CN104104390A - Signal compression method, signal reconstruction method, and correlation apparatus and system - Google Patents

Abstract

The embodiments of the invention disclose a signal compression method, a signal reconstruction method, and a correlation apparatus and system, for reducing influences exerted by quantification errors on signal construction and improving signal construction accuracy. One method comprises: first of all, performing low-speed sampling on input signals by use of a sampling matrix to obtain sampling signals, then performing amplitude quantification on M sampling values of the sampling signals by use of a quantification function to obtain quantification signals, next, determining whether M quantification values in the quantification signals exceed a quantification error tolerance, obtaining an m-th quantification value which does not exceed the quantification error tolerance from the M quantification values, according to all the m-th quantification value, the sampling signals and the sampling matrix, generating a sampling reservation matrix, and finally, sending quantification reservation signals composed of all the m-th quantification values and the sampling reservation matrix to a signal reconstruction apparatus.

Description

A kind of compression method and reconstructing method and relevant apparatus and system

Technical field

The present invention relates to communication technical field, relate in particular to a kind of compression method and reconstructing method and relevant apparatus and system.

Background technology

For the signal with sparse characteristic, the scholars such as Emmanuel Candes proposed compressed sensing (CS, Compressed Sensing) theory in 2004.The potential sparse property of CS technology based on signal, it is in the situation that guaranteeing that signal is not suffered a loss, use the speed requiring far below Nyquist sampling thheorem to carry out down-sampled to signal, signal vector is carried out to dimensionality reduction linear projection, the data that obtain after low speed sampling are sampling output vectors (dimension of this sampled data vector is less than the dimension of original signal vector) of a dimensionality reduction, and the potential sparse property that signal has makes this problem of owing to determine, (be the number that the number of equation or equation is less than unknown quantity, wherein, the number of equation or equation equals the dimension of the sampled data vector of low speed sampling output, the number of unknown quantity equals the dimension of primary signal) can solve by methods such as protruding optimization or greedy searchs, by less low-dimensional sampled data, rebuild original high dimensional signal.Because the efficient information processing manner of CS technology can significantly reduce acquisition of information expense, can correctly recover former sparse signal with higher probability simultaneously, attracted at present the close attention of academia and industrial quarters, in real system, be with a wide range of applications, for example: image processing, channel estimating, wireless sensor network, cognitive radio frequency spectrum detection, target localization etc.

Existing employing CS technology is in the method for signal reconstruction, most research is mainly to carry out signal reconstruction for the sampled data of desirable non-quantification, suppose that the dimensionality reduction sampled data obtaining after the sampling of CS low speed is not through quantization operation, the amplitude of the sampled data that obtains is not carried out discretization processing, still the continuity in maintenance amplitude is then carried out CS signal reconstruction according to the continuous sampled data of these amplitudes in CS signal reconfiguring method.Yet, this hypothesis is too desirable, in actual applications, because just can be convenient to carry out the operations such as subsequent treatment, transmission, storage after conventionally all needing that the signal of analog domain is transformed into numeric field, therefore must need obtained sampled data signal to carry out quantification treatment, carry out original amplitude of approximate representation sampled data by some discrete amplitudes.In concrete quantization operation, consider that the data volume that coding side sends to decoding end is multiplied along with the increase of quantized level, cause transport overhead to increase, therefore the quantizing bit number of quantizer is normally limited, so quantized level is also limited, therefore in the practical application of CS technology, can not avoid and need to carry out quantization operation to CS sampled data.

To sum up, available technology adopting CS technology is on ignoring the impact quantizing CS technology in the method for signal reconstruction, directly the sampled data after quantizing is used for to signal reconstruction, but the quantization error that when quantized level causing when quantizing bit number is less is less, CS coding side is introduced can have a strong impact on the accuracy of CS decoding end signal reconstructed results.Therefore when quantizing bit number is less, the accuracy that how to improve signal reconstruction need to inquire into.

Summary of the invention

The embodiment of the present invention provides a kind of compression method and reconstructing method and relevant apparatus and system, for reducing the impact of quantization error on signal reconstruction, improves the accuracy of signal reconstruction.

For solving the problems of the technologies described above, the embodiment of the present invention provides following technical scheme:

First aspect, the embodiment of the present invention provides a kind of compression method, comprising:

Use sampling matrix to carry out low speed sampling to input signal, obtain sampled signal, wherein, described sampling matrix comprises M the row vector being comprised of N row sampling matrix value, described input signal is the column vector being comprised of N input signal values, described sampled signal is the column vector being comprised of M sampled value, the natural number that described M, N are non-zero, and M < N;

Use quantization function to carry out amplitude quantizing to the M of a described sampled signal sampled value, obtain quantized signal, described quantized signal is the column vector being comprised of M quantized value;

Judge respectively whether M quantized value surpasses preset quantization error tolerance;

From a described M quantized value, obtain m the quantized value that does not surpass quantization error tolerance, according to all described m quantized values, described sampled signal and described sampling matrix generate sampling and retain matrix, wherein, described m quantized value is in a described M quantized value, not surpass a quantized value of quantization error tolerance, and in a described M quantized value, at least there is a described m quantized value, described m quantized value is corresponding with m sampled value in described sampled signal, the row vector that described m the sampled value m by described sampling matrix is comprised of N row sampling matrix value is multiplied by described input signal and obtains, described sampling retains m the row vector being comprised of N row sampling matrix value that matrix comprises all described sampling matrixs,

Quantification stick signal and described sampling reservation matrix are sent to signal reconstruction device, and described quantification stick signal comprises all described m quantized values.

Second aspect, the embodiment of the present invention also provides a kind of signal reconfiguring method, comprising:

Receive quantification stick signal and sampling reservation matrix that Signal Compression device sends, wherein, described quantification stick signal comprises m all quantized values, described m quantized value is in M quantized value, not surpass a quantized value of quantization error tolerance, a described M quantized value is that described Signal Compression device is used sampling matrix to carry out after low speed sampling acquisition sampled signal input signal, use quantization function sampled signal to be carried out to the result of amplitude quantizing, the number of the row vector that described M is described sampling matrix;

According to the quantification stick signal receiving and sampling reservation matrix, carry out signal reconstruction, obtain signal reconstruction result;

Export described signal reconstruction result.

The third aspect, the embodiment of the present invention provides a kind of Signal Compression device, comprising:

Sampling module, be used for using sampling matrix to carry out low speed sampling to input signal, obtain sampled signal, wherein, described sampling matrix comprises M the row vector being comprised of N row sampling matrix value, and described input signal is the column vector being comprised of N input signal values, and described sampled signal is the column vector being comprised of M sampled value, described M, N are the natural number of non-zero, and M < N;

Quantization modules, for using quantization function to carry out amplitude quantizing to the M of a described sampled signal sampled value, obtains quantized signal, and described quantized signal is the column vector being comprised of M quantized value;

Judge module, for judging respectively whether M quantized value surpasses preset quantization error tolerance;

Acquisition module, for obtain m the quantized value that does not surpass quantization error tolerance from a described M quantized value, according to all described m quantized values, described sampled signal and described sampling matrix generate sampling and retain matrix, wherein, described m quantized value is in a described M quantized value, not surpass a quantized value of quantization error tolerance, and in a described M quantized value, at least there is a described m quantized value, described m quantized value is corresponding with m sampled value in described sampled signal, the row vector that described m the sampled value m by described sampling matrix is comprised of N row sampling matrix value is multiplied by described input signal and obtains, described sampling retains m the row vector being comprised of N row sampling matrix value that matrix comprises all described sampling matrixs,

Sending module, for quantification stick signal and described sampling reservation matrix are sent to signal reconstruction device, described quantification stick signal comprises all described m quantized values.

Fourth aspect, the embodiment of the present invention provides a kind of signal reconstruction device, comprising:

Receiver module, the quantification stick signal and the sampling reservation matrix that for receiving Signal Compression device, send, wherein, described quantification stick signal comprises m all quantized values, described m quantized value is in M quantized value, not surpass a quantized value of quantization error tolerance, a described M quantized value is that described Signal Compression device is used sampling matrix to carry out after low speed sampling acquisition sampled signal input signal, use quantization function sampled signal to be carried out to the result of amplitude quantizing, the number of the row vector that described M is described sampling matrix;

Rebuild module, for carrying out signal reconstruction according to the quantification stick signal receiving and sampling reservation matrix, obtain signal reconstruction result;

Output module, for exporting described signal reconstruction result.

The 5th aspect, the embodiment of the present invention provides a kind of signal processing system, comprise: the Signal Compression device described in the aforementioned third aspect and the signal reconstruction device described in aforementioned fourth aspect, wherein, Signal Compression device with signal reconstruction device being connected by communication mode.

As can be seen from the above technical solutions, the embodiment of the present invention has the following advantages:

Therefore, in the embodiment of the present invention, first use sampling matrix to carry out low speed sampling to input signal, obtain sampled signal, then use quantization function to carry out amplitude quantizing to the M of a sampled signal sampled value, obtain quantized signal, next judge whether M quantized value in quantized signal has surpassed quantization error tolerance, from M quantized value, choose m the quantized value that does not surpass quantization error tolerance, according to m all quantized values, sampled signal and sampling matrix generate sampling and retain matrix, the quantification stick signal finally m the quantized value by all being formed and sampling retain matrix and send to signal reconstruction device.Due to what send to signal reconstruction device, be not surpass the quantized value of quantization error tolerance, and quantize second-rate quantized value, do not send to signal reconstruction device, therefore what send to signal reconstruction device is all to quantize the higher quantized value of quality, and according to the quantized value that does not surpass quantization error tolerance, former sampling matrix is screened, generate sampling and retain matrix, after receiving not over the quantized value of quantization error tolerance and the reservation matrix of sampling, signal reconstruction device carries out signal reconstruction, can reduce the impact of quantization error on signal reconstruction result, improve the accuracy of signal reconstruction.In addition, due to what send to signal reconstruction device, be not surpass the quantized value of quantization error tolerance, and quantize second-rate quantized value, do not send to signal reconstruction device, therefore saved, Signal Compression device is stored, the expense of transmission quantized value.

