CN107147396B - Signal sampling method, signal sampling system and signal sampling device - Google Patents

Signal sampling method, signal sampling system and signal sampling device Download PDF

Info

Publication number
CN107147396B
CN107147396B CN201610116285.4A CN201610116285A CN107147396B CN 107147396 B CN107147396 B CN 107147396B CN 201610116285 A CN201610116285 A CN 201610116285A CN 107147396 B CN107147396 B CN 107147396B
Authority
CN
China
Prior art keywords
sampling
matrix
signal
boolean
row
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Active
Application number
CN201610116285.4A
Other languages
Chinese (zh)
Other versions
CN107147396A (en
Inventor
余浩
王雨豪
王侃文
杨伟
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Huawei Technologies Co Ltd
Nanyang Technological University
Original Assignee
Huawei Technologies Co Ltd
Nanyang Technological University
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Huawei Technologies Co Ltd, Nanyang Technological University filed Critical Huawei Technologies Co Ltd
Priority to CN201610116285.4A priority Critical patent/CN107147396B/en
Publication of CN107147396A publication Critical patent/CN107147396A/en
Application granted granted Critical
Publication of CN107147396B publication Critical patent/CN107147396B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • H—ELECTRICITY
    • H03—ELECTRONIC CIRCUITRY
    • H03M—CODING; DECODING; CODE CONVERSION IN GENERAL
    • H03M7/00—Conversion of a code where information is represented by a given sequence or number of digits to a code where the same, similar or subset of information is represented by a different sequence or number of digits
    • H03M7/30—Compression; Expansion; Suppression of unnecessary data, e.g. redundancy reduction
    • H03M7/3059—Digital compression and data reduction techniques where the original information is represented by a subset or similar information, e.g. lossy compression
    • H03M7/3062—Compressive sampling or sensing

Abstract

The embodiment of the invention provides a signal sampling method, a signal sampling system and a signal sampling device. The signal sampling method comprises the following steps: obtaining a real number matrix psi of m rows and n columns related to the application data according to the application data corresponding to the target signal, wherein m and n are integers greater than or equal to 1; obtaining a Boolean sampling matrix satisfying condition (1) from Ψ
Figure DDA0000932783420000011
Figure DDA0000932783420000012
Wherein T is a matrix with m rows and m columns, and D is a diagonal matrix with m rows and m columns; according to
Figure DDA0000932783420000013
And sampling the target signal to obtain a sampling value. According to the technical scheme, when the signal is sampled, the complexity and the power consumption of hardware used for sampling the signal according to the data-related Boolean sampling matrix can be reduced.

