CN105027519B - A kind of signal processing method and device - Google Patents

Abstract

Embodiments of the present invention provide a signal processing method and device, which relate to the field of communication technology, and can solve the problem of reduced signal reconstruction performance due to limited sparsity prior information, reduce or basically eliminate the deviation generated by the reconstructed signal, and improve the single bit rate. Practicality and accuracy of compressive sensing techniques. The method is as follows: by obtaining input parameters and encoding end data, after initializing the process parameters and related parameter values, performing an adaptive sparsity estimation operation, and after obtaining the target sparsity, reconstructing the signal according to the target sparsity to obtain the original Reconstruction of sparse signals. Embodiments of the present invention are used for signal processing in the scenario of single-bit compressed sensing technology.

Description

A signal processing method and device

technical field

The present invention relates to the field of communication technology, in particular to a signal processing method and device.

Background technique

In traditional signal processing theory, according to Shannon sampling theorem, the acquisition rate should be at least twice the signal bandwidth to restore the original signal without distortion, but the analog-to-digital conversion and signal processing capabilities of existing hardware equipment can no longer meet the growing signal High-speed sampling is required, and massive data acquisition with high energy consumption is not essential. Therefore, the latest signal processing method uses Compressive Sensing (CS) technology, and the data collected at a lower sampling rate can still be accurately restored. original signal.

CS technology is based on the premise that the signal is sparse. After low-speed sampling of the input signal, the dimension-reduced sampling signal is obtained. The decoding process reconstructs the original input signal according to the dimension-reducing sampling signal. Reconstruct the original high-dimensional signal, significantly reducing signal acquisition overhead. In the application, the single-bit CS technology can be extended. After the input signal is sampled at a low speed, the sampling signal is quantized so that the input signal in the analog domain can be converted to the digital domain for subsequent processing, transmission, storage, etc., which can reduce hardware equipment. The complexity of analog-to-digital conversion, the complexity of information reception and acquisition, the amount of data transmitted and stored by the system, and the data robustness make CS technology more practical.

However, in the prior art, in the above-mentioned process of reconstructing the signal, it is necessary to predict the sparsity of the original signal and use it as an input parameter. The prior information of sparsity is usually limited. Even if the maximum sparsity is set through long-term statistical observation of the signal, there will be deviations from the actual sparsity, which will seriously affect the reconstruction performance of the signal. The obtained reconstruction The signal has a large deviation.

Contents of the invention

Embodiments of the present invention provide a signal processing method and device, which are used in the scenario of single-bit compressed sensing technology, which solves the problem of reduced signal reconstruction performance due to limited sparsity prior information, and reduces or basically eliminates the generation of reconstructed signals. The deviation improves the practicability and accuracy of single-bit compressed sensing technology.

In order to achieve the above object, embodiments of the present invention adopt the following technical solutions:

In a first aspect, a signal processing method is provided, the method comprising:

Acquiring input parameters and encoding end data, and initializing the process parameters and the first iteration signal, the encoding end data including the measurement signal obtained by the encoding end performing low-speed sampling processing and symbol quantization processing on the original sparse signal;

performing an adaptive sparsity estimation operation according to the input parameters, the encoding end data, the initialized process parameters, and the first iteration signal to obtain a target sparsity;

Perform signal reconstruction on the measurement signal according to the target sparsity to obtain a reconstructed signal of the original sparse signal.

With reference to the first aspect, in a first possible implementation manner, the acquiring input parameters and encoding end data includes:

Acquiring the input parameters, the input parameters include a maximum sparsity and a decision threshold;

Receive the encoding end data transmitted by the encoding end, the encoding end data includes the measurement signal and the sampling matrix; the measurement signal is the low speed sampling of the original sparse signal by the encoding end in combination with the sampling matrix After the sampling signal is acquired, the sampling signal is obtained by performing sign quantization processing on the sampling signal.

With reference to the first aspect or the first possible implementation manner of the first aspect, in a second possible implementation manner, the process parameters include a first estimated value and a first used value of the target sparsity, and the Initializing process parameters and first iteration signals involves:

assigning the first estimated value as the maximum sparsity, and assigning the first use value as the maximum sparsity;

Assigning the initial value of the first iteration signal as a zero vector.

With reference to any one of the first aspect to the second possible implementation manner of the first aspect, in a third possible implementation manner, the Performing an adaptive sparsity estimation operation on the process parameters and the first iteration signal, and obtaining the target sparsity includes:

According to the measurement signal and the sampling matrix, obtain the difference between the first iteration signal and the measurement signal after low-speed sampling and symbol quantization, and multiply the difference by the transpose matrix of the sampling matrix on the left , get the gradient of the first iteration;

Obtaining a process signal by a gradient descent method, where the process signal is a difference between the first iteration signal and the first iteration gradient;

Judging whether the first iterative gradient approaches a zero vector, and if the first iterative gradient approaches a zero vector, assigning the second used value of the target sparsity as the first estimated value; Or, if the first iterative gradient does not approach the zero vector, assigning the second use value as the first use value;

According to the process signal and the second use value, a second iterative signal is obtained through a threshold function, and the threshold function is used to reserve the second use value and element values with the largest element amplitude in the process signal, At the same time, all other element values in the process signal except the second use value elements with the largest element amplitude are set to zero, and the processing result of the threshold function is assigned to the second iteration signal;

Carrying out unit normalization on the second iterative signal to obtain a normalized iterative signal, counting the number of non-zero elements whose absolute value of the element amplitude in the normalized iterative signal exceeds the decision threshold, and calculating the Assigning the number of elements to the second estimated value of the target sparsity;

If the first iterative gradient does not approach the zero vector or the second estimated value is not equal to the first used value, assign the first estimated value to the second estimated value, the first estimated value A use value is assigned as the second use value, the first iteration signal is assigned as the second iteration signal, and the adaptive sparsity estimation operation is re-executed; or, if the gradient of the first iteration approaches If it is a zero vector and the second estimated value is equal to the first used value, assign the target sparsity as the second estimated value to obtain the target sparsity.

With reference to any one of the first aspect to the third possible implementation manner of the first aspect, in a fourth possible implementation manner, performing signal reconstruction on the measurement signal according to the target sparsity to obtain The reconstructed signal of the original sparse signal includes:

Perform a signal reconstruction operation according to the encoding end data, the first iteration signal, and the target sparsity, and the signal reconstruction operation specifically includes:

According to the measurement signal and the sampling matrix, obtain the difference between the first iteration signal and the measurement signal after low-speed sampling and symbol quantization, and multiply the difference by the transpose matrix of the sampling matrix on the left , get the gradient of the first iteration;

Obtaining a process signal by a gradient descent method, where the process signal is a difference between the first iteration signal and the first iteration gradient;

Combining the process signal and the target sparsity, a second iterative signal is obtained through a threshold function, the threshold function is used to retain the target sparsity element values with the largest element amplitude in the process signal, and at the same time In the process signal, all element values except the target sparsity elements with the largest element amplitude are set to zero, and the processing result of the threshold function is assigned to the second iteration signal;

According to the measurement signal, the sampling matrix and the second iteration signal, the Hamming distance is obtained through a non-zero statistical function; the non-zero statistical function is used to obtain the second iteration signal through low-speed sampling processing and the difference between the measurement signal after sign quantization processing and counting the number of non-zero elements in the difference, assigning the number of non-zero elements to the Hamming distance; the Hamming distance is the number of elements whose value is different from that of the measurement signal at the corresponding position of the second iterative signal after low-speed sampling processing and sign quantization processing;

If the Hamming distance is greater than a preset threshold value, assign the first iterative signal to the second iterative signal, and re-execute the signal reconstruction operation; or, if the Hamming distance is less than or equal to For the preset threshold value, unit normalization processing is performed on the second iteration signal to obtain the reconstructed signal of the original sparse signal.