Accompanying drawing explanation

In order to be illustrated more clearly in the technical scheme in the embodiment of the present invention, below the accompanying drawing of required use during embodiment is described is briefly described, apparently, accompanying drawing in the following describes is only some embodiments of the present invention, to those skilled in the art, can also obtain according to these accompanying drawings other accompanying drawing.

The process blocks schematic diagram of a kind of compression method that Fig. 1 provides for the embodiment of the present invention;

The process blocks schematic diagram of a kind of signal reconfiguring method that Fig. 2 provides for the embodiment of the present invention;

The CS coding side that Fig. 3 provides for the embodiment of the present invention and the reciprocal process schematic diagram between CS decoding end;

The data point distribution schematic diagram of the original input signal that Fig. 4-a provides for the embodiment of the present invention;

The data point distribution schematic diagram of Fig. 4-b for obtaining according to the implementation reconstruction signal of prior art;

The data point distribution schematic diagram of Fig. 4-c for obtaining according to the implementation reconstruction signal of the embodiment of the present invention;

Fig. 5 for the embodiment of the present invention provide in Monte-Carlo Simulation repeatedly under different quantizing bit number scenes the embodiment of the present invention than prior art scheme the lifting situation schematic diagram in signal reconstruction performance;

The composition structural representation of a kind of Signal Compression device that Fig. 6 provides for the embodiment of the present invention;

The composition structural representation of a kind of signal reconstruction device that Fig. 7 provides for the embodiment of the present invention;

The composition structural representation of a kind of signal processing system that Fig. 8 provides for the embodiment of the present invention;

The composition structural representation of the another kind of Signal Compression device that Fig. 9 provides for the embodiment of the present invention;

The composition structural representation of the another kind of signal reconstruction device that Figure 10 provides for the embodiment of the present invention.

Embodiment

The embodiment of the present invention provides a kind of compression method and reconstructing method and relevant apparatus and system, for reducing the impact of quantization error on signal reconstruction, improves the accuracy of signal reconstruction.

For making goal of the invention of the present invention, feature, advantage can be more obvious and understandable, below in conjunction with the accompanying drawing in the embodiment of the present invention, technical scheme in the embodiment of the present invention is clearly and completely described, obviously, the embodiments described below are only the present invention's part embodiment, but not whole embodiment.Embodiment based in the present invention, the every other embodiment that those skilled in the art obtains, belongs to the scope of protection of the invention.

Below be elaborated respectively.

The embodiment of the present invention provides a kind of signal processing system, in this system, comprise Signal Compression device and signal reconstruction device, Signal Compression device is by after input signal compression, by the transmission channel between Signal Compression device and signal reconstruction device, Signal Compression result is sent to signal reconstruction device, signal reconstruction device carries out signal reconstruction according to Signal Compression result.Concrete, the signal processing system that the embodiment of the present invention provides can be applied to compressed sensing (CS, Compressed Sensing) in technology, wherein Signal Compression device just can be used as CS coding side, signal reconstruction device just can be used as CS decoding end, and the embodiment that provides as follows and the detailed description in application scenarios are provided its specific implementation.

An embodiment of compression method of the present invention can comprise following method: use sampling matrix to carry out low speed sampling to input signal, obtain sampled signal, wherein, described sampling matrix comprises M the row vector being comprised of N row sampling matrix value, described input signal is the column vector being comprised of N input signal values, described sampled signal is the column vector being comprised of M sampled value, the natural number that described M, N are non-zero, and M < N, use quantization function to carry out amplitude quantizing to the M of a described sampled signal sampled value, obtain quantized signal, described quantized signal is the column vector being comprised of M quantized value, judge respectively whether M quantized value surpasses preset quantization error tolerance, from a described M quantized value, obtain m the quantized value that does not surpass quantization error tolerance, according to all described m quantized values, described sampled signal and described sampling matrix generate sampling and retain matrix, wherein, described m quantized value is in a described M quantized value, not surpass a quantized value of quantization error tolerance, and in a described M quantized value, at least there is a described m quantized value, described m quantized value is corresponding with m sampled value in described sampled signal, described m sampled value is multiplied by described input signal by the m in described sampling matrix the row vector being comprised of N row sampling matrix value and obtains, described sampling retains matrix and comprises all m the row vector being comprised of N row sampling matrix value, quantification stick signal and described sampling reservation matrix are sent to signal reconstruction device, and described quantification stick signal comprises all described m quantized values.

Refer to shown in Fig. 1, a kind of compression method that one embodiment of the invention provides, can following steps comprise:

101, use sampling matrix to carry out low speed sampling to input signal, obtain sampled signal.

Wherein, above-mentioned sampling matrix comprises M the row vector being comprised of N row sampling matrix value, and above-mentioned input signal is the column vector being comprised of N input signal values, and above-mentioned sampled signal is the column vector being comprised of M sampled value, M, N are the natural number of non-zero, and M < N.

In some embodiments of the invention, input signal can be expressed as the column vector of N * 1, wherein each element in column vector is referred to as input signal values, sampling matrix can be expressed as the matrix of M * N, M, N are the natural number of non-zero, and wherein each element in sampling matrix is referred to as sampling matrix value, due to what input signal was carried out, are low speed samplings, also referred to as down-sampled, therefore M < N.Using sampling matrix to carry out low speed sampling to input signal is exactly that all M row vectors of sampling matrix and this column vector of input signal are multiplied each other in fact, the corresponding sampled value of multiplied result of each row vector and input signal, all M multiplied result form sampled signal, sampled signal can be expressed as the column vector of M * 1, and wherein each element in column vector is sampled value or claims sampled point.

102, use quantization function to carry out amplitude quantizing to the M of an above-mentioned sampled signal sampled value, obtain quantized signal.

Wherein, above-mentioned quantized signal is the column vector being comprised of M quantized value.

In some embodiments of the invention, the sampled signal representing for the column vector with M * 1 is carried out amplitude quantizing, can obtain quantized signal, and this quantized signal can be expressed as the column vector of M * 1, and wherein each element in column vector is referred to as quantized value.

In addition, the quantized level using in above-mentioned quantization function can be 2 t power, and wherein, t is quantizing bit number.For example, quantizing bit number is that 2 quantification progression are 4=2 ², quantized value represents with 2 bits, can represent altogether 4 kinds of amplitude state.

It should be noted that, amplitude quantizing operation is original amplitude to be similar to the operation (former amplitude is shone upon to immediate quantized level) of value, this approximate value operation is irreversible, the amplitude after quantizing cannot revert to original amplitude (amplitude before quantification), so in signal reconstruction device, the quantized data that utilizes process to quantize (approximate value) carries out signal reconstruction reconstruction errors can occur, and quantizing bit number is fewer, quantized level is fewer, the potential quantization error that approximate value operation causes is larger, so signal reconstruction accuracy is poorer.In addition, the scale of quantizer is to increase with the increase exponentially of its quantizing bit number, and when quantizing bit number increases, the complexity of quantizer can increase fast.After sampled signal obtains, along with the increase of the quantizing bit number of quantizer, the quantized data amount after quantizing will be exponent increase, so the data volume that sends to signal reconstruction device from Signal Compression device will increase thereupon, cause transport overhead to increase.

103, judge respectively whether M quantized value surpasses preset quantization error tolerance.

Wherein, above-mentioned quantization error tolerance can be used for the quantification quality that judgement is carried out amplitude quantizing to above-mentioned sampled signal.

In some embodiments of the invention, quantize quality and can weigh the quality of sampled value being carried out to amplitude quantizing, if the amplitude of the sampled value before the amplitude of the quantized value after quantizing and quantification approaches very much, just can think that quantification quality is better, if the amplitude of the sampled value before the amplitude of the quantized value after quantizing and quantification differs, just can think that greatly quantification is second-rate.Wherein, preset quantization error tolerance is for judging the quantification quality of above-mentioned amplitude quantizing, in actual applications, if quantized value A surpasses quantization error tolerance, think that its quantification is second-rate, if quantized value B surpasses quantization error tolerance, think that its quantification quality is better.Preset quantization error tolerance specifically can have one or more criterion and judgment mode, specifically can set in conjunction with actual application scenarios, as long as can screen M quantized value by quantizing fault tolerance, therefrom select the quantized value that does not surpass quantization error tolerance.For example, if quantization error tolerance can surpass a threshold value for the difference of the sampled value before the quantification corresponding with it of quantized value, this quantized value surpasses quantization error tolerance; Quantization error tolerance also can for the sampled value before two adjacent quantifications, all nearest two quantized levels be far away, in two sampled values, larger sampled value is less than the first higher quantized level, less sampled value is greater than the second lower quantized level, so these two sampled values are all away from quantized level, two quantized values that obtain after amplitude quantizing all surpass quantization error tolerance.Multiple implementation just illustrates above, might not be applicable to a certain concrete application scenarios, specifically can set quantification fault tolerance in conjunction with actual application scenarios, does not limit herein.