Description

Signal sampling method, signal sampling system and signal sampling device
Technical Field
The present invention relates to the field of signal processing, and in particular, to a signal sampling method, a signal sampling system, and a signal sampling apparatus.
Background
The conventional signal processing process may include four parts of sampling, compressing, storing/transmitting and decompressing, in the sampling process, data acquisition is performed on an original signal according to Nyquist sampling theorem, then the acquired data is compressed, in the compressing process, the acquired data is firstly transformed, then a few coefficients with larger absolute values are compressed and encoded, and other coefficients with zero or close to zero are discarded, wherein most of the data obtained by sampling is discarded in the compressing process.
In recent years, a new signal acquisition technology, i.e., compressive sensing (compressive sensing), has been proposed, which utilizes the sparsity of the original signal s, i.e., only K non-zero terms whose positions are unknown after the original signal is orthogonally transformed. The data volume measured by the compressive sensing technology is far smaller than that required by traditional sampling.
In the process of compressed sensing implementation, first, sampling values of the original signal s are obtained (non-adaptive linear projection), where the sampling values are represented by equation (1.1):
y＝Φα (1.1)
wherein y is a sampling value, phi is a sampling matrix, and alpha is an original signal. The original signal α is represented as a column vector of N terms, and there is a sparse representation, that is, Ψ α ═ x obtained after orthogonal transformation Ψ has non-zero terms whose positions are unknown at K (K < < N). The measurement y is a column vector of M terms (M < < N & M >2K), and the sampling matrix Φ is a matrix of M rows and N columns.
After the sampling value y of the original signal is acquired and stored/transmitted, signal reconstruction may be performed.
Therefore, the processing of the signal by using the compressed sensing technology is divided into a signal sampling phase and a signal recovery phase. The signal sampling phase includes constructing a sampling matrix and sampling the signal data using the sampling matrix. The existing sampling matrix construction methods can be divided into two types, the first type is to construct a sampling matrix irrelevant to application data, and the second type is to construct a sampling matrix relevant to application data. The sampling matrix independent of the application data may be a real matrix following a gaussian distribution or a random boolean matrix following a bernoulli distribution. From the hardware point of view, the random boolean matrix is generally preferred, but the data is sampled by using the random boolean matrix, and the inherent characteristics of the data cannot be well utilized, so that the accuracy of the sampling value is poor. And the sampling matrix related to the application data is adopted to sample the signals, so that the internal mode of the application data can be learned, the data information can be better utilized, more acquisition is carried out in an information dense area, less acquisition is carried out in an information sparse area, and higher signal quality or less sampling number is achieved.
However, a sampling matrix related to data is constructed based on application data corresponding to the learning original signal, a real number sampling matrix related to the application data is generated, and hardware used for sampling signals according to the real number sampling matrix is difficult to implement to a certain extent and has high power consumption.
Disclosure of Invention
The invention provides a signal sampling method, a signal sampling system and a signal sampling device, which can reduce the complexity and power consumption of hardware used for sampling signals according to a data related sampling matrix.
In a first aspect, the present invention provides a signal sampling method. Firstly, acquiring a real number matrix psi of m rows and n columns related to application data according to the application data corresponding to a target signal, wherein m and n are integers which are more than or equal to 1. Then obtaining a Boolean sampling matrix satisfying the condition (1) according to psi
Figure BDA0000932783400000021
Figure BDA0000932783400000022
Wherein T is a matrix with m rows and m columns, and D is a diagonal matrix with m rows and m columns. Finally, according to
Figure BDA0000932783400000023
And sampling the target signal to obtain a sampling value.
The signal sampling method of the invention converts a real number sampling matrix related to application data into a Boolean sampling matrix related to the application data, and then samples a target signal according to the Boolean sampling matrix. The target signal is sampled according to the Boolean sampling matrix related to the application data, so that the information of the data can be better utilized, and the complexity and the power consumption of hardware used in sampling can be reduced.
In one possible implementation, TT×T＝D2The method comprises the following steps: t isTAnd x T is I, I is an identity matrix.
In one possible implementation, a Boolean sampling matrix satisfying condition (1) is obtained from Ψ
Figure BDA0000932783400000024
First, the first row T of the T is initialized randomly1A value of (d); randomly initializing the
Figure BDA0000932783400000025
First row of
Figure BDA0000932783400000026
Then according to psi and T of T1To line i-1 ti-1Obtaining the
Figure BDA0000932783400000027
Row i of (2)
Figure BDA0000932783400000028
Figure BDA0000932783400000031
Wherein i is an integer greater than 1 and less than or equal to m.
The implementation method initializes the first line of T and then obtains the first line of T in sequence through formula (2)
Figure BDA0000932783400000032
And in other rows, the nonlinear constraint condition is converted into a linear constraint condition, so that the complexity of the solution can be reduced.