In a second aspect, a signal processing device is provided, wherein the device includes:

A parameter acquisition unit, configured to acquire input parameters and encoding end data, and initialize process parameters and the first iteration signal. The encoding end data includes measurement signals acquired by the encoding end through low-speed sampling processing and symbol quantization processing on the original sparse signal ;

A sparsity estimation unit, configured to perform an adaptive sparsity estimation operation according to the input parameters, the encoding end data, the initialized process parameters, and the first iteration signal to obtain a target sparsity;

The signal reconstruction unit is configured to perform signal reconstruction on the measurement signal according to the target sparsity to obtain a reconstructed signal of the original sparse signal.

With reference to the second aspect, in a first possible implementation manner, the parameter acquisition unit is specifically configured to:

Acquiring the input parameters, the input parameters include a maximum sparsity and a decision threshold;

Receive the encoding end data transmitted by the encoding end, the encoding end data includes the measurement signal and the sampling matrix; the measurement signal is the low speed sampling of the original sparse signal by the encoding end in combination with the sampling matrix After the sampling signal is acquired, the sampling signal is obtained by performing sign quantization processing on the sampling signal.

With reference to the second aspect or the first possible implementation manner of the second aspect, in the second possible implementation manner, the process parameters include a first estimated value and a first used value of the target sparsity, and the The parameter acquisition unit also includes an initialization unit, which is specifically used for:

assigning the first estimated value as the maximum sparsity, and assigning the first use value as the maximum sparsity;

Assigning the initial value of the first iteration signal as a zero vector.

With reference to any one of the second aspect to the second possible implementation manner of the second aspect, in a third possible implementation manner, the sparsity estimation unit is configured to perform the adaptive sparsity estimation operation, Specifically include:

According to the measurement signal and the sampling matrix, obtain the difference between the first iteration signal and the measurement signal after low-speed sampling and symbol quantization, and multiply the difference by the transpose matrix of the sampling matrix on the left , get the gradient of the first iteration;

Obtaining a process signal by a gradient descent method, where the process signal is a difference between the first iteration signal and the first iteration gradient;

Judging whether the first iterative gradient approaches a zero vector, and if the first iterative gradient approaches a zero vector, assigning the second used value of the target sparsity as the first estimated value; Or, if the first iterative gradient does not approach the zero vector, assigning the second use value as the first use value;

According to the process signal and the second use value, a second iterative signal is obtained through a threshold function, and the threshold function is used to reserve the second use value and element values with the largest element amplitude in the process signal, At the same time, all other element values in the process signal except the second use value elements with the largest element amplitude are set to zero, and the processing result of the threshold function is assigned to the second iteration signal;

Carrying out unit normalization on the second iterative signal to obtain a normalized iterative signal, counting the number of non-zero elements whose absolute value of the element amplitude in the normalized iterative signal exceeds the decision threshold, and calculating the Assigning the number of elements to the second estimated value of the target sparsity;

If the first iterative gradient does not approach the zero vector or the second estimated value is not equal to the first used value, assign the first estimated value to the second estimated value, the first estimated value A use value is assigned as the second use value, the first iteration signal is assigned as the second iteration signal, and the adaptive sparsity estimation operation is re-executed; or, if the gradient of the first iteration approaches If it is a zero vector and the second estimated value is equal to the first used value, assign the target sparsity as the second estimated value to obtain the target sparsity.

With reference to any one of the second aspect to the third possible implementation manner of the second aspect, in a fourth possible implementation manner, the signal reconstruction unit is specifically configured to:

Perform a signal reconstruction operation according to the encoding end data, the first iteration signal, and the target sparsity, and the signal reconstruction operation specifically includes:

According to the measurement signal and the sampling matrix, obtain the difference between the first iteration signal and the measurement signal after low-speed sampling and symbol quantization, and multiply the difference by the transpose matrix of the sampling matrix on the left , get the gradient of the first iteration;

Obtaining a process signal by a gradient descent method, where the process signal is a difference between the first iteration signal and the first iteration gradient;

Combining the process signal and the target sparsity, a second iterative signal is obtained through a threshold function, the threshold function is used to retain the target sparsity element values with the largest element amplitude in the process signal, and at the same time In the process signal, all element values except the target sparsity elements with the largest element amplitude are set to zero, and the processing result of the threshold function is assigned to the second iteration signal;

According to the measurement signal, the sampling matrix and the second iteration signal, the Hamming distance is obtained through a non-zero statistical function; the non-zero statistical function is used to obtain the second iteration signal through low-speed sampling processing and the difference between the measurement signal after sign quantization processing and counting the number of non-zero elements in the difference, assigning the number of non-zero elements to the Hamming distance; the Hamming distance is the number of elements whose value is different from that of the measurement signal at the corresponding position of the second iterative signal after low-speed sampling processing and sign quantization processing;

If the Hamming distance is greater than a preset threshold value, assign the first iterative signal to the second iterative signal, and re-execute the signal reconstruction operation; or, if the Hamming distance is less than or equal to For the preset threshold value, unit normalization processing is performed on the second iteration signal to obtain the reconstructed signal of the original sparse signal.

In a third aspect, a signal processing device is provided, and the signal processing device includes: a bus, a processor connected to the bus, a memory, and an interface, wherein the interface is used for communicating with external devices; the memory uses In order to store instructions, the processor executes the instructions for:

Acquiring input parameters and encoding end data, and initializing the process parameters and the first iteration signal, the encoding end data including the measurement signal obtained by the encoding end performing low-speed sampling processing and symbol quantization processing on the original sparse signal;

performing an adaptive sparsity estimation operation according to the input parameters, the encoding end data, the initialized process parameters, and the first iteration signal to obtain a target sparsity;

Perform signal reconstruction on the measurement signal according to the target sparsity to obtain a reconstructed signal of the original sparse signal.

With reference to the third aspect, in a first possible implementation manner, the processor executes the instruction specifically for:

Acquiring the input parameters, the input parameters include a maximum sparsity and a decision threshold;

Receive the encoding end data transmitted by the encoding end, the encoding end data includes the measurement signal and the sampling matrix; the measurement signal is the low speed sampling of the original sparse signal by the encoding end in combination with the sampling matrix After the sampling signal is acquired, the sampling signal is obtained by performing sign quantization processing on the sampling signal.

With reference to the third aspect or the first possible implementation manner of the third aspect, in a second possible implementation manner, the processor executes the instruction specifically for:

assigning the first estimated value as the maximum sparsity, and assigning the first use value as the maximum sparsity;

Assigning the initial value of the first iteration signal as a zero vector.