In addition, in some embodiments of the invention, can be according to above-mentioned quantization error tolerance generating quantification error tolerance thresholding, judge respectively whether M quantification difference is greater than quantization error tolerance thresholding, for the first quantized value corresponding to quantification difference that is greater than quantization error tolerance thresholding, surpassed quantization error tolerance, second quantized value corresponding for the quantification difference that is less than or equal to quantization error tolerance thresholding do not surpass quantization error tolerance, wherein, above-mentioned quantification difference is the absolute value of the difference between the first sampled value and above-mentioned the first quantized value, or the absolute value of the second difference between sampled value and above-mentioned the second quantized value, above-mentioned the first quantized value is for to carry out to above-mentioned the first sampled value the quantized value that amplitude quantizing obtains, above-mentioned the first quantized value is corresponding with above-mentioned the first sampled value, above-mentioned the second quantized value is for to carry out to above-mentioned the second sampled value the quantized value that amplitude quantizing obtains, above-mentioned the second quantized value is corresponding with above-mentioned the second sampled value, that is to say whether M the quantized value that quantized signal is comprised is greater than quantization error tolerance thresholding according to quantification difference and divides the quantized value of two classes, one class is the first quantized value, another kind of is the second quantized value, the first quantized value is to be greater than quantization error to tolerate that the quantification difference of thresholding is corresponding, the second quantized value is to be less than or equal to quantization error to tolerate that the quantification difference of thresholding is corresponding.For example, quantization error tolerance is that the difference of two range values before and after amplitude quantizing can not surpass a predetermined threshold value, if having surpassed default thresholding thinks and has surpassed quantization error tolerance, if not surpassing default thresholding thinks not over quantization error tolerance, according to this quantization error tolerance, first generate a quantization error tolerance thresholding (being predetermined threshold value η), calculate respectively difference between the sampled value before M quantized value and quantification, for quantizing difference, total M, for example, M sampled value is respectively A ₁, A ₂..., A _m, M quantized value is respectively B ₁, B ₂..., B _m, quantized value B ₁to sampled value A ₁carry out the quantized value obtaining after amplitude quantizing, same, quantized value B _mto sampled value A _mcarry out the quantized value obtaining after amplitude quantizing, quantized value B ₁with sampled value A ₁between quantification difference C ₁=∣ B ₁-A ₁∣, same, quantized value B _mwith sampled value A _mbetween quantification difference C _m=∣ B _m-A _m∣, then judges respectively that this M quantification difference (is respectively C ₁, C ₂..., C _m) whether be greater than quantization error tolerance thresholding η, if only have a quantification difference, be C _i(i is the natural number that is less than M, C _ic ₁, C ₂..., C _min i value) be greater than quantization error tolerance thresholding, think quantized value B _isurpassed quantization error tolerance, if quantized value B ₁, B ₂..., B _i-1, B _i+1, B _mall do not surpass quantization error tolerance, therefore whether surpass preset quantization error tolerance according to quantized value, the quantized value of the M in quantized signal is divided into the value of two types, surpass the quantized value B of quantization error tolerance _i, can not be transmitted to signal reconstruction device, in order to avoid the excessive accuracy that affects signal reconstruction of quantization error, and do not surpass the quantized value B of quantization error tolerance ₁, B ₂..., B _{i -1}, B _i+1, B _mcan be transmitted to signal reconstruction device, signal reconstruction device carries out signal reconstruction according to the quantized value receiving, and can improve the accuracy of signal reconstruction.

104, from an above-mentioned M quantized value, obtain m the quantized value that does not surpass quantization error tolerance, according to all above-mentioned m quantized values, above-mentioned sampled signal and above-mentioned sampling matrix, generate sampling and retain matrix.

Wherein, above-mentioned m quantized value is in an above-mentioned M quantized value, not surpass a quantized value of quantization error tolerance, and in an above-mentioned M quantized value, at least there is an above-mentioned m quantized value, that is to say, in the embodiment of the present invention, the quantized value that does not surpass quantization error tolerance in M quantized value is all defined as to m quantized value, the quantized value that does not surpass quantization error tolerance in M quantized value has when a plurality of, " m " represents a plurality of quantized values, the quantized value that does not for example surpass quantization error tolerance in M quantized value is the 2nd quantized value, the 4th quantized value, the 7th quantized value, the value of " m " can be just 2, 4, 7, therefore m quantized value just can represent the 2nd quantized value, the 4th quantized value, the 7th quantized value.

Above-mentioned m quantized value is corresponding with m sampled value in above-mentioned sampled signal, the row vector that above-mentioned m the sampled value m by sampling matrix is comprised of N row sampling matrix value is multiplied by above-mentioned input signal and obtains, and to retain matrix be to comprise the individual row vector being comprised of N row sampling matrix value of m in above-mentioned sampling.What the above-mentioned explanation that retains matrix to sampling referred in fact is exactly to generate the process that sampling retains matrix, first according to m quantized value, from sampled signal, find corresponding m sampled value, then according to this m sampled value, from sampling matrix, find for multiplying each other and obtain the row vector of this m sampled value with above-mentioned input signal, be defined as m the row vector being formed by N row sampling matrix value, owing at least there being a m quantized value in M quantized value, that is to say and in M quantized value, include a plurality of m quantized values, due to corresponding m the sampled value of each m quantized value, and each m sampled value row vector that also a corresponding m is comprised of N row sampling matrix value, therefore m the row vector being comprised of N row sampling matrix value equally also has a plurality of, these all m the row vectors that are comprised of N row sampling matrix value are combined, the sampling that just obtains the present invention's proposition has retained matrix.

In some embodiments of the invention, m above-mentioned quantized value specifically refers to the quantized value that in M quantized value, ID is m, it should be noted that, in embodiments of the present invention the quantized value that does not surpass quantization error tolerance in M quantized value is referred to as to m quantized value, therefore the value of m is not unique, and may be that a value may be also a plurality of values, this depends in M quantized value has how many quantized values not surpass quantization error tolerance.Still take aforesaid example as example explanation, quantized value B ₁, B ₂..., B _i-1, B _i+1, B _mm the quantized value getting all do not surpass quantization error tolerance, therefore just can refer to quantized value B from M ₁, B ₂..., B _i-1, B _i+1, B _min a quantized value, in M quantized value, include (M-1) individual m such quantized value.After obtaining m quantized value, just can individual above-mentioned m quantized value, above-mentioned sampled signal and above-mentioned sampling matrix generate sampling reservation matrix according to all (M-1).For example, quantized value B ₁, B ₂..., B _i-1, B _i+1, B _mcorresponding sampled value is A respectively ₁, A ₂..., A _i-1, A _i+1, A _m.

Suppose that sampling matrix is $\left[\begin{array}{cccc}{X}_{11}& X_{12}& ...& X_{1N}\\ X_{21}& X_{22}& ...& X_{2N}\\ ...& ...& ...& ...\\ X_{M1}& X_{M2}& ...& X_{\mathrm{MN}}\end{array}\right],$ Suppose that input signal is θ, for sampled value A ₁, be the row vector [X by being formed by N row sampling matrix value ₁₁x ₁₂x _1N] be multiplied by input signal θ and obtain, for sampled value A ₂, be the row vector [X by being formed by N row sampling matrix value ₂₁x ₂₂x _2N] be multiplied by input signal θ and obtain ..., for sampled value A _m, be the row vector [X by being formed by N row sampling matrix value _m1x _m2x _mN] be multiplied by input signal θ and obtain, by removing i these row vectors beyond capable, combine, just can obtain sampling reservation matrix as follows:

$\left[\begin{array}{cccc}{X}_{11}& X_{12}& ...& X_{1N}\\ X_{21}& X_{22}& ...& X_{2N}\\ ...& ...& ...& ...\\ X_{(i-1)1}& X_{(i-1)2}& ...& X_{(i-1)N}\\ X_{(i+1)1}& X_{(i+1)2}& ...& X_{(i+1)N}\\ ...& ...& ...& ...\\ X_{M1}& X_{M2}& ...& X_{\mathrm{MN}}\end{array}\right].$

In addition, in some embodiments of the invention, can obtain in sampled signal for obtaining m sampled value of m quantized value according to above-mentioned m quantized value; According to m sampled value, obtain in above-mentioned sampling matrix m the row vector being formed by N row sampling matrix value of sampling and using while calculating above-mentioned m sampled value by low speed; All m the row vector being comprised of N row sampling matrix value combined, obtain described sampling and retain matrix, or, other row vectors that are comprised of N row sampling matrix value except all above-mentioned m the row vector being comprised of N row sampling matrix value in above-mentioned sampling matrix are rejected, again the remaining row vector being comprised of N row sampling matrix value in above-mentioned sampling matrix is combined, obtained above-mentioned sampling and retain matrix.

In addition, in some embodiments of the invention, can also obtain n quantized value according to above-mentioned m quantized value, above-mentioned n quantized value is a quantized value except all above-mentioned m quantized values in an above-mentioned M quantized value, and above-mentioned n quantized value is in an above-mentioned M quantized value, to surpass a quantized value of quantization error tolerance; According to above-mentioned n quantized value, obtain in sampled signal for obtaining n sampled value of n quantized value; According to above-mentioned n sampled value, obtain in above-mentioned sampling matrix n the row vector being formed by N row sampling matrix value of using while calculating above-mentioned n sampled value sampling by low speed; Other row vectors that are comprised of N row sampling matrix value except all above-mentioned n the row vector being comprised of N row sampling matrix value in above-mentioned sampling matrix are combined, obtain above-mentioned sampling and retain matrix, or, above-mentioned n in above-mentioned sampling matrix the row vector being comprised of N row sampling matrix value rejected, again the remaining row vector being comprised of N row sampling matrix value in above-mentioned sampling matrix is combined, obtained above-mentioned sampling and retain matrix.

105, the quantification stick signal above-mentioned m the quantized value by all being formed and above-mentioned sampling retain matrix and send to signal reconstruction device.

Wherein, after getting quantification stick signal and above-mentioned sampling reservation matrix, Signal Compression device can send to signal reconstruction device by quantification stick signal and sampling reservation matrix by the transmission channel between Signal Compression device and signal reconstruction device, and signal reconstruction device retains matrix according to the quantification stick signal receiving and sampling and carries out signal reconstruction.