In a second aspect, a signal sampling system is provided. The signal acquisitionThe sampling system comprises a real number sampling matrix acquisition module, a Boolean sampling matrix acquisition module and a sampling module. The real number sampling matrix acquisition module is used for acquiring a sampling matrix psi related to application data according to the application data corresponding to the target signal, psi is a real number matrix with m rows and n columns, and m and n are integers greater than or equal to 1. The Boolean sampling matrix acquisition module is used for acquiring the Boolean sampling matrix meeting the condition (1) according to psi
Figure BDA0000932783400000033
Figure BDA0000932783400000034
Wherein T is a matrix with m rows and m columns, and D is a diagonal matrix with m rows and m columns. The sampling module is used for
Figure BDA0000932783400000035
And sampling the target signal to obtain a sampling value.
The signal sampling system converts a real number sampling matrix related to application data into a Boolean sampling matrix related to the application data, and then samples a target signal according to the Boolean sampling matrix. The target signal is sampled according to the Boolean sampling matrix related to the application data, so that the information of the data can be better utilized, and the complexity and the power consumption of hardware used in sampling can be reduced.
In one possible implementation, TT×T＝D2The method comprises the following steps: t isTAnd x T is I, I is an identity matrix.
In a possible implementation, the boolean sampling matrix acquisition module is specifically configured to randomly initialize the first row T of T1A value of (d); randomly initializing the
Figure BDA0000932783400000036
First row of
Figure BDA0000932783400000037
And according to T of psi and T in turn1To line i-1 ti-1Obtaining
Figure BDA0000932783400000038
Row i of (2)
Figure BDA0000932783400000039
Figure BDA00009327834000000310
Wherein i is an integer greater than 1 and less than or equal to m.
The signal sampling system in the above implementation initializes the first row of T and
Figure BDA00009327834000000311
then the other rows of T are sequentially obtained by equation (2). When each line of T is acquired, the T can be acquired simultaneously
Figure BDA0000932783400000041
The corresponding line converts the nonlinear constraint condition into a linear constraint condition, so that the complexity of solution can be reduced.
In a third aspect, a signal sampling apparatus is provided. The signal sampling device comprises a real number sampling matrix acquisition module, a Boolean sampling matrix acquisition module and a sampling module. The real number sampling matrix acquisition module is used for acquiring a sampling matrix psi related to application data according to the application data corresponding to the target signal, psi is a real number matrix with m rows and n columns, and m and n are integers greater than or equal to 1. The Boolean sampling matrix acquisition module is used for acquiring the Boolean sampling matrix meeting the condition (1) according to psi
Figure BDA0000932783400000042
Figure BDA0000932783400000043
Wherein T is a matrix with m rows and m columns, and D is a diagonal matrix with m rows and m columns. The sampling module is used for
Figure BDA0000932783400000044
And sampling the target signal to obtain a sampling value.
The signal sampling device converts a real number sampling matrix related to application data into a Boolean sampling matrix related to the application data, and then samples a target signal according to the Boolean sampling matrix. The target signal is sampled according to the Boolean sampling matrix related to the application data, so that the information of the data can be better utilized, and the complexity and the power consumption of hardware used in sampling can be reduced.
In one possible implementation, TT×T＝D2The method comprises the following steps: t isTAnd x T is I, I is an identity matrix.
In a possible implementation, the boolean sampling matrix acquisition module is specifically configured to randomly initialize the first row T of T1A value of (d); randomly initializing the
Figure BDA0000932783400000045
First row of
Figure BDA0000932783400000046
And according to T of psi and T in turn1To line i-1 ti-1Obtaining
Figure BDA0000932783400000047
Row i of (2)
Figure BDA0000932783400000048
Figure BDA0000932783400000049
Wherein i is an integer greater than 1 and less than or equal to m.
The signal sampling apparatus in the above implementation initializes the first row of T and
Figure BDA00009327834000000410
the first row ofThen, the other rows of T are sequentially obtained by formula (2). When each line of T is acquired, the T can be acquired simultaneously
Figure BDA00009327834000000411
The corresponding line converts the nonlinear constraint condition into a linear constraint condition, so that the complexity of solution can be reduced.
In a fourth aspect, a signal sampling apparatus is provided that includes a memory and a processor. The memory is used for storing programs and the processor is used for executing the programs stored in the memory. The processor is configured to perform the method of the first aspect when the memory stores a program for execution.
Drawings
In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings needed to be used in the embodiments of the present invention will be briefly described below, and it is obvious that the drawings described below are only some embodiments of the present invention, and it is obvious for those skilled in the art to obtain other drawings based on these drawings without creative efforts.
FIG. 1 is a schematic flow chart diagram of a signal sampling method of an embodiment of the present invention;
FIG. 2 is a schematic flow chart diagram of a method of obtaining a Boolean sampling matrix of an embodiment of the present invention;
fig. 3 is a schematic configuration diagram of a signal sampling system of an embodiment of the present invention.
Fig. 4 is a schematic configuration diagram of a signal sampling apparatus according to an embodiment of the present invention.