With reference to any one of the third aspect to the second possible implementation manner of the third aspect, in a third possible implementation manner, the processor executing the instruction is further specifically used for:

According to the measurement signal and the sampling matrix, obtain the difference between the first iteration signal and the measurement signal after low-speed sampling and symbol quantization, and multiply the difference by the transpose matrix of the sampling matrix on the left , get the gradient of the first iteration;

Obtaining a process signal by a gradient descent method, where the process signal is a difference between the first iteration signal and the first iteration gradient;

Judging whether the first iterative gradient approaches a zero vector, and if the first iterative gradient approaches a zero vector, assigning the second used value of the target sparsity as the first estimated value; Or, if the first iterative gradient does not approach the zero vector, assigning the second use value as the first use value;

According to the process signal and the second use value, a second iterative signal is obtained through a threshold function, and the threshold function is used to reserve the second use value and element values with the largest element amplitude in the process signal, At the same time, all other element values in the process signal except the second use value elements with the largest element amplitude are set to zero, and the processing result of the threshold function is assigned to the second iteration signal;

Carrying out unit normalization on the second iterative signal to obtain a normalized iterative signal, counting the number of non-zero elements whose absolute value of the element amplitude in the normalized iterative signal exceeds the decision threshold, and calculating the Assigning the number of elements to the second estimated value of the target sparsity;

If the first iterative gradient does not approach the zero vector or the second estimated value is not equal to the first used value, assign the first estimated value to the second estimated value, the first estimated value A use value is assigned as the second use value, the first iteration signal is assigned as the second iteration signal, and the adaptive sparsity estimation operation is re-executed; or, if the gradient of the first iteration approaches If it is a zero vector and the second estimated value is equal to the first used value, assign the target sparsity as the second estimated value to obtain the target sparsity.

With reference to any one of the third aspect to the third possible implementation manner of the third aspect, in a fourth possible implementation manner, the processor executing the instruction is further specifically used for:

Perform a signal reconstruction operation according to the encoding end data, the first iteration signal, and the target sparsity, and the signal reconstruction operation specifically includes:

According to the measurement signal and the sampling matrix, obtain the difference between the first iteration signal and the measurement signal after low-speed sampling and symbol quantization, and multiply the difference by the transpose matrix of the sampling matrix on the left , get the gradient of the first iteration;

Obtaining a process signal by a gradient descent method, where the process signal is a difference between the first iteration signal and the first iteration gradient;

Combining the process signal and the target sparsity, a second iterative signal is obtained through a threshold function, the threshold function is used to retain the target sparsity element values with the largest element amplitude in the process signal, and at the same time In the process signal, all element values except the target sparsity elements with the largest element amplitude are set to zero, and the processing result of the threshold function is assigned to the second iteration signal;

According to the measurement signal, the sampling matrix and the second iteration signal, the Hamming distance is obtained through a non-zero statistical function; the non-zero statistical function is used to obtain the second iteration signal through low-speed sampling processing and the difference between the measurement signal after sign quantization processing and counting the number of non-zero elements in the difference, assigning the number of non-zero elements to the Hamming distance; the Hamming distance is the number of elements whose value is different from that of the measurement signal at the corresponding position of the second iterative signal after low-speed sampling processing and sign quantization processing;

If the Hamming distance is greater than a preset threshold value, assign the first iterative signal to the second iterative signal, and re-execute the signal reconstruction operation; or, if the Hamming distance is less than or equal to For the preset threshold value, unit normalization processing is performed on the second iteration signal to obtain the reconstructed signal of the original sparse signal.

Embodiments of the present invention provide a signal processing method and device. By acquiring input parameters and encoding end data, after initializing process parameters and related parameter values, an adaptive sparsity estimation operation is performed, and after acquiring the target sparsity , reconstruct the signal according to the target sparsity to obtain the reconstructed signal of the original sparse signal. In this way, the problem of signal reconstruction performance degradation due to limited sparsity prior information is solved, the deviation caused by reconstruction signal is reduced or basically eliminated, and the practicability and accuracy of single-bit compressed sensing technology are improved.

Description of drawings

In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following will briefly introduce the drawings that need to be used in the description of the embodiments or the prior art. Obviously, the accompanying drawings in the following description are only These are some embodiments of the present invention. Those skilled in the art can also obtain other drawings based on these drawings without creative work.

FIG. 1 is a schematic flowchart of a signal processing method provided by an embodiment of the present invention;

FIG. 2 is a first schematic flowchart of another signal processing method provided by an embodiment of the present invention;

FIG. 3 is a second schematic flow diagram of another signal processing method provided by an embodiment of the present invention;

FIG. 4 is a third schematic flowchart of another signal processing method provided by an embodiment of the present invention;

FIG. 5 is a first structural schematic diagram of a signal processing device provided by an embodiment of the present invention;

FIG. 6 is a second schematic structural diagram of a signal processing device provided by an embodiment of the present invention;

FIG. 7 is a schematic structural diagram III of a signal processing device provided by an embodiment of the present invention.

Detailed ways

The following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only some, not all, embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by persons of ordinary skill in the art without creative efforts fall within the protection scope of the present invention.

Embodiments of the present invention provide a signal processing method, as shown in FIG. 1, the method includes:

S101. Obtain input parameters and encoding end data, and initialize the process parameters and the first iteration signal;

Wherein, the input parameters include the maximum sparsity and the decision threshold; the data at the encoding end include the measurement signal and the sampling matrix; the process parameters include the first estimated value and the first use value of the target sparsity.

The maximum sparsity, decision threshold, and sampling matrix are predetermined calculation parameters related to target sparsity estimation; the first iteration signal, the first estimated value, and the first used value are intermediate variables involved in the calculation; the measurement signal is the encoding end pair The original sparse signal is obtained after low-speed sampling processing and symbol quantization processing.

It is worth mentioning that the original sparse signal can be video signal, audio signal, image processing signal, channel estimation signal, wireless sensor network signal, cognitive radio spectrum detection signal, etc., and the original sparse signal listed above is only exemplary, Including but not limited to this.

S102. Perform an adaptive sparsity estimation operation according to the input parameters, encoding end data, initialized process parameters, and the first iteration signal to obtain the target sparsity;

S103. Perform signal reconstruction on the measurement signal according to the target sparsity to obtain a reconstructed signal of the original sparse signal.

Embodiments of the present invention provide a signal processing method. By acquiring input parameters and encoding end data, after initializing process parameters and related parameter values, an adaptive sparsity estimation operation is performed, and after acquiring the target sparsity, according to The target sparsity is used to reconstruct the signal to obtain the reconstructed signal of the original sparse signal. In this way, the problem of signal reconstruction performance degradation due to limited sparsity prior information is solved, the deviation caused by reconstruction signal is reduced or basically eliminated, and the practicability and accuracy of single-bit compressed sensing technology are improved.

In order to enable those skilled in the art to more clearly understand the technical solutions provided by the embodiments of the present invention, another signal processing method provided by the embodiments of the present invention will be described in detail below through specific embodiments, as shown in FIG. 2 , the Methods include:

S201. Obtain input parameters and encoding end data.

Specifically, the acquired input parameters include the maximum sparsity Km and the decision threshold λ; wherein, the maximum sparsity Km and the decision threshold λ can be the default settings of the system, or can be given by external input, which are not limited here.

Receive the encoding end data transmitted by the encoding end. The encoding end data includes the measurement signal y and the sampling matrix Φ; where the measurement signal y is the low-speed sampling process of the original sparse signal θ at the encoding end, and the symbol quantization is performed on the obtained sampling signal x After processing, it contains symbol information that can be represented by a single bit, and the calculation formula is:

y=sign(Φθ),

Among them, θ is the original high-dimensional input signal, expressed as an N×1 column vector, and N represents the signal dimension; it is worth mentioning that, as a prerequisite for applying compressed sensing technology, the input signal θ is considered to be sparse, that is, the input signal Except for the only K non-zero elements in θ, the other elements are all zero, and the K value is much smaller than the signal dimension N. The sampling matrix Φ is expressed as an M×N matrix, and the number of rows M is smaller than the number of columns N. The input signal θ is multiplied by the sampling matrix Φ to the left, which can realize the function of low-speed sampling. The input signal θ is mapped from the original N-dimensional dimension reduction to M-dimensional, and the dimension-reduced sampling signal x=Φθ is obtained, and then the sampling signal Perform single-bit quantization on x to obtain measurement signal y. Specifically, sign() is a single-bit sign quantization function, which is used to perform single-bit sign quantization processing on sampled signal x, and quantize positive values in sampled signal x to 1 and negative values. Quantized to -1, and represented by bit 1 and bit 0 respectively, so that each measurement value in the measurement signal y can be represented by one bit.