Therefore, in some embodiments of the present invention, first use sampling matrix to carry out low speed sampling to input signal, obtain sampled signal, then use quantization function to carry out amplitude quantizing to the M of a sampled signal sampled value, obtain quantized signal, next judge whether M quantized value in quantized signal has surpassed quantization error tolerance, from M quantized value, choose m the quantized value that does not surpass quantization error tolerance, according to m all quantized values, sampled signal and sampling matrix generate sampling and retain matrix, the quantification stick signal finally m the quantized value by all being formed and sampling retain matrix and send to signal reconstruction device.Due to what send to signal reconstruction device, be not surpass the quantized value of quantization error tolerance, and quantize second-rate quantized value, do not send to signal reconstruction device, therefore what send to signal reconstruction device is all to quantize the higher quantized value of quality, and according to the quantized value that does not surpass quantization error tolerance, former sampling matrix is screened, generate sampling and retain matrix, after receiving not over the quantized value of quantization error tolerance and the reservation matrix of sampling, signal reconstruction device carries out signal reconstruction, can reduce the impact of quantization error on signal reconstruction result, improve the accuracy of signal reconstruction.In addition, due to what send to signal reconstruction device, be not surpass the quantized value of quantization error tolerance, and quantize second-rate quantized value, do not send to signal reconstruction device, therefore saved, Signal Compression device is stored, the expense of transmission quantized value.

Above embodiment has introduced compression method provided by the invention, next introduce the signal reconfiguring method that the embodiment of the present invention provides, this signal reconfiguring method can comprise following method: receive quantification stick signal and sampling reservation matrix that Signal Compression device sends, wherein, quantize stick signal and comprise m all quantized values, described m quantized value is in M quantized value, not surpass a quantized value of quantization error tolerance, a described M quantized value is that described Signal Compression device is used sampling matrix to carry out after low speed sampling obtains sampled signal using quantization function sampled signal to be carried out to the result of amplitude quantizing to input signal, described M is the number of the row vector of described sampling matrix, according to the quantification stick signal receiving and sampling reservation matrix, carry out signal reconstruction, obtain signal reconstruction result, export described signal reconstruction result.

Refer to shown in Fig. 2, a kind of signal reconfiguring method that one embodiment of the invention provides, can comprise:

201, receive quantification stick signal and the sampling reservation matrix that Signal Compression device sends.

Wherein, above-mentioned quantification stick signal comprises m all quantized values, above-mentioned m quantized value is in M quantized value, not surpass all quantized values of quantization error tolerance, an above-mentioned M quantized value is that above-mentioned Signal Compression device is used sampling matrix to carry out after low speed sampling acquisition sampled signal input signal, use quantization function sampled signal to be carried out to the result of amplitude quantizing, the number of the row vector that above-mentioned M is above-mentioned sampling matrix.

In some embodiments of the invention, Signal Compression device can send to signal reconstruction device by Signal Compression result by the transmission channel between Signal Compression device and signal reconstruction device, and first signal reconstruction device receives by transmission channel quantification stick signal and the sampling reservation matrix that Signal Compression device sends.For example, Signal Compression device just can be used as CS coding side, and signal reconstruction device just can be used as CS decoding end, and CS decoding end receives and quantizes stick signal and sampling reservation matrix from CS coding side.

202, according to the quantification stick signal receiving and sampling reservation matrix, carry out signal reconstruction, obtain signal reconstruction result.

In some embodiments of the invention, signal reconstruction device can retain matrix setting constraints according to above-mentioned quantification stick signal and sampling; Then, calculate the column vector while making target function reach minimum value under above-mentioned constraints, wherein, above-mentioned column vector is signal reconstruction result.

In addition, according to above-mentioned quantification stick signal and sampling reservation matrix setting constraints, specifically can obtain in the following way:

$s.t.:{\left|\right|\tilde{y}-\widetilde{\mathrm{\&Phi;}}\mathrm{\&theta;}\left|\right|}_2\&le;\mathrm{\&epsiv;},$

Wherein, above-mentioned s.t. represents constraints, above-mentioned for described quantification stick signal, above-mentioned for above-mentioned sampling retains matrix, above-mentioned θ is column vector to be reconstructed, and above-mentioned ε is preset error constraints parameter.

Column vector when in addition, above-mentioned calculating makes target function reach minimum value under above-mentioned constraints specifically can be obtained in the following way:

$\widehat{\mathrm{\&theta;}}=\underset{\mathrm{\&theta;}}{\arg \min }\left\{{\left|\right|\mathrm{\&theta;}\left|\right|}_1\right\}$ ，

$s.t.:{\left|\right|\tilde{y}-\widetilde{\mathrm{\&Phi;}}\mathrm{\&theta;}\left|\right|}_2\&le;\mathrm{\&epsiv;}$

Wherein, above-mentioned $\begin{array}{c}\widehat{\mathrm{\&theta;}}=\underset{\mathrm{\&theta;}}{\arg \min }\{\&CenterDot;\}\\ s.t.,...\end{array}$ For solving, reach the column vector of minimum value meeting the above-mentioned constraints target function of ordering

203, export above-mentioned signal reconstruction result.

Wherein, signal reconstruction device, after getting signal reconstruction result, is exported this signal reconstruction result, and this signal reconstruction result is preserved so that other follow-up coherent signals are processed.

Therefore, in some embodiments of the present invention, first receive quantification stick signal and sampling reservation matrix that Signal Compression device sends, quantize stick signal and comprise m all quantized values, m quantized value for choosing the quantized value that does not surpass quantization error tolerance from M quantized value, and it is according to m all quantized values that sampling retains matrix, sampled signal and sampling matrix generate, due to what send to signal reconstruction device, be not surpass the quantized value of quantization error tolerance, and quantize second-rate quantized value, do not send to signal reconstruction device, therefore what send to signal reconstruction device is all to quantize the higher quantized value of quality, and according to the quantized value that does not surpass quantization error tolerance, former sampling matrix is screened, generate sampling and retain matrix, signal reconstruction device receives not over carrying out signal reconstruction after the quantized value of quantization error tolerance and the reservation matrix of sampling, can reduce the impact of quantization error on signal reconstruction result, improve the accuracy of signal reconstruction.In addition, due to what send to signal reconstruction device, be not surpass the quantized value of quantization error tolerance, and quantize second-rate quantized value, do not send to signal reconstruction device, therefore signal reconstruction device only need to receive, quantize the good quantized value of quality, saved storage, transmitted the expense of quantized value.

For ease of better understanding and implement the such scheme of the embodiment of the present invention, an application scenarios is specifically described for example below.

Below the system architecture of mainly take based on aforesaid signal processing system be example, the Signal Compression device that signal processing system comprises be take its specific implementation as CS coding side example, and the signal reconstruction device that signal processing system comprises be take its specific implementation as CS decoding end example.

Refer to the schematic flow sheet of the compression method shown in Fig. 3 and signal reconfiguring method, wherein step 301 is to 307 being the compression method that CS coding side one side realizes, and step 308 is to 310 be the signal reconfiguring method of CS decoding end one side realization.

301, after system powers on and starts, first input signal is input to CS coding side.

Wherein, input signal can be expressed as the input signal θ that signal dimension is N (being expressed as the column vector θ of N * 1).

302, CS coding side carries out low speed sampling to input signal, obtains sampled signal.

Wherein, input signal θ is carried out to low speed sampling, this process can be expressed as the form of sampling matrix and input signal product:

m=Φθ （1）

Wherein, in formula (1), θ is the column vector of corresponding N * 1 of input signal, Φ is that (in CS technology, signal sampling process is the process of down-sampled (low speed sampling) to M * N sampling matrix, be M < N), m is the sampled signal (being sampled result column vector) of M * 1 through obtaining after low speed sampling.In other words, each sampled value in m (being each element in sampled result column vector) is that row vector in sampling matrix Φ and this column vector of input signal θ multiply each other and obtain, and these row vectors form sampling matrix Φ.

303, CS coding side carries out amplitude quantizing to the sampled value of sampled signal, obtains quantized signal.

Wherein, sampled signal m is quantized to obtain quantized signal y:

y=Q(m) （2）

Wherein, in formula (2), Q () is quantization function, and sampled signal is carried out to quantization operation, each element value in sampled result column vector m is selected to an immediate quantized level nearby and using this as new element value corresponding in quantized signal y, i.e. quantized value.In actual quantization operation, because the quantizing bit number of quantizer is limited, so quantized level (carrying out the amplitude of approximate representation sampled data by how many centrifugal pumps) is also limited, the numerical relation of quantization bit and quantized level is: quantize the quantizing bit number power that progression equals 2.For example, quantizing bit number is that 2 quantification progression are 4=2 ².

304, judge whether quantized value is greater than quantization error tolerance thresholding, if be less than or equal to execution step 305, if be greater than execution step 307.

305, CS coding side retains the quantized value that is less than or equal to quantization error tolerance thresholding, then performs step 306.

307, CS coding side abandons being greater than the quantized value of quantization error tolerance thresholding.

Wherein, quantize difference and predetermined threshold value is relatively adjudicated, i.e. each quantized value y relatively _i(being i the element of quantized result column vector y) with quantize before corresponding sampled value m _iwhether the difference of (for i the element of sampled signal column vector m) is greater than default quantization error tolerance thresholding η, this predetermined threshold value can be set according to many practical factors such as quantizing bit number and system tolerable quantization errors, in the embodiment of the present invention, this is not limited, when the absolute value of both differences is greater than default thresholding η, | y _i-m _i| > η, abandon this quantized value, when the absolute value of both differences is not more than default thresholding η, | y _i-m _i|≤η, retains this quantized value.

306, CS coding side obtains corresponding sampling reservation matrix according to the quantized value retaining, and sends to CS decoding end.

