoration unit that performs signal restoration using the signal received from the remote station and the random restoration matrix;
A signal detection unit that detects a signal component included in the entire band of the restored signal output from the signal restoration unit, and calculates a ratio between the total bandwidth and the signal bandwidth;
Based on the ratio between the total bandwidth and the signal bandwidth, the positive / negative pattern generation unit is controlled so as to generate the positive / negative conversion patterns having the same period for each branch unit, and the modulation speed of the modulation unit a signal processing system characterized by comprising said control unit for controlling the passband of the low-pass filter and the AD conversion section of the AD conversion speed and. The positive / negative pattern generation unit can change the number of positive / negative conversion patterns included in one cycle and the cycle of the positive / negative conversion pattern,
The modulation unit can change the modulation speed to a constant multiple of Nyquist rate of the input signal,
The low-pass filter unit may change the passband,
The AD conversion unit can change the AD conversion speed,
The control unit determines the period of the positive / negative conversion pattern and the number of positive / negative patterns included in one period, the modulation speed, the passband, and the AD conversion speed. Item 3. The signal processing system according to Item 1 or 2. Wherein the control unit includes a number of positive and negative patterns that are included in the period of the positive and negative conversion pattern _{(T period)} and single period _{(T period} / _{T modulation),} and the modulation rate (1 / _{T modulation),} the When determining the passband (f _LPF ) and the AD conversion speed (f _ADC ), the number of demultiplexing is N (N is a natural number), and the modulation speed is the _Nyquist rate (f _Nyquist ) of the input signal. Is set to K times (K is a natural number) times, λ is a coefficient for determining the amount of information satisfying the requirements for the restoration performance from the detected signal bandwidth (f _occupied ), and a random code length L that can maintain randomness If you want to
(N / L) × (f _Nyquist / f _occupied ) = N × (f _ADC / f _occupied ) ≧ λ
1 / T _modulation = K × f _Nyquist ,
f _ADC = f _LPF = 1 / T _period ,
_4. The signal processing system according to claim 1, wherein T _period / T _modulation ≧ L is established. 5. A signal processing method performed by a signal processing system that compresses and decompresses signals having a sparse property in the frequency domain or in the time domain,
A demultiplexing step for demultiplexing the input signal into a plurality of identical signals;
The same number of branch portions as the demultiplexed signal respectively generate a positive / negative pattern generation step for generating a random positive / negative conversion pattern, a modulation step for modulating the demultiplexed signal with the positive / negative conversion pattern, and an output of the modulation step a low pass filtering step of passing the low frequency components of a branch processing step of performing the output of the low pass filtering step by concurrent branch processing comprising the AD conversion step of AD conversion,
A combining unit that combines the outputs of the AD conversion steps;
A random restoration matrix generation unit, which generates a random restoration matrix from a random measurement matrix and an IFFT matrix composed of response characteristics of the positive / negative conversion pattern and the low-pass filtering step;
A signal restoration step , wherein a signal restoration unit performs signal restoration using the output signal of the combining step and the random restoration matrix;
A signal detection unit that detects a signal component included in the entire band of the restored signal output by the signal restoration step, and calculates a ratio between the total bandwidth and the signal bandwidth;
The control unit controls the positive / negative pattern generation step so as to generate the positive / negative conversion pattern having the same period for each branch unit based on the ratio between the total bandwidth and the signal bandwidth, and the modulation signal processing method, wherein the step of modulating the speed executes a control step for controlling the AD conversion speed of the AD conversion step and the pass band of the low-pass filtering step. A signal processing method performed by a signal processing system composed of a central station and a remote station that compresses and decompresses signals having a sparse property in the frequency domain or in the time domain,
In the remote station ,
A demultiplexing step for demultiplexing the input signal into a plurality of identical signals;
The same number of branch portions as the demultiplexed signal respectively generate a positive / negative pattern generation step for generating a random positive / negative conversion pattern, a modulation step for modulating the demultiplexed signal with the positive / negative conversion pattern, and an output of the modulation step a low pass filtering step of passing the low frequency components of a branch processing step of performing the output of the low pass filtering step by concurrent branch processing comprising the AD conversion step of AD conversion,
A combining unit that combines outputs of the AD conversion steps and transmits the combined outputs to the central station;
Run
In the central office ,
A random restoration matrix generation unit , which generates a random restoration matrix from a random measurement matrix and an IFFT matrix composed of response characteristics of the positive / negative conversion pattern and the low-pass filtering step;
A signal restoration step , wherein the signal restoration unit performs signal restoration using the signal received from the remote station and the random restoration matrix;
A signal detection unit that detects a signal component included in the entire band of the restored signal output by the signal restoration step, and calculates a ratio between the total bandwidth and the signal bandwidth;
The control unit controls the positive / negative pattern generation step so as to generate the positive / negative conversion pattern having the same period for each branch unit based on the ratio between the total bandwidth and the signal bandwidth, and the modulation signal processing method, wherein the step of modulating the speed executes a control step for controlling the AD conversion speed of the AD conversion step and the pass band of the low-pass filtering step. In the positive / negative pattern generation step , the positive / negative pattern generation unit of the branching unit can change the cycle of the positive / negative conversion pattern and the number of positive / negative conversion patterns included in one cycle,
In the modulation step , the modulation unit of the branch unit can change the modulation speed to a constant multiple of the Nyquist rate of the input signal,
In the low-pass filtering step , the low-pass filter unit of the branching unit can change the pass band,
In the AD conversion step , the AD conversion unit of the branching unit can change the AD conversion speed,
In the control step , the control unit determines the cycle of the positive / negative conversion pattern, the number of positive / negative patterns included in one cycle, the modulation speed, the passband, and the AD conversion speed. The signal processing method according to claim 5, wherein the signal processing method is a signal processing method. Wherein in the control step, the control section, the period of the positive and negative conversion pattern _{(T period)} and the number of positive and negative patterns that are contained in one period _{(T period} / _{T modulation),} the modulation rate (1 / When determining T _modulation ), the passband (f _LPF ), and the AD conversion speed (f _ADC ), the number of demultiplexing is N (N is a natural number), and the modulation speed is the input signal. Set to K times (K is a natural number) times the _Nyquist rate (f _Nyquist ) , the coefficient that determines the amount of information that satisfies the restoration performance requirements from the detected signal bandwidth (f _occupied ) is λ, and randomness is maintained When the random code length L is possible,
(N / L) × (f _Nyquist / f _occupied ) = N × (f _ADC / f _occupied ) ≧ λ
1 / T _modulation = K × f _Nyquist ,
f _ADC = f _LPF = 1 / T _period ,
_8. The signal processing method according to claim 5, wherein T _period / T _modulation ≧ L is established.

Images