Exemplarily, the encoding end data transmitted by the receiving encoding end may be the encoding end data transmitted by the decoding end receiving the encoding end, wherein the decoding end and the encoding end may be two independent devices, wired and/or wireless It can also be transmitted in a wired manner for two functional modules in one device, which is not limited here.

It is worth mentioning that the original sparse signal θ can be video signal, audio signal, image processing signal, channel estimation signal, wireless sensor network signal, cognitive radio spectrum detection signal, etc., and the original sparse signal listed above is only exemplary , including but not limited to.

S202. Initialize process parameters and initial values of the first iteration signal.

Specifically, the process parameters include a first estimated value pre_Est_S of the target sparsity K and a first usage value pre_Use_S. During initialization, both the first estimated value pre_Est_S and the first use value pre_Use_S are assigned as the maximum sparsity K _m in the input parameters, and the first iteration signal initial value of Assignment is a zero vector.

Among them, the above pair of process parameters and the first iteration signal The initial value during initialization is only exemplary, and other available initial values can also be selected for initialization according to actual applications, including but not limited to this.

S203. Perform an adaptive sparsity estimation operation according to the input parameters, the encoding end data, the initialized process parameters, and the first iteration signal to obtain a target sparsity.

Specifically, as shown in FIG. 3, performing an adaptive sparsity estimation operation includes:

S2031. Calculate and obtain the first iteration gradient.

Specifically, according to the measurement signal y and the sampling matrix Φ received in S201, and the first iteration signal Calculate the first iteration gradient corresponding to the current iteration process Among them, the first iteration signal It is the signal column vector obtained by updating the last iteration process. If the iteration process is performed for the first time, the first iteration signal That is, the initial value of the first iteration signal after initialization in S202

first iteration gradient In order to reflect the parameters of the iterative signal change rate in the iterative update process, the calculation formula is:

Specifically, combined with the sampling matrix Φ, in the first iterative signal After low-speed sampling processing and symbol quantization processing, the difference with the measurement signal y is obtained, and the difference is multiplied by the transpose matrix of the sampling matrix Φ to the left to obtain the first iteration gradient

S2032. Calculate and acquire the process signal.

Specifically, through the gradient descent method, according to the first iteration gradient with the first iteration signal Update the iterative intermediate process signal at along the direction of gradient _descent , the calculation formula is:

S2033. Determine whether the gradient of the first iteration approaches the zero vector.

Specifically, if the first iteration gradient tends to zero, execute S2034; or, if the gradient of the first iteration If it is not close to zero, execute S2035.

S2034. Update the second usage value to the first estimated value.

Specifically, the second use value cur_Use_S of the target sparsity K is used as a threshold of the threshold function in the subsequent process of updating the iterative signal.

If the first iteration gradient approaching zero, update the second use value cur_Use_S to the result of the sparsity estimation in the last iteration process, which is the first estimated value pre_Est_S of the target sparsity K; if this iteration process is performed for the first time, the second An estimated value pre_Est_S is equal to the maximum sparsity K _m given by the input parameters.

S2035. Update the second usage value to the first usage value.

Specifically, the second use value cur_Use_S of the target sparsity K is used as a threshold of the threshold function in the subsequent process of updating the iterative signal.

If the first iteration gradient If it is not close to zero, the second usage value cur_Use_S follows the threshold value used by the threshold function in the last iteration, that is, the second usage value cur_Use_S is assigned as the first usage value pre_Use_S of the target sparsity K; In an iterative process, the first use value pre_Use_S is equal to the maximum sparsity K _m given by the input parameter.

S2036. Acquire a second iteration signal by performing a hard threshold operation through a threshold function.

Specifically, according to the process signal at and the second use value _{cur_Use_S} , the second iteration signal is obtained through the threshold function The calculation formula is:

Among them, Γ() represents the threshold function, and the hard threshold operation is used to retain the second use value _{cur_Use_S} element values with the largest element amplitude in the process signal at, and at the same _time to remove the first element with the largest amplitude in the process signal at 2. Set the value of all other elements except cur_Use_S elements to zero, and assign the processing result of the threshold function Γ() to the second iteration signal

S2037. Acquire a second estimated value of the target sparsity.

Specifically, for the second iteration signal Perform unit normalization to obtain normalized iterative signals The calculation formula is:

Among them, |||| ₂ represents the two-norm of the vector; then, according to the target sparsity decision threshold λ given by the input parameters, compare and count the normalized iteration signals The number of non-zero elements in which the absolute value of the element amplitude exceeds the decision threshold λ, and the number of non-zero elements is recorded as the second estimated value cur_Est_S of the target sparsity K, which can be expressed as:

Among them, count() represents a counting function, which is used to record the number of elements in the vector that satisfy the condition.

S2038. Determine whether the first iteration gradient approaches zero and whether the second estimated value is equal to the first used value.

Specifically, if the first iteration gradient If the vector does not approach zero or the second estimated value cur_Est_S is not equal to the first use value pre_Use_S, execute S2039;

Or, if the first iteration gradient The vector approaches to zero and the second estimated value cur_Est_S is equal to the first usage value pre_Use_S, then execute S2030.

S2039 , update the process parameters, and continue to perform subsequent iterations.

Specifically, if the first iteration gradient If the vector does not approach zero or the second estimated value cur_Est_S is not equal to the first use value pre_Use_S, then the first estimate value pre_Est_S is assigned as the second estimate value cur_Est_S, the first use value pre_Use_S is assigned as the second use value cur_Use_S, and Add 1 to the subscript t of the iterative signal, that is, the second iterative signal in the current iterative process as the first iteration signal during the next iterative update Participate in the calculation, and re-execute the operations from S2031 to S2038.

S2030. Obtain the target sparsity.

Specifically, if the first iteration gradient If the vector is close to zero and the second estimated value cur_Est_S is equal to the first use value pre_Use_S, then assign the target sparsity K to the second estimated value cur_Est_S to obtain the target sparsity K.

S204. Perform signal reconstruction on the measurement signal according to the target sparsity to obtain a reconstructed signal of the original sparse signal.

Specifically, according to the measurement signal y, the sampling matrix Φ, the first iteration signal The target sparsity is K, and the signal reconstruction operation is performed to obtain the reconstructed signal, as shown in Figure 4, including:

S2041. Calculate and obtain the first iteration gradient.

Specifically, according to the measurement signal y and the sampling matrix Φ received in S201, and the first iteration signal Calculate the first iteration gradient corresponding to the current iteration process

It is worth mentioning that the first iteration signals It can be the value retained after the last iteration process in S203, or the initial value of the first iteration signal after initialization in S202

first iteration gradient In order to reflect the parameters of the iterative signal change rate in the iterative update process, the calculation formula is:

Specifically, combined with the sampling matrix Φ, in the first iterative signal After low-speed sampling processing and symbol quantization processing, the difference with the measurement signal y is obtained, and the difference is multiplied by the transpose matrix of the sampling matrix Φ to the left to obtain the first iteration gradient

S2042. Calculate and acquire the process signal.