By by the quantification stick signal that forms of quantized value with a grain of salt and corresponding sampling retains matrix send to CS decoding end, wherein, the corresponding column vector forming for those elements that are retained after step 305 in column vector y, for the sampling forming corresponding to those row vectors of retained quantized value in sampling matrix Φ retains the matrix (quantized value that only has those to be retained and the sampling that in sampling matrix, corresponding line vector forms retains matrix just being sent to CS decoding end rebuilds for follow-up signal).

Next the signal reconfiguring method that the embodiment of the present invention provides is described:

308, CS decoding termination is received the quantized value of reservation and corresponding sampling reservation matrix.

Wherein, CS decoding end specifically receives that CS coding side sends with

309, CS decoding end carries out signal reconstruction according to the quantized value of the reservation receiving and corresponding sampling reservation matrix.

Wherein, CS decoding end obtains according to reception with , carry out CS signal reconstruction, in CS technology, to the signal reconstruction of unknown input signal vector, can be to realize by solving at the constraints vector that target function reaches minimum value of ordering:

$\widehat{\mathrm{\&theta;}}=\underset{\mathrm{\&theta;}}{\arg \min }\left\{{\left|\right|\mathrm{\&theta;}\left|\right|}_1\right\}$ （3）

$s.t.:{\left|\right|\tilde{y}-\widetilde{\mathrm{\&Phi;}}\mathrm{\&theta;}\left|\right|}_2\&le;\mathrm{\&epsiv;}$

Wherein, in above-mentioned formula (3), the content in { } is target function, s.t. heel constraints, $\begin{array}{c}\widehat{\mathrm{\&theta;}}=\underset{\mathrm{\&theta;}}{\arg \min }\{\&CenterDot;\}\\ s.t.,...\end{array}$ For solving, reach the unknown vector of minimum value meeting the constraints target function of ordering in target function || || ₁a norm that represents vector, the object that minimizes a norm is the sparse property of utilizing signal θ to be reconstructed to have, in constraints || || ₂two norms that represent vector, two norm constraint conditions are that constraint is retained the error in quantized result, ε is default error constraints parameter.

310, CS decoding end output signal reconstructed results.

Wherein, CS decoding end output signal reconstructed results .

Therefore, in embodiment provided by the invention, reduced the potential impact of quantization error to CS signal reconstruction, and adopt the mechanism that quantitative differences is relatively adjudicated to screen quantized data, thereby retained quantized data that quality is higher for signal reconstruction, can improve the signal reconstruction accuracy that quantizes compressed sensing, meanwhile, owing to only retaining, send, receiving, quantize the higher quantized value of quality, obviously can save the expenses such as storage, transmission, processing.

For further illustrate the embodiment of the present invention with respect to prior art scheme at the remarkable result aspect signal reconstruction accuracy, the signal reconstruction performance gain of comparison and the analysis embodiment of the present invention will be carried out by emulation experiment below, prior art refers to and adopts CS technology on ignoring the impact of quantization operation on CS technology in the method for signal reconstruction herein, directly the sampled data after quantizing is used for to signal reconstruction.Wherein, signal reconstruction performance is characterized by signal reconstruction signal to noise ratio (SNR, Signal-to-Noise Ratio), and the numerical value of this signal to noise ratio is larger, represents that signal reconstruction method is better to the reconstruction performance of signal, specifically can use following formula (4):

$\mathrm{\&gamma;}=10\lg \left(\frac{{\left|\right|\mathrm{\&theta;}\left|\right|}_2^2}{{\left|\right|\mathrm{\&theta;}-\widehat{\mathrm{\&theta;}}\left|\right|}_2^2}\right)---\left(4\right)$

Wherein, molecular moiety in formula (4) for two norms of original sparse signal vector square, characterize the energy of original input signal; Denominator part for two norms of the error vector between original sparse signal vector and reconstruction signal vector square, represent the energy of reconstruction signal error.

Refer to Fig. 4-a, Fig. 4-b, Fig. 4-c is with the explanation embodiment of the present invention and the accuracy of prior art when to signal reconstruction, the data point distribution schematic diagram that wherein Fig. 4-a is original input signal, in Fig. 4-a, each circle represents a data point of original input signal, 5 circles that upper and lower sides distributes in Fig. 4-a represent the concrete value of nonzero element in sparse signal, the circle that a large amount of values that distribute in Fig. 4-a are 0 represents in sparse signal it except a small amount of nonzero element, is all that value is 0 data point, Fig. 4-a generates by experiment simulation method, this figure laterally represents that (degree of rarefication is 5 i.e. 5 nonzero elements for the value of each element in an input signal vector (column vector) here, other elements are 0).The data point distribution schematic diagram of Fig. 4-b for obtaining according to the implementation reconstruction signal of prior art, the data point distribution schematic diagram of Fig. 4-c for obtaining according to the implementation reconstruction signal of the embodiment of the present invention.More known by 3 accompanying drawings, the embodiment of the present invention has and improves intuitively effect in signal reconstruction accuracy with respect to prior art scheme, so the signal reconstruction result of the embodiment of the present invention approaches original input signal more than the signal reconstruction result of prior art.

Therefore, Fig. 4-a, Fig. 4-b, Fig. 4-c is only the result comparison to an emulation experiment, for embodiment of the present invention lifting in signal reconstruction accuracy than prior art is described from the statistical significance, thereby further illustrate the beneficial effect of the embodiment of the present invention, refer to shown in Fig. 5, for the embodiment of the present invention under different quantizing bit number scenes in Monte-Carlo Simulation repeatedly than prior art scheme the lifting situation in signal reconstruction performance, simulation result from Fig. 5, under given quantizing bit number, the signal accuracy of the embodiment of the present invention is better than prior art scheme.

Therefore, in the embodiment of the present invention, first use sampling matrix to carry out low speed sampling to input signal, obtain sampled signal, then use quantization function to carry out amplitude quantizing to the M of a sampled signal sampled value, obtain quantized signal, next judge whether M quantized value in quantized signal has surpassed quantization error tolerance, from M quantized value, choose m the quantized value that does not surpass quantization error tolerance, according to m all quantized values, sampled signal and sampling matrix generate sampling and retain matrix, the quantification stick signal finally m the quantized value by all being formed and sampling retain matrix and send to signal reconstruction device.Due to what send to signal reconstruction device, be not surpass the quantized value of quantization error tolerance, and quantize second-rate quantized value, do not send to signal reconstruction device, therefore what send to signal reconstruction device is all to quantize the higher quantized value of quality, and according to the quantized value that does not surpass quantization error tolerance, former sampling matrix is screened, generate sampling and retain matrix, after receiving not over the quantized value of quantization error tolerance and the reservation matrix of sampling, signal reconstruction device carries out signal reconstruction, can reduce the impact of quantization error on signal reconstruction result, improve the accuracy of signal reconstruction.In addition, due to what send to signal reconstruction device, be not surpass the quantized value of quantization error tolerance, and quantize second-rate quantized value, do not send to signal reconstruction device, therefore saved, Signal Compression device is stored, the expense of transmission quantized value.

It should be noted that, for aforesaid each embodiment of the method, for simple description, therefore it is all expressed as to a series of combination of actions, but those skilled in the art should know, the present invention is not subject to the restriction of described sequence of movement, because according to the present invention, some step can adopt other orders or carry out simultaneously.Secondly, those skilled in the art also should know, the embodiment described in specification all belongs to preferred embodiment, and related action and module might not be that the present invention is necessary.

For ease of better implementing the such scheme of the embodiment of the present invention, be also provided for implementing the relevant apparatus of such scheme below.

Refer to as shown in Figure 6, a kind of Signal Compression device 600 that the embodiment of the present invention provides, can comprise: sampling module 601, quantization modules 602, judge module 603, acquisition module 604 and sending module 605, wherein,

Sampling module 601, be used for using sampling matrix to carry out low speed sampling to input signal, obtain sampled signal, wherein, described sampling matrix comprises M the row vector being comprised of N row sampling matrix value, and described input signal is the column vector being comprised of N input signal values, and described sampled signal is the column vector being comprised of M sampled value, described M, N are the natural number of non-zero, and M < N;

Quantization modules 602, for using quantization function to carry out amplitude quantizing to the M of a described sampled signal sampled value, obtains quantized signal, and described quantized signal is the column vector being comprised of M quantized value;

Judge module 603, for judging respectively whether M quantized value surpasses preset quantization error tolerance;

Acquisition module 604, for obtain m the quantized value that does not surpass quantization error tolerance from a described M quantized value, according to all described m quantized values, described sampled signal and described sampling matrix generate sampling and retain matrix, wherein, described m quantized value is in a described M quantized value, not surpass a quantized value of quantization error tolerance, and in a described M quantized value, at least there is a described m quantized value, described m quantized value is corresponding with m sampled value in described sampled signal, the row vector that described m the sampled value m by described sampling matrix is comprised of N row sampling matrix value is multiplied by described input signal and obtains, described sampling retains m the row vector being comprised of N row sampling matrix value that matrix comprises all described sampling matrixs,

Sending module 605, for quantification stick signal and described sampling reservation matrix are sent to signal reconstruction device, described quantification stick signal comprises all described m quantized values.

In some embodiments of the invention, judge module 603, specifically can comprise (not shown in Figure 6):

Generate submodule, for tolerating thresholding according to described quantization error tolerance generating quantification error;

Judgement submodule, for judging respectively whether M quantification difference is greater than quantization error tolerance thresholding, for the first quantized value corresponding to quantification difference that is greater than quantization error tolerance thresholding, surpassed quantization error tolerance, second quantized value corresponding for the quantification difference that is less than or equal to quantization error tolerance thresholding do not surpass quantization error tolerance, wherein, described quantification difference is the absolute value of the difference between the first sampled value and described the first quantized value, or the absolute value of the second difference between sampled value and described the second quantized value, described the first quantized value is for to carry out to described the first sampled value the quantized value that amplitude quantizing obtains, described the first quantized value is corresponding with described the first sampled value, described the second quantized value is for to carry out to described the second sampled value the quantized value that amplitude quantizing obtains, described the second quantized value is corresponding with described the second sampled value.