Fig. 5 is a schematic configuration diagram of a signal sampling apparatus according to an embodiment of the present invention.
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are some, not all, embodiments of the present invention. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
At present, when a signal is sampled by a compressive sensing technology, application data corresponding to the signal is trained first, an internal mode of the application data is learned, so that a real number sampling matrix related to the application data is obtained, and then a target signal corresponding to the application data is sampled by using the real number sampling matrix. Because the real number sampling matrix is used for sampling the target signal, the hardware has higher complexity and power consumption, and the like, the invention provides that Boolean optimization is carried out on the real number sampling matrix to obtain the Boolean matrix, the obtained Boolean matrix is ensured to keep the data-related information of the real number matrix (namely the Boolean matrix is also a data-related matrix for application data), and then the Boolean matrix is used for sampling the target signal. Therefore, a key point of the technical solution of the present invention is how to obtain a data-dependent boolean sampling matrix from a data-dependent real sampling matrix.
In the compressive sensing technology, it is required that, for all possible signals x, after sampling x by using a real number sampling matrix Ψ associated with data to obtain sampling values Ψ x, the sampling values Ψ x are satisfied with the following finite equivalent Property (RIP), so as to ensure the recoverability of the signal:
Figure BDA0000932783400000061
from the geometric perspective, namely after the signal x in the original n-dimensional space is subjected to compressed sensing sampling to the data of the m-dimensional subspace, the length of the signal corresponding to the data can only occur at mostkAnd (6) deformation.kIs a constant and ranges between 0 and 1.kThe smaller the value of (A), the smaller the deformation of the signal after mapping, and the higher the quality after recovery.kThe quality of the sampling matrix Ψ can be characterized.
The most important information in the real sampling matrix associated with the data is that RIP is guaranteed for signal x. In order to keep the result of data correlation after the real matrix is transformed into the boolean matrix, the boolean matrix obtained by the transformation still satisfies the RIP property. Thus, assuming that the transform operator is a matrix T, the following inequality needs to be satisfied to ensure the recoverability of the signal sampled by the boolean matrix:
Figure BDA0000932783400000062
the above inequality is satisfied if and only if T is an orthonormal (orthonormal) matrix. This is because the orthonormal matrix is used as an operator and does not change the sampling value
Figure BDA0000932783400000063
Length of (d).
In addition to satisfying orthonormal constraints, the T transform matrix needs to transform Ψ into a boolean matrix, and therefore T needs to satisfy:
Figure BDA0000932783400000064
wherein the content of the first and second substances,
Figure BDA0000932783400000065
is a boolean matrix. Since T is desired to
Figure BDA0000932783400000066
As close as possible, therefore, the following optimization problem can be planned:
Figure BDA0000932783400000067
in this problem, T and
Figure BDA0000932783400000068
are all unknown numbers, and are all provided with the same number,
Figure BDA0000932783400000069
is a Boolean matrix with m rows and n columns,
Figure BDA00009327834000000610
i.e. the cloth finally to be acquired for sampling the signalA matrix of samples is sampled. Normally, the values included in the boolean matrix are 0 and 1, but since-1 is generally used instead of 0 when multiplication of the matrix is implemented in a hardware device, the boolean matrix in the embodiment of the present invention
Figure BDA00009327834000000611
Is a matrix of m rows and n columns, and the elements in the matrix are-1 or 1.
In solving the problem by the above optimization
Figure BDA00009327834000000612
Then, the relaxed orthonormal constraint (orthonormal) that T needs to satisfy can be converted to an orthogonal constraint (orthogonal), so the above optimization problem can be converted to:
Figure BDA0000932783400000071
where D is a diagonal matrix. At the same time, can resume passing according to D
Figure BDA0000932783400000072
Sampling value obtained by sampling signal x
Figure BDA0000932783400000073
So that relaxation of the constraints does not change the quality of the recovered signal. Such as when using the following algorithm to derive the sample values from the samples
Figure BDA0000932783400000074
When the signal is recovered, the quality of the recovered signal is not changed:
Figure BDA0000932783400000075
where y is the recovered signal and is a value infinitely close to 0.
In obtaining
Figure BDA0000932783400000076
When, firstly, obtainT meeting orthogonal constraint is taken, and T can be obtained at the same time
Figure BDA0000932783400000077
Obtaining T satisfying the orthogonal constraint may use a line generation algorithm, i.e., generate line by line in an iterative manner. Specifically, the ith row of T is acquired from the first i-1 row of T under the condition that the ith row of T is orthogonal to the first i-1 row of T, and of course, the ith row of T is acquired at the same time as the ith row of T is acquired
Figure BDA0000932783400000078
Row i of (1)
Figure BDA0000932783400000079
In a specific implementation manner, the first row T of T may be initialized randomly first1And
Figure BDA00009327834000000710
first row of
Figure BDA00009327834000000711
Then T and are obtained line by line according to the following formula
Figure BDA00009327834000000712
The other rows of (2):
Figure BDA00009327834000000713
wherein, tiThe ith row of the representation T is,
Figure BDA00009327834000000714
to represent
Figure BDA00009327834000000715
Row i of (