Specifically, through the gradient descent method, according to the first iteration gradient with the first iteration signal Update the iterative intermediate process signal at along the direction of gradient _descent , the calculation formula is:

S2043. Acquire a second iteration signal according to the process signal and the target sparsity.

Specifically, according to the target sparsity K obtained in S203 and the process signal at obtained in _S2042 , the hard threshold operation is performed through the threshold function to obtain the second iteration signal The calculation formula is:

Among them, Γ() represents the threshold function, and the hard threshold operation is used to retain the K element values of the target _sparsity with the largest element amplitude in the process signal at, and at the same time divide the target _sparseness with the largest element amplitude in the process signal at Set the value of all other elements except K elements to zero, and assign the processing result of the threshold function Γ() to the second iteration signal

S2044. Calculate and obtain the Hamming distance.

Specifically, according to the measurement signal y and the sampling matrix Φ, the second iteration signal Carry out the same signal processing operation as that of the encoder, that is, first perform low-speed sampling processing and then perform symbol quantization processing, and further calculate the difference with the measured signal y, and count the number of non-zero elements in the difference as the Hamming distance hd, calculate The formula is:

Among them, the function nnz() represents the number of non-zero elements in the solution vector, and the Hamming distance hd is the second iteration signal After low-speed sampling processing and symbol quantization processing The number of elements whose value is different from that of the corresponding position element of the measurement signal y, which is a non-negative integer.

S2045. Determine whether the Hamming distance is greater than a preset threshold.

Specifically, it is judged whether the Hamming distance is greater than the preset threshold value, if the Hamming distance hd is greater than the preset threshold value, then execute S2046; or, if the Hamming distance hd is less than or equal to the preset threshold value, then Execute S2047.

Wherein, the preset threshold value can be a default setting of the system, or can be given by an external input, which can be set and subsequently adjusted according to the requirements for the accuracy of the reconstructed signal, and there is no limitation here.

S2046. Jump to execute the next iterative update process.

If the Hamming distance hd is greater than the preset threshold value, add 1 to the subscript t of the iterative signal, that is, the second iterative signal in the current iterative process as the first iteration signal during the next iterative update Participate in the calculation, and re-execute the operations from S2041 to S2045.

S2047. Acquire a reconstruction signal.

Specifically, if the Hamming distance is less than or equal to the preset threshold value, then for the second iteration signal Perform unit normalization to obtain the reconstructed signal of the original sparse signal The calculation formula is:

S205. Output the acquired reconstructed signal as a final result.

Embodiments of the present invention provide a signal processing method. By acquiring input parameters and encoding end data, after initializing process parameters and related parameter values, an adaptive sparsity estimation operation is performed, and after acquiring the target sparsity, according to The target sparsity is used to reconstruct the signal to obtain the reconstructed signal of the original sparse signal. In this way, the problem of signal reconstruction performance degradation due to limited sparsity prior information is solved, the deviation caused by reconstruction signal is reduced or basically eliminated, and the practicability and accuracy of single-bit compressed sensing technology are improved.

The embodiment of the present invention also provides a signal processing device 00, as shown in Figure 5, the signal processing device 00 includes:

The parameter acquisition unit 001 is used to acquire input parameters and encoding end data, and initialize the process parameters and the first iteration signal. The encoding end data includes the measurement signal obtained by the encoding end performing low-speed sampling processing and symbol quantization processing on the original sparse signal;

The sparsity estimation unit 002 is used to perform an adaptive sparsity estimation operation according to the input parameters, the encoding end data, the initialized process parameters and the first iteration signal to obtain the target sparsity;

The signal reconstruction unit 003 is configured to perform signal reconstruction on the measurement signal according to the target sparsity to obtain a reconstructed signal of the original sparse signal.

Optionally, the parameter acquisition unit 001 is specifically used for:

Obtain input parameters, including the maximum sparsity K _m and the decision threshold λ;

Receive the encoding end data transmitted by the encoding end, the encoding end data includes the measurement signal y, the sampling matrix Φ; the measurement signal y is the encoding end combined with the sampling matrix Φ to perform low-speed sampling processing on the original sparse signal θ to obtain the sampling signal x, and then the sampling signal x is obtained by sign quantization.

Optionally, the process parameters include the first estimated value pre_Est_S and the first use value pre_Use_S of the target sparsity K;

The parameter acquisition unit 001 also includes an initialization unit 004, as shown in Figure 6, specifically for:

Assign the first estimated value pre_Est_s as the maximum sparsity K _m , and assign the first use value pre_Use_S as the maximum sparsity K _m ;

The first iteration signal initial value of Assignment is a zero vector.

Optionally, the sparsity estimation unit 002 is specifically configured to perform an adaptive sparsity estimation operation, where the adaptive sparsity estimation operation includes:

According to the measurement signal y and the sampling matrix Φ, the first iteration signal is obtained The difference between the measured signal y and the measured signal y after low-speed sampling processing and symbol quantization processing, and the difference is multiplied by the transpose matrix of the sampling matrix Φ to the left to obtain the first iteration gradient

_Obtain the process signal at by gradient _descent method, and the process signal at is the first iteration signal with the first iteration gradient the difference;

For the first iteration gradient It is judged whether it is close to the zero vector, if the gradient of the first iteration approaching the zero vector, assign the second use value cur_Use_S of the target sparsity K to the first estimated value pre_Est_S; or, if the first iteration gradient If it does not approach the zero vector, assign the second use value cur_Use_S to the first use value pre_Use_S;

According to the process signal at and the second use value _{cur_Use_S} , the second iteration signal is obtained through the threshold function Γ() The threshold function Γ() is used to keep the second use value _{cur_Use_S} element values with the largest element amplitude in the process signal at, and at the same time save the process signal at except the second use value _{cur_Use_S} elements with the largest element amplitude All other element values of are set to zero, and the processing result of the threshold function Γ() is assigned to the second iteration signal

For the second iteration signal Perform unit normalization to obtain normalized iterative signals Statistically normalized iterative signal The number of non-zero elements whose absolute value of the element amplitude exceeds the decision threshold λ, and assign the number of non-zero elements to the second estimated value cur_Est_S of the target sparsity K;

If the first iteration gradient If the vector does not approach zero or the second estimated value cur_Est_S is not equal to the first use value pre_Use_S, then the first estimate value pre_Est_S is assigned as the second estimate value cur_Est_S, the first use value pre_Use_S is assigned as the second use value cur_Use_S, and Add 1 to the subscript t of the iterative signal, that is, the second iterative signal in the current iterative process as the first iteration signal during the next iterative update Participate in the calculation and re-execute the adaptive sparsity estimation operation;

Or, if the first iteration gradient is approaching the zero vector and the second estimated value cur_Est_S is equal to the first use value pre_Use_S, then the target sparsity K is assigned as the second estimated value cur_Est_S to obtain the target sparsity K.