In some embodiments of the invention, acquisition module 604, specifically can comprise (not shown in Figure 6):

First obtains submodule, for obtaining described sampled signal for obtaining m sampled value of described m quantized value according to described m quantized value;

Second obtains submodule, for obtaining described sampling matrix according to described m sampled value in m the row vector being comprised of N row sampling matrix value of sampling by low speed and using while calculating described m sampled value;

The 3rd obtains submodule, for the row vector that all described m is comprised of N row sampling matrix value, combine, obtain described sampling and retain matrix, or, other row vectors that are comprised of N row sampling matrix value except all described m the row vector being comprised of N row sampling matrix value in described sampling matrix are rejected, again the remaining row vector being comprised of N row sampling matrix value in described sampling matrix is combined, obtained described sampling and retain matrix.

In some embodiments of the invention, acquisition module 604, specifically can comprise (not shown in Figure 6):

The 4th obtains submodule, for obtaining n quantized value according to all described m quantized values, described n quantized value is a quantized value except all described m quantized values in a described M quantized value, and described n quantized value is in a described M quantized value, to surpass a quantized value of quantization error tolerance;

The 5th obtains submodule, for obtaining described sampled signal for obtaining n sampled value of described n quantized value according to described n quantized value;

The 6th obtains submodule, for obtaining described sampling matrix according to described n sampled value in n the row vector being comprised of N row sampling matrix value of sampling by low speed and using while calculating described n sampled value;

The 7th obtains submodule, being used for other row vectors that are comprised of N row sampling matrix value except all described n the row vector being comprised of N row sampling matrix value by described sampling matrix combines, obtain described sampling and retain matrix, or, all described n in described sampling matrix the row vector being comprised of N row sampling matrix value rejected, again the remaining row vector being comprised of N row sampling matrix value in described sampling matrix is combined, obtained described sampling and retain matrix.

In addition, the t power that in the described quantization function that quantization modules 602 is used, quantized level is 2, described t is quantizing bit number.

Refer to shown in Fig. 7, a kind of signal reconstruction device 700 that the embodiment of the present invention provides, can comprise: receiver module 701, reconstruction module 702, output module 703, wherein,

Receiver module 701, the quantification stick signal and the sampling reservation matrix that for receiving Signal Compression device, send, wherein, described quantification stick signal comprises m all quantized values, described m quantized value is in M quantized value, not surpass a quantized value of quantization error tolerance, a described M quantized value is that described Signal Compression device is used sampling matrix to carry out after low speed sampling acquisition sampled signal input signal, use quantization function sampled signal to be carried out to the result of amplitude quantizing, the number of the row vector that described M is described sampling matrix;

Rebuild module 702, for carrying out signal reconstruction according to the quantification stick signal receiving and sampling reservation matrix, obtain signal reconstruction result;

Output module 703, for exporting described signal reconstruction result.

In some embodiments of the invention, rebuild module 702, specifically can comprise (not shown in Figure 7):

Obtain submodule, for retaining matrix setting constraints according to described quantification stick signal and sampling;

Calculating sub module, the column vector while making target function reach minimum value for calculating under described constraints, described column vector is signal reconstruction result.

Further, obtain submodule, specifically can be for obtaining in the following way constraints:

$s.t.:{\left|\right|\tilde{y}-\widetilde{\mathrm{\&Phi;}}\mathrm{\&theta;}\left|\right|}_2\&le;\mathrm{\&epsiv;},$

Wherein, described s.t. represents constraints, described in for described quantification stick signal, described in for described sampling retains matrix, described θ is column vector to be reconstructed, and described ε is preset error constraints parameter;

Calculating sub module, specifically can be for obtaining in the following way column vector:

$\widehat{\mathrm{\&theta;}}=\underset{\mathrm{\&theta;}}{\arg \min }\left\{{\left|\right|\mathrm{\&theta;}\left|\right|}_1\right\}$ ，

$s.t.:{\left|\right|\tilde{y}-\widetilde{\mathrm{\&Phi;}}\mathrm{\&theta;}\left|\right|}_2\&le;\mathrm{\&epsiv;}$

Wherein, described in $\begin{array}{c}\widehat{\mathrm{\&theta;}}=\underset{\mathrm{\&theta;}}{\arg \min }\{\&CenterDot;\}\\ s.t.,...\end{array}$ For solving, reach the column vector of minimum value meeting the described constraints target function of ordering

Refer to shown in Fig. 8, a kind of signal processing system 800 that the embodiment of the present invention provides, can comprise: Signal Compression device 600 as shown in the previous embodiment and signal reconstruction device 700 as shown in the previous embodiment, wherein, Signal Compression device 600 with signal reconstruction device 700 being connected by communication mode.

In the above-described embodiments, the description of each embodiment is all emphasized particularly on different fields, in certain embodiment, there is no the part of detailed description, can be referring to the associated description of other embodiment.

To sum up, in the embodiment of the present invention, first use sampling matrix to carry out low speed sampling to input signal, obtain sampled signal, then use quantization function to carry out amplitude quantizing to the M of a sampled signal sampled value, obtain quantized signal, next judge whether M quantized value in quantized signal has surpassed quantization error tolerance, from M quantized value, choose m the quantized value that does not surpass quantization error tolerance, according to m all quantized values, sampled signal and sampling matrix generate sampling and retain matrix, the quantification stick signal finally m the quantized value by all being formed and sampling retain matrix and send to signal reconstruction device.Due to what send to signal reconstruction device, be not surpass the quantized value of quantization error tolerance, and quantize second-rate quantized value, do not send to signal reconstruction device, therefore what send to signal reconstruction device is all to quantize the higher quantized value of quality, and according to the quantized value that does not surpass quantization error tolerance, former sampling matrix is screened, generate sampling and retain matrix, after receiving not over the quantized value of quantization error tolerance and the reservation matrix of sampling, signal reconstruction device carries out signal reconstruction, can reduce the impact of quantization error on signal reconstruction result, improve the accuracy of signal reconstruction.In addition, due to what send to signal reconstruction device, be not surpass the quantized value of quantization error tolerance, and quantize second-rate quantized value, do not send to signal reconstruction device, therefore saved, Signal Compression device is stored, the expense of transmission quantized value.

The embodiment of the present invention also provides a kind of computer-readable storage medium, and wherein, this computer-readable storage medium has program stored therein, and this program is carried out and comprised the part or all of layout of recording in said method embodiment.

Next introduce the another kind of Signal Compression device that the embodiment of the present invention provides, refer to shown in Fig. 9, Signal Compression device 900 comprises:

Input unit 901, output device 902, processor 903 and memory 904 (wherein the quantity of the processor 903 in positioner 900 can be one or more, and the processor of take in Fig. 9 is example).In some embodiments of the invention, input unit 901, output device 902, processor 903 and memory 904 can be connected by bus or other modes, wherein, in Fig. 9 to be connected to example by bus.

Wherein, input unit 901 is for being input to input signal processor 903;

Processor 903, be used for carrying out following steps: use sampling matrix to carry out low speed sampling to input signal, obtain sampled signal, wherein, described sampling matrix comprises M the row vector being comprised of N row sampling matrix value, and described input signal is the column vector being comprised of N input signal values, and described sampled signal is the column vector being comprised of M sampled value, described M, N are the natural number of non-zero, and M < N, use quantization function to carry out amplitude quantizing to the M of a described sampled signal sampled value, obtain quantized signal, described quantized signal is the column vector being comprised of M quantized value, judge respectively whether M quantized value surpasses preset quantization error tolerance, and described quantization error tolerance is for judging the quantification quality of described sampled signal being carried out to amplitude quantizing, from a described M quantized value, obtain m the quantized value that does not surpass quantization error tolerance, according to all described m quantized values, described sampled signal and described sampling matrix generate sampling and retain matrix, wherein, described m quantized value is in a described M quantized value, not surpass a quantized value of quantization error tolerance, and in a described M quantized value, at least there is a described m quantized value, described m quantized value is corresponding with m sampled value in described sampled signal, the row vector that described m the sampled value m by sampling matrix is comprised of N row sampling matrix value is multiplied by described input signal and obtains, described sampling retains m the row vector being comprised of N row sampling matrix value that matrix comprises all sampling matrixs.

Output device 902, for quantification stick signal and described sampling reservation matrix are sent to signal reconstruction device, described quantification stick signal comprises all described m quantized values.

In some embodiments of the invention, processor 903 specifically can be for carrying out following steps: according to described quantization error tolerance generating quantification error tolerance thresholding, judge respectively whether M quantification difference is greater than quantization error tolerance thresholding, for the first quantized value corresponding to quantification difference that is greater than quantization error tolerance thresholding, surpassed quantization error tolerance, second quantized value corresponding for the quantification difference that is less than or equal to quantization error tolerance thresholding do not surpass quantization error tolerance, wherein, described quantification difference is the absolute value of the difference between the first sampled value and described the first quantized value, or the absolute value of the second difference between sampled value and described the second quantized value, described the first quantized value is for to carry out to described the first sampled value the quantized value that amplitude quantizing obtains, described the first quantized value is corresponding with described the first sampled value, described the second quantized value is for to carry out to described the second sampled value the quantized value that amplitude quantizing obtains, described the second quantized value is corresponding with described the second sampled value.