Optionally, the signal reconstruction unit 003 is specifically used for:

According to the encoder data, the first iteration signal With the target sparsity K, the signal reconstruction operation is performed, and the signal reconstruction operation specifically includes:

According to the measurement signal y and the sampling matrix Φ, the first iteration signal is obtained The difference between the measured signal y and the measured signal y after low-speed sampling processing and symbol quantization processing, and the difference is multiplied by the transpose matrix of the sampling matrix Φ to the left to obtain the first iteration gradient

_Obtain the process signal at by gradient _descent method, and the process signal at is the first iteration signal with the first iteration gradient the difference;

Combining the process signal at and the target _sparsity K, the second iteration signal is obtained through the threshold function Γ() The threshold function Γ() is used to keep the target sparsity K elements with the largest element amplitude in the process signal a _t , and at the same time save the other elements in the process signal a _t except the target sparsity K elements with the largest element amplitude All element values are set to zero, and the processing result of the threshold function Γ() is assigned to the second iteration signal

According to the measurement signal y, the sampling matrix Φ and the second iteration signal Obtain the Hamming distance hd through the non-zero statistical function; the non-zero statistical function is used to obtain the second iteration signal The difference between the measured signal y and the measured signal y after low-speed sampling processing and symbol quantization processing, and the number of non-zero elements in the difference are counted, and the number of non-zero elements is assigned to the Hamming distance hd; the Hamming distance hd is the first Second iteration signal After low-speed sampling and symbol quantization processing, the number of elements is different from the corresponding position element of the measurement signal y;

If the Hamming distance hd is greater than the preset threshold value, add 1 to the subscript t of the iterative signal, that is, the second iterative signal in the current iterative process as the first iteration signal during the next iterative update Participate in the calculation, and re-execute the signal reconstruction operation; or, if the Hamming distance hd is less than or equal to the preset threshold value, then for the second iteration signal Perform unit normalization to obtain the reconstructed signal of the original sparse signal

An embodiment of the present invention provides a signal processing device, which performs an adaptive sparsity estimation operation after initializing the process parameters and related parameter values by acquiring input parameters and encoding end data, and after acquiring the target sparsity, according to The target sparsity is used to reconstruct the signal to obtain the reconstructed signal of the original sparse signal. In this way, the problem of signal reconstruction performance degradation due to limited sparsity prior information is solved, the deviation caused by reconstruction signal is reduced or basically eliminated, and the practicability and accuracy of single-bit compressed sensing technology are improved.

The embodiment of the present invention also provides a signal processing device 01, as shown in Figure 7, the signal processing device 01 includes:

A bus 011, and a processor 012 connected to the bus 011, a memory 013 and an interface 014, wherein the interface 014 is used to communicate with external devices;

The memory 013 is used to store instructions, and the processor 012 is used to execute the instructions to obtain input parameters and encoding end data, and initialize process parameters and the first iteration signal, and the encoding end data includes the original sparse signal performed by the encoding end Measurement signals obtained by low-speed sampling processing and symbol quantization processing;

The processor 012 executes the instruction and is also used to perform an adaptive sparsity estimation operation according to the input parameters, the encoding end data, the initialized process parameters and the first iteration signal, so as to obtain the target sparsity;

The processor 012 executing the instruction is also used to perform signal reconstruction on the measurement signal according to the target sparsity to obtain a reconstructed signal of the original sparse signal.

In the embodiment of the present invention, optionally, the execution of the instruction by the processor 012 may be specifically used to obtain input parameters, the input parameters include the maximum sparsity K _m and the decision threshold λ;

Receive the encoding end data transmitted by the encoding end, the encoding end data includes the measurement signal y, the sampling matrix Φ; the measurement signal y is the encoding end combined with the sampling matrix Φ to the original sparse signal θ after low-speed sampling processing to obtain the sampling signal x, and then the sampling signal x is obtained after sign quantization processing.

In the embodiment of the present invention, optionally, the process parameters include the first estimated value pre_Est_S of the target sparsity K and the first use value pre_Use_S, and the processor 012 executing the instruction can be specifically used for:

Assign the first estimated value pre_Est_S as the maximum sparsity K _m , and assign the first use value pre_Use_S as the maximum sparsity K _m ;

The first iteration signal initial value of Assignment is a zero vector.

In this embodiment of the present invention, optionally, the processor 012 executing the instruction may be specifically used to perform an adaptive sparsity estimation operation, where the adaptive sparsity estimation operation includes:

According to the measurement signal y and the sampling matrix Φ, the first iteration signal is obtained The difference between the measured signal y and the measured signal y after low-speed sampling processing and symbol quantization processing, and the difference is multiplied by the transpose matrix of the sampling matrix Φ to the left to obtain the first iteration gradient

_Obtain the process signal at by gradient _descent method, and the process signal at is the first iteration signal with the first iteration gradient the difference;

For the first iteration gradient It is judged whether it is close to the zero vector, if the gradient of the first iteration approaching the zero vector, assign the second use value cur_Use_S of the target sparsity K to the first estimated value pre_Est_S; or, if the first iteration gradient If it does not approach the zero vector, assign the second use value cur_Use_S to the first use value pre_Use_S;

According to the process signal at and the second use value _{cur_Use_S} , the second iteration signal is obtained through the threshold function Γ() The threshold function Γ() is used to keep the second use value _{cur_Use_S} element values with the largest element amplitude in the process signal at, and at the same time save the process signal at except the second use value _{cur_Use_S} elements with the largest element amplitude All other element values of are set to zero, and the processing result of the threshold function Γ() is assigned to the second iteration signal

For the second iteration signal Perform unit normalization to obtain normalized iterative signals Statistically normalized iterative signal The number of non-zero elements whose absolute value of the element amplitude exceeds the decision threshold λ, and assign the number of non-zero elements to the second estimated value cur_Est_S of the target sparsity K;

If the first iteration gradient If the vector does not approach zero or the second estimated value cur_Est_S is not equal to the first use value pre_Use_S, then the first estimate value pre_Est_S is assigned as the second estimate value cur_Est_S, the first use value pre_Use_S is assigned as the second use value cur_Use_S, and Add 1 to the subscript t of the iterative signal, that is, the second iterative signal in the current iterative process as the first iteration signal during the next iterative update Participate in the calculation and re-execute the adaptive sparsity estimation operation;

Or, if the first iteration gradient is approaching the zero vector and the second estimated value cur_Est_S is equal to the first use value pre_Use_S, then the target sparsity K is assigned as the second estimated value cur_Est_S to obtain the target sparsity K.

In this embodiment of the present invention, optionally, the processor 012 executing the instruction may be specifically used to With the target sparsity K, the signal reconstruction operation is performed, and the signal reconstruction operation specifically includes:

According to the measurement signal y and the sampling matrix Φ, the first iteration signal is obtained The difference between the measured signal y and the measured signal y after low-speed sampling processing and symbol quantization processing, and the difference is multiplied by the transpose matrix of the sampling matrix Φ to the left to obtain the first iteration gradient

_Obtain the process signal at by gradient _descent method, and the process signal at is the first iteration signal with the first iteration gradient the difference;

Combining the process signal at and the target _sparsity K, the second iteration signal is obtained through the threshold function Γ() The threshold function Γ() is used to keep the target sparsity K elements with the largest element amplitude in the process signal a _t , and at the same time save the other elements in the process signal a _t except the target sparsity K elements with the largest element amplitude All element values are set to zero, and the processing result of the threshold function Γ() is assigned to the second iteration signal

According to the measurement signal y, the sampling matrix Φ and the second iteration signal Obtain the Hamming distance hd through the non-zero statistical function; the non-zero statistical function is used to obtain the second iteration signal The difference between the measured signal y and the measured signal y after low-speed sampling processing and symbol quantization processing, and the number of non-zero elements in the difference are counted, and the number of non-zero elements is assigned to the Hamming distance hd; the Hamming distance hd is the first Second iteration signal After low-speed sampling and symbol quantization processing, the number of elements is different from the corresponding position element of the measurement signal y;

If the Hamming distance hd is greater than the preset threshold value, add 1 to the subscript t of the iterative signal, that is, the second iterative signal in the current iterative process as the first iteration signal during the next iterative update Participate in the calculation, and re-execute the signal reconstruction operation; or, if the Hamming distance hd is less than or equal to the preset threshold value, then for the second iteration signal Perform unit normalization to obtain the reconstructed signal of the original sparse signal

An embodiment of the present invention provides a signal processing device, which performs an adaptive sparsity estimation operation after initializing the process parameters and related parameter values by acquiring input parameters and encoding end data, and after acquiring the target sparsity, according to The target sparsity is used to reconstruct the signal to obtain the reconstructed signal of the original sparse signal. In this way, the problem of signal reconstruction performance degradation due to limited sparsity prior information is solved, the deviation caused by reconstruction signal is reduced or basically eliminated, and the practicability and accuracy of single-bit compressed sensing technology are improved.

The above is only a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Anyone skilled in the art can easily think of changes or substitutions within the technical scope disclosed in the present invention. Should be covered within the protection scope of the present invention. Therefore, the protection scope of the present invention should be based on the protection scope of the claims.

Claims (12)

1. A signal processing method, characterized in that the method comprises: Acquiring input parameters and encoding end data, and initializing the process parameters and the first iteration signal, the encoding end data including the measurement signal obtained by the encoding end performing low-speed sampling processing and symbol quantization processing on the original sparse signal; performing an adaptive sparsity estimation operation according to the input parameters, the encoding end data, the initialized process parameters, and the first iteration signal to obtain a target sparsity; performing signal reconstruction on the measurement signal according to the target sparsity to obtain a reconstructed signal of the original sparse signal; Wherein, the input parameters include a maximum sparsity and a decision threshold; the encoding end data includes the measurement signal and a sampling matrix; the initialized process parameters include the first estimated value of the target sparsity, the second a use value; The performing an adaptive sparsity estimation operation according to the input parameters, the encoding end data, the initialized process parameters, and the first iteration signal, and obtaining the target sparsity includes: According to the measurement signal and the sampling matrix, obtain the difference between the first iteration signal and the measurement signal after low-speed sampling and sign quantization, and multiply the difference by the transpose matrix of the sampling matrix to the left, Get the gradient of the first iteration; Obtaining a process signal by a gradient descent method, where the process signal is a difference between the first iteration signal and the first iteration gradient; Judging whether the first iterative gradient approaches a zero vector, and if the first iterative gradient approaches a zero vector, assigning the second used value of the target sparsity as the first estimated value; or, If the first iterative gradient does not approach the zero vector, assigning the second use value as the first use value; According to the process signal and the second use value, a second iterative signal is obtained through a threshold function, and the threshold function is used to reserve the second use value and element values with the largest element amplitude in the process signal, At the same time, all other element values in the process signal except the second use value elements with the largest element amplitude are set to zero, and the processing result of the threshold function is assigned to the second iteration signal; Carrying out unit normalization to the second iterative signal to obtain a normalized iterative signal, counting the number of non-zero elements whose absolute value of the element amplitude exceeds the decision threshold in the normalized iterative signal, and counting the number of elements assigning a value to the second estimate of the target sparsity; If the first iterative gradient does not approach the zero vector or the second estimated value is not equal to the first used value, assign the first estimated value to the second estimated value, the first estimated value A use value is assigned as the second use value, the first iteration signal is assigned as the second iteration signal, and the adaptive sparsity estimation operation is re-executed; or, if the gradient of the first iteration approaches If it is a zero vector and the second estimated value is equal to the first used value, assign the target sparsity as the second estimated value to obtain the target sparsity. 2. The method according to claim 1, wherein the measurement signal is that after the encoding end performs low-speed sampling processing on the original sparse signal in conjunction with the sampling matrix to obtain the sampling signal, the sampling signal is then obtained by sign quantization. 3. The method according to claim 1 or 2, wherein said initializing the process parameters and the first iteration signal comprises: assigning the first estimated value as the maximum sparsity, and assigning the first use value as the maximum sparsity; Assigning the initial value of the first iteration signal as a zero vector. 4. The method according to claim 1 or 2, wherein, according to the target sparsity, performing signal reconstruction on the measurement signal to obtain a reconstructed signal of the original sparse signal comprises: Perform a signal reconstruction operation according to the encoding end data, the first iteration signal, and the target sparsity, and the signal reconstruction operation specifically includes: According to the measurement signal and the sampling matrix, obtain the difference between the first iteration signal and the measurement signal after low-speed sampling and symbol quantization, and multiply the difference by the transpose matrix of the sampling matrix on the left , get the gradient of the first iteration; Obtaining a process signal by a gradient descent method, where the process signal is a difference between the first iteration signal and the first iteration gradient; Combining the process signal and the target sparsity, a second iterative signal is obtained through a threshold function, the threshold function is used to retain the target sparsity element values with the largest element amplitude in the process signal, and at the same time In the process signal, all element values except the target sparsity elements with the largest element amplitude are set to zero, and the processing result of the threshold function is assigned to the second iteration signal; According to the measurement signal, the sampling matrix and the second iteration signal, the Hamming distance is obtained through a non-zero statistical function; the non-zero statistical function is used to obtain the second iteration signal through low-speed sampling processing and the difference between the measurement signal after sign quantization processing and counting the number of non-zero elements in the difference, assigning the number of non-zero elements to the Hamming distance; the Hamming distance is the number of elements whose value is different from that of the measurement signal at the corresponding position of the second iterative signal after low-speed sampling processing and sign quantization processing; If the Hamming distance is greater than a preset threshold value, assign the first iterative signal to the second iterative signal, and re-execute the signal reconstruction operation; or, if the Hamming distance is less than or equal to For the preset threshold value, unit normalization processing is performed on the second iteration signal to obtain the reconstructed signal of the original sparse signal. 5. A signal processing device, characterized in that the device comprises: A parameter acquisition unit, configured to acquire input parameters and encoding end data, and initialize process parameters and the first iteration signal. The encoding end data includes measurement signals acquired by the encoding end through low-speed sampling processing and symbol quantization processing on the original sparse signal ; A sparsity estimation unit, configured to perform an adaptive sparsity estimation operation according to the input parameters, the encoding end data, the initialized process parameters, and the first iteration signal to obtain a target sparsity; A signal reconstruction unit, configured to perform signal reconstruction on the measurement signal according to the target sparsity to obtain a reconstructed signal of the original sparse signal; Wherein, the input parameters include a maximum sparsity and a decision threshold; the encoding end data includes the measurement signal and a sampling matrix; the initialized process parameters include the first estimated value of the target sparsity, the second a use value; The sparsity estimation unit is configured to perform the adaptive sparsity estimation operation, specifically including: According to the measurement signal and the sampling matrix, the difference between the first iteration signal and the measurement signal after low-speed sampling and symbol quantization is obtained, and the difference is multiplied by the transpose matrix of the sampling matrix to the left to obtain first iteration gradient; Obtaining a process signal by a gradient descent method, where the process signal is a difference between the first iteration signal and the first iteration gradient; Judging whether the first iterative gradient approaches a zero vector, and if the first iterative gradient approaches a zero vector, assigning the second used value of the target sparsity as the first estimated value; or, If the first iterative gradient does not approach the zero vector, assigning the second use value as the first use value; According to the process signal and the second use value, a second iterative signal is obtained through a threshold function, and the threshold function is used to reserve the second use value and element values with the largest element amplitude in the process signal, At the same time, all other element values in the process signal except the second use value elements with the largest element amplitude are set to zero, and the processing result of the threshold function is assigned to the second iteration signal; Carrying out unit normalization to the second iterative signal to obtain a normalized iterative signal, counting the number of non-zero elements whose absolute value of the element amplitude exceeds the decision threshold in the normalized iterative signal, and counting the number of elements assigning a value to the second estimate of the target sparsity; If the first iterative gradient does not approach the zero vector or the second estimated value is not equal to the first used value, assign the first estimated value to the second estimated value, the first estimated value A use value is assigned as the second use value, the first iteration signal is assigned as the second iteration signal, and the adaptive sparsity estimation operation is re-executed; or, if the gradient of the first iteration approaches If it is a zero vector and the second estimated value is equal to the first used value, assign the target sparsity as the second estimated value to obtain the target sparsity. 6. The device according to claim 5, wherein the parameter acquisition unit is specifically used for: Obtain the input parameters; receiving the encoding end data transmitted by the encoding end; the measurement signal is that the encoding end performs low-speed sampling processing on the original sparse signal in combination with the sampling matrix to obtain the sampling signal, and then signs the sampling signal obtained by quantitative processing. 7. The device according to claim 5 or 6, wherein the parameter acquisition unit also includes an initialization unit, specifically for: assigning the first estimated value as the maximum sparsity, and assigning the first use value as the maximum sparsity; Assigning the initial value of the first iteration signal as a zero vector. 8. The device according to claim 5 or 6, wherein the signal reconstruction unit is specifically used for: Perform a signal reconstruction operation according to the encoding end data, the first iteration signal, and the target sparsity, and the signal reconstruction operation specifically includes: According to the measurement signal and the sampling matrix, obtain the difference between the first iteration signal and the measurement signal after low-speed sampling and symbol quantization, and multiply the difference by the transpose matrix of the sampling matrix on the left , get the gradient of the first iteration; Obtaining a process signal by a gradient descent method, where the process signal is a difference between the first iteration signal and the first iteration gradient; Combining the process signal and the target sparsity, a second iterative signal is obtained through a threshold function, the threshold function is used to retain the target sparsity element values with the largest element amplitude in the process signal, and at the same time In the process signal, all element values except the target sparsity elements with the largest element amplitude are set to zero, and the processing result of the threshold function is assigned to the second iteration signal; According to the measurement signal, the sampling matrix and the second iteration signal, the Hamming distance is obtained through a non-zero statistical function; the non-zero statistical function is used to obtain the second iteration signal through low-speed sampling processing and the difference between the measurement signal after sign quantization processing and counting the number of non-zero elements in the difference, assigning the number of non-zero elements to the Hamming distance; the Hamming distance is the number of elements whose value is different from that of the measurement signal at the corresponding position of the second iterative signal after low-speed sampling processing and sign quantization processing; If the Hamming distance is greater than a preset threshold value, assign the first iterative signal to the second iterative signal, and re-execute the signal reconstruction operation; or, if the Hamming distance is less than or equal to For the preset threshold value, unit normalization processing is performed on the second iteration signal to obtain the reconstructed signal of the original sparse signal. 9. A signal processing device, characterized in that the signal processing device comprises: a bus, and a processor connected to the bus, a memory, and an interface, wherein the interface is used to communicate with an external device; the memory for storing instructions, the processor executes the instructions for: Acquiring input parameters and encoding end data, and initializing the process parameters and the first iteration signal, the encoding end data including the measurement signal obtained by the encoding end performing low-speed sampling processing and symbol quantization processing on the original sparse signal; performing an adaptive sparsity estimation operation according to the input parameters, the encoding end data, the initialized process parameters, and the first iteration signal to obtain a target sparsity; performing signal reconstruction on the measurement signal according to the target sparsity to obtain a reconstructed signal of the original sparse signal; Wherein, the input parameters include a maximum sparsity and a decision threshold; the encoding end data includes the measurement signal and a sampling matrix; the initialized process parameters include the first estimated value of the target sparsity, the second a use value; The processor executes the instruction and is also specifically used for: According to the measurement signal and the sampling matrix, the difference between the first iteration signal and the measurement signal after low-speed sampling and symbol quantization is obtained, and the difference is multiplied by the transpose matrix of the sampling matrix to the left to obtain first iteration gradient; Obtaining a process signal by a gradient descent method, where the process signal is a difference between the first iteration signal and the first iteration gradient; Judging whether the first iterative gradient approaches a zero vector, and if the first iterative gradient approaches a zero vector, assigning the second used value of the target sparsity as the first estimated value; or, If the first iterative gradient does not approach the zero vector, assigning the second use value as the first use value; According to the process signal and the second use value, a second iterative signal is obtained through a threshold function, and the threshold function is used to reserve the second use value and element values with the largest element amplitude in the process signal, At the same time, all other element values in the process signal except the second use value elements with the largest element amplitude are set to zero, and the processing result of the threshold function is assigned to the second iteration signal; Carrying out unit normalization to the second iterative signal to obtain a normalized iterative signal, counting the number of non-zero elements whose absolute value of the element amplitude exceeds the decision threshold in the normalized iterative signal, and counting the number of elements assigning a value to the second estimate of the target sparsity; If the first iterative gradient does not approach the zero vector or the second estimated value is not equal to the first used value, assign the first estimated value to the second estimated value, the first estimated value A use value is assigned as the second use value, the first iteration signal is assigned as the second iteration signal, and the adaptive sparsity estimation operation is re-executed; or, if the gradient of the first iteration approaches If it is a zero vector and the second estimated value is equal to the first used value, assign the target sparsity as the second estimated value to obtain the target sparsity. 10. The device according to claim 9, wherein the processor executes the instructions for: Obtain the input parameters; receiving the encoding end data transmitted by the encoding end; the measurement signal is that the encoding end performs low-speed sampling processing on the original sparse signal in combination with the sampling matrix to obtain the sampling signal, and then signs the sampling signal obtained by quantitative processing. 11. The device according to claim 9 or 10, wherein the processor executes the instruction for: assigning the first estimated value as the maximum sparsity, and assigning the first use value as the maximum sparsity; Assigning the initial value of the first iteration signal as a zero vector. 12. The device according to claim 9 or 10, wherein the processor executing the instruction is also specifically used for: Perform a signal reconstruction operation according to the encoding end data, the first iteration signal, and the target sparsity, and the signal reconstruction operation specifically includes: According to the measurement signal and the sampling matrix, obtain the difference between the first iteration signal and the measurement signal after low-speed sampling and symbol quantization, and multiply the difference by the transpose matrix of the sampling matrix on the left , get the gradient of the first iteration; Obtaining a process signal by a gradient descent method, where the process signal is a difference between the first iteration signal and the first iteration gradient; Combining the process signal and the target sparsity, a second iterative signal is obtained through a threshold function, the threshold function is used to retain the target sparsity element values with the largest element amplitude in the process signal, and at the same time In the process signal, all element values except the target sparsity elements with the largest element amplitude are set to zero, and the processing result of the threshold function is assigned to the second iteration signal; According to the measurement signal, the sampling matrix and the second iteration signal, the Hamming distance is obtained through a non-zero statistical function; the non-zero statistical function is used to obtain the second iteration signal through low-speed sampling processing and the difference between the measurement signal after sign quantization processing and counting the number of non-zero elements in the difference, assigning the number of non-zero elements to the Hamming distance; the Hamming distance is the number of elements whose value is different from that of the measurement signal at the corresponding position of the second iterative signal after low-speed sampling processing and sign quantization processing; If the Hamming distance is greater than a preset threshold value, assign the first iterative signal to the second iterative signal, and re-execute the signal reconstruction operation; or, if the Hamming distance is less than or equal to For the preset threshold value, unit normalization processing is performed on the second iteration signal to obtain the reconstructed signal of the original sparse signal.