In some embodiments of the invention, processor 903 specifically can be for carrying out following steps: according to described m quantized value, obtain in described sampled signal for obtaining m sampled value of described m quantized value; According to described m sampled value, obtain in described sampling matrix m the row vector being formed by N row sampling matrix value of using while calculating described m sampled value sampling by low speed; All described m the row vector being comprised of N row sampling matrix value combined, obtain described sampling and retain matrix, or, other row vectors that are comprised of N row sampling matrix value except all described m the row vector being comprised of N row sampling matrix value in described sampling matrix are rejected, again the remaining row vector being comprised of N row sampling matrix value in described sampling matrix is combined, obtained described sampling and retain matrix.

In some embodiments of the invention, processor 903 specifically can be for carrying out following steps: according to all described m quantized values, obtain n quantized value, described n quantized value is a quantized value except all described m quantized values in a described M quantized value, and described n quantized value is in a described M quantized value, to surpass a quantized value of quantization error tolerance; According in sampled signal described in described n quantized value for obtaining n sampled value of described n quantized value; According to described n sampled value, obtain in described sampling matrix n the row vector being formed by N row sampling matrix value of using while calculating described n sampled value sampling by low speed; Other row vectors that are comprised of N row sampling matrix value except all described n the row vector being comprised of N row sampling matrix value in described sampling matrix are combined, obtain described sampling and retain matrix, or, all described n in described sampling matrix the row vector being comprised of N row sampling matrix value rejected, again the remaining row vector being comprised of N row sampling matrix value in described sampling matrix is combined, obtained described sampling and retain matrix.

Next introduce the another kind of signal reconstruction device that the embodiment of the present invention provides, refer to shown in Figure 10, signal reconstruction device 1000 comprises:

Input unit 1001, output device 1002, processor 1003 and memory 1004 (wherein the quantity of the processor 1003 in positioner 1000 can be one or more, and the processor of take in Figure 10 is example).In some embodiments of the invention, input unit 1001, output device 1002, processor 1003 and memory 1004 can be connected by bus or other modes, wherein, in Figure 10 to be connected to example by bus.

Wherein, quantification stick signal and sampling reservation matrix that input unit 1001 sends for receiving Signal Compression device, to quantize stick signal and sampling and retain Input matrix in processor 1003, wherein, described quantification stick signal comprises m all quantized values, described m quantized value is in M quantized value, not surpass a quantized value of quantization error tolerance, a described M quantized value is that described Signal Compression device is used sampling matrix to carry out after low speed sampling acquisition sampled signal input signal, use quantization function sampled signal to be carried out to the result of amplitude quantizing, described M is the number of the row vector of described sampling matrix,

Processor 1003, for carrying out following steps: carry out signal reconstruction according to the quantification stick signal receiving and sampling reservation matrix, obtain signal reconstruction result;

Output device 1002, for exporting described signal reconstruction result.

In some embodiments of the invention, processor 1003 specifically can be for carrying out following steps: according to described quantification stick signal and sampling, retain matrix and set constraints; Column vector when calculating makes target function reach minimum value under described constraints, described column vector is signal reconstruction result.

Further, processor 1003 specifically can be for carrying out following steps: obtain in the following way constraints:

$s.t.:{\left|\right|\tilde{y}-\widetilde{\mathrm{\&Phi;}}\mathrm{\&theta;}\left|\right|}_2\&le;\mathrm{\&epsiv;},$

Wherein, described s.t. represents constraints, described in for described quantification stick signal, described in for described sampling retains matrix, described θ is column vector to be reconstructed, and described ε is preset error constraints parameter;

Obtain in the following way signal Output rusults:

$\widehat{\mathrm{\&theta;}}=\underset{\mathrm{\&theta;}}{\arg \min }\left\{{\left|\right|\mathrm{\&theta;}\left|\right|}_1\right\}$ ，

$s.t.:{\left|\right|\tilde{y}-\widetilde{\mathrm{\&Phi;}}\mathrm{\&theta;}\left|\right|}_2\&le;\mathrm{\&epsiv;}$

Wherein, described in $\begin{array}{c}\widehat{\mathrm{\&theta;}}=\underset{\mathrm{\&theta;}}{\arg \min }\{\&CenterDot;\}\\ s.t.,...\end{array}$ For solving, reach the column vector of minimum value meeting the described constraints target function of ordering

To sum up, in the embodiment of the present invention, first use sampling matrix to carry out low speed sampling to input signal, obtain sampled signal, then use quantization function to carry out amplitude quantizing to the M of a sampled signal sampled value, obtain quantized signal, next judge whether M quantized value in quantized signal has surpassed quantization error tolerance, from M quantized value, choose m the quantized value that does not surpass quantization error tolerance, according to m all quantized values, sampled signal and sampling matrix generate sampling and retain matrix, the quantification stick signal finally m the quantized value by all being formed and sampling retain matrix and send to signal reconstruction device.Due to what send to signal reconstruction device, be not surpass the quantized value of quantization error tolerance, and quantize second-rate quantized value, do not send to signal reconstruction device, therefore what send to signal reconstruction device is all to quantize the higher quantized value of quality, and according to the quantized value that does not surpass quantization error tolerance, former sampling matrix is screened, generate sampling and retain matrix, after receiving not over the quantized value of quantization error tolerance and the reservation matrix of sampling, signal reconstruction device carries out signal reconstruction, can reduce the impact of quantization error on signal reconstruction result, improve the accuracy of signal reconstruction.In addition, due to what send to signal reconstruction device, be not surpass the quantized value of quantization error tolerance, and quantize second-rate quantized value, do not send to signal reconstruction device, therefore saved, Signal Compression device is stored, the expense of transmission quantized value.

One of ordinary skill in the art will appreciate that all or part of step realizing in above-described embodiment method is to come the hardware that instruction is relevant to complete by program, described program can be stored in a kind of computer-readable recording medium, the above-mentioned storage medium of mentioning can be read-only memory, disk or CD etc.

Above a kind of compression method provided by the present invention and reconstructing method and relevant apparatus and system are described in detail, for one of ordinary skill in the art, thought according to the embodiment of the present invention, all will change in specific embodiments and applications, in sum, this description should not be construed as limitation of the present invention.

Claims (17)

1. a compression method, is characterized in that, comprising:

Use sampling matrix to carry out low speed sampling to input signal, obtain sampled signal, wherein, described sampling matrix comprises M the row vector being comprised of N row sampling matrix value, described input signal is the column vector being comprised of N input signal values, described sampled signal is the column vector being comprised of M sampled value, the natural number that described M, N are non-zero, and M < N;

Use quantization function to carry out amplitude quantizing to the M of a described sampled signal sampled value, obtain quantized signal, described quantized signal is the column vector being comprised of M quantized value;

Judge respectively whether M quantized value surpasses preset quantization error tolerance;

From a described M quantized value, obtain m the quantized value that does not surpass quantization error tolerance, according to all described m quantized values, described sampled signal and described sampling matrix generate sampling and retain matrix, wherein, described m quantized value is in a described M quantized value, not surpass a quantized value of quantization error tolerance, and in a described M quantized value, at least there is a described m quantized value, described m quantized value is corresponding with m sampled value in described sampled signal, the row vector that described m the sampled value m by described sampling matrix is comprised of N row sampling matrix value is multiplied by described input signal and obtains, described sampling retains m the row vector being comprised of N row sampling matrix value that matrix comprises all described sampling matrixs,

Quantification stick signal and described sampling reservation matrix are sent to signal reconstruction device, and described quantification stick signal comprises all described m quantized values.

2. method according to claim 1, is characterized in that, described judge respectively M quantized value whether preset surpass quantization error tolerance, comprising:

According to described quantization error tolerance generating quantification error tolerance thresholding;

Judge respectively whether M quantification difference is greater than quantization error tolerance thresholding, for the first quantized value corresponding to quantification difference that is greater than quantization error tolerance thresholding, surpassed quantization error tolerance, second quantized value corresponding for the quantification difference that is less than or equal to quantization error tolerance thresholding do not surpass quantization error tolerance, wherein, described quantification difference is the absolute value of the difference between the first sampled value and described the first quantized value, or the absolute value of the second difference between sampled value and described the second quantized value, described the first quantized value is for to carry out to described the first sampled value the quantized value that amplitude quantizing obtains, described the first quantized value is corresponding with described the first sampled value, described the second quantized value is for to carry out to described the second sampled value the quantized value that amplitude quantizing obtains, described the second quantized value is corresponding with described the second sampled value.

3. method according to claim 1 and 2, is characterized in that, describedly according to all described m quantized values, described sampled signal and described sampling matrix, generates sampling and retains matrix and comprise:

According to described m quantized value, obtain in described sampled signal for obtaining m sampled value of described m quantized value;

According to described m sampled value, obtain in described sampling matrix m the row vector being formed by N row sampling matrix value of using while calculating described m sampled value sampling by low speed;

All described m the row vector being comprised of N row sampling matrix value combined, obtain described sampling and retain matrix, or, other row vectors that are comprised of N row sampling matrix value except all described m the row vector being comprised of N row sampling matrix value in described sampling matrix are rejected, again the remaining row vector being comprised of N row sampling matrix value in described sampling matrix is combined, obtained described sampling and retain matrix.

4. method according to claim 1 and 2, is characterized in that, describedly according to all described m quantized values, described sampled signal and described sampling matrix, generates sampling and retains matrix and comprise:

According to all described m quantized values, obtain n quantized value, described n quantized value is a quantized value except all described m quantized values in a described M quantized value, and described n quantized value is in a described M quantized value, to surpass a quantized value of quantization error tolerance;

According to described n quantized value, obtain in described sampled signal for obtaining n sampled value of described n quantized value;

According to described n sampled value, obtain in described sampling matrix n the row vector being formed by N row sampling matrix value of using while calculating described n sampled value sampling by low speed;

Other row vectors that are comprised of N row sampling matrix value except all described n the row vector being comprised of N row sampling matrix value in described sampling matrix are combined, obtain described sampling and retain matrix, or, all described n in described sampling matrix the row vector being comprised of N row sampling matrix value rejected, again the remaining row vector being comprised of N row sampling matrix value in described sampling matrix is combined, obtained described sampling and retain matrix.

5. according to the method described in any one in claim 1 to 4, it is characterized in that, the t power that the quantized level using in described quantization function is 2, described t is quantizing bit number.

6. a signal reconfiguring method, is characterized in that, comprising:

Receive quantification stick signal and sampling reservation matrix that Signal Compression device sends, wherein, described quantification stick signal comprises m all quantized values, described m quantized value is in M quantized value, not surpass a quantized value of quantization error tolerance, a described M quantized value is that described Signal Compression device is used sampling matrix to carry out after low speed sampling acquisition sampled signal input signal, use quantization function sampled signal to be carried out to the result of amplitude quantizing, the number of the row vector that described M is described sampling matrix;

According to the quantification stick signal receiving and sampling reservation matrix, carry out signal reconstruction, obtain signal reconstruction result;

Export described signal reconstruction result.

7. method according to claim 6, is characterized in that, described carries out signal reconstruction according to the quantification stick signal receiving and sampling reservation matrix, obtains signal reconstruction result, comprising:

According to described quantification stick signal and described sampling, retain matrix setting constraints;

Column vector when calculating makes target function reach minimum value under described constraints, described column vector is signal reconstruction result.

8. method according to claim 7, is characterized in that, describedly according to described quantification stick signal and described sampling, retains matrix and sets constraints and specifically obtain in the following way:

$s.t.:{\left|\right|\tilde{y}-\widetilde{\mathrm{\&Phi;}}\mathrm{\&theta;}\left|\right|}_2\&le;\mathrm{\&epsiv;},$ Wherein, described s.t. represents constraints, described in for described quantification stick signal, described in for described sampling retains matrix, described θ is column vector to be reconstructed, and described ε is preset error constraints parameter;

Column vector when described calculating makes target function reach minimum value under described constraints is specifically obtained in the following way: $\widehat{\mathrm{\&theta;}}=\underset{\mathrm{\&theta;}}{\arg \min }\left\{{\left|\right|\mathrm{\&theta;}\left|\right|}_1\right\}$ ，

$s.t.:{\left|\right|\tilde{y}-\widetilde{\mathrm{\&Phi;}}\mathrm{\&theta;}\left|\right|}_2\&le;\mathrm{\&epsiv;}$

Wherein, described in $\begin{array}{c}\widehat{\mathrm{\&theta;}}=\underset{\mathrm{\&theta;}}{\arg \min }\{\&CenterDot;\}\\ s.t.,...\end{array}$ For solving, reach the column vector of minimum value meeting the described constraints target function of ordering

9. a Signal Compression device, is characterized in that, comprising:

Sampling module, be used for using sampling matrix to carry out low speed sampling to input signal, obtain sampled signal, wherein, described sampling matrix comprises M the row vector being comprised of N row sampling matrix value, and described input signal is the column vector being comprised of N input signal values, and described sampled signal is the column vector being comprised of M sampled value, described M, N are the natural number of non-zero, and M < N;

Quantization modules, for using quantization function to carry out amplitude quantizing to the M of a described sampled signal sampled value, obtains quantized signal, and described quantized signal is the column vector being comprised of M quantized value;

Judge module, for judging respectively whether M quantized value surpasses preset quantization error tolerance;

Acquisition module, for obtain m the quantized value that does not surpass quantization error tolerance from a described M quantized value, according to all described m quantized values, described sampled signal and described sampling matrix generate sampling and retain matrix, wherein, described m quantized value is in a described M quantized value, not surpass a quantized value of quantization error tolerance, and in a described M quantized value, at least there is a described m quantized value, described m quantized value is corresponding with m sampled value in described sampled signal, the row vector that described m the sampled value m by described sampling matrix is comprised of N row sampling matrix value is multiplied by described input signal and obtains, described sampling retains m the row vector being comprised of N row sampling matrix value that matrix comprises all described sampling matrixs,

Sending module, for quantification stick signal and described sampling reservation matrix are sent to signal reconstruction device, described quantification stick signal comprises all described m quantized values.

10. device according to claim 9, is characterized in that, described judge module, specifically comprises:

Generate submodule, for tolerating thresholding according to described quantization error tolerance generating quantification error;

Judgement submodule, for judging respectively whether M quantification difference is greater than quantization error tolerance thresholding, for the first quantized value corresponding to quantification difference that is greater than quantization error tolerance thresholding, surpassed quantization error tolerance, second quantized value corresponding for the quantification difference that is less than or equal to quantization error tolerance thresholding do not surpass quantization error tolerance, wherein, described quantification difference is the absolute value of the difference between the first sampled value and described the first quantized value, or the absolute value of the second difference between sampled value and described the second quantized value, described the first quantized value is for to carry out to described the first sampled value the quantized value that amplitude quantizing obtains, described the first quantized value is corresponding with described the first sampled value, described the second quantized value is for to carry out to described the second sampled value the quantized value that amplitude quantizing obtains, described the second quantized value is corresponding with described the second sampled value.

11. according to the device described in claim 9 or 10, it is characterized in that, described acquisition module, specifically comprises:

First obtains submodule, for obtaining described sampled signal for obtaining m sampled value of described m quantized value according to described m quantized value;

Second obtains submodule, for obtaining described sampling matrix according to described m sampled value in m the row vector being comprised of N row sampling matrix value of sampling by low speed and using while calculating described m sampled value;

The 3rd obtains submodule, for the row vector that all described m is comprised of N row sampling matrix value, combine, obtain described sampling and retain matrix, or, other row vectors that are comprised of N row sampling matrix value except all described m the row vector being comprised of N row sampling matrix value in described sampling matrix are rejected, again the remaining row vector being comprised of N row sampling matrix value in described sampling matrix is combined, obtained described sampling and retain matrix.

12. according to the device described in claim 9 or 10, it is characterized in that, described acquisition module, specifically comprises:

The 4th obtains submodule, for obtaining n quantized value according to all described m quantized values, described n quantized value is a quantized value except all described m quantized values in a described M quantized value, and described n quantized value is in a described M quantized value, to surpass a quantized value of quantization error tolerance;

The 5th obtains submodule, for obtaining described sampled signal for obtaining n sampled value of described n quantized value according to described n quantized value;

The 6th obtains submodule, for obtaining described sampling matrix according to described n sampled value in n the row vector being comprised of N row sampling matrix value of sampling by low speed and using while calculating described n sampled value;

The 7th obtains submodule, being used for other row vectors that are comprised of N row sampling matrix value except all described n the row vector being comprised of N row sampling matrix value by described sampling matrix combines, obtain described sampling and retain matrix, or, all described n in described sampling matrix the row vector being comprised of N row sampling matrix value rejected, again the remaining row vector being comprised of N row sampling matrix value in described sampling matrix is combined, obtained described sampling and retain matrix.

13. according to the device described in any one in claim 9 to 12, it is characterized in that, the t power that in the described quantization function that described quantization modules is used, quantized level is 2, and described t is quantizing bit number.

14. 1 kinds of signal reconstruction devices, is characterized in that, comprising:

Receiver module, the quantification stick signal and the sampling reservation matrix that for receiving Signal Compression device, send, wherein, described quantification stick signal comprises m all quantized values, described m quantized value is in M quantized value, not surpass a quantized value of quantization error tolerance, a described M quantized value is that described Signal Compression device is used sampling matrix to carry out after low speed sampling acquisition sampled signal input signal, use quantization function sampled signal to be carried out to the result of amplitude quantizing, the number of the row vector that described M is described sampling matrix;

Rebuild module, for carrying out signal reconstruction according to the quantification stick signal receiving and sampling reservation matrix, obtain signal reconstruction result;

Output module, for exporting described signal reconstruction result.

15. devices according to claim 14, is characterized in that, described reconstruction module, specifically comprises:

Obtain submodule, for retaining matrix setting constraints according to described quantification stick signal and described sampling;

Calculating sub module, the column vector while making target function reach minimum value for calculating under described constraints, described column vector is signal reconstruction result.

16. devices according to claim 15, is characterized in that, described in obtain submodule, specifically for obtaining in the following way constraints:

$s.t.:{\left|\right|\tilde{y}-\widetilde{\mathrm{\&Phi;}}\mathrm{\&theta;}\left|\right|}_2\&le;\mathrm{\&epsiv;},$

Wherein, described s.t. represents constraints, described in for described quantification stick signal, described in for described sampling retains matrix, described θ is column vector to be reconstructed, and described ε is preset error constraints parameter;

Described calculating sub module, specifically for obtaining in the following way column vector:

$\widehat{\mathrm{\&theta;}}=\underset{\mathrm{\&theta;}}{\arg \min }\left\{{\left|\right|\mathrm{\&theta;}\left|\right|}_1\right\}$ ，

$s.t.:{\left|\right|\tilde{y}-\widetilde{\mathrm{\&Phi;}}\mathrm{\&theta;}\left|\right|}_2\&le;\mathrm{\&epsiv;}$

Wherein, described in $\begin{array}{c}\widehat{\mathrm{\&theta;}}=\underset{\mathrm{\&theta;}}{\arg \min }\{\&CenterDot;\}\\ s.t.,...\end{array}$ For solving, reach the column vector of minimum value meeting the described constraints target function of ordering

17. 1 kinds of signal processing systems, it is characterized in that, comprise: the Signal Compression device as described in any one claim in claim 9 to 13 and the signal reconstruction device as described in any one claim in claim 14 to 16, wherein, described Signal Compression device with described signal reconstruction device being connected by communication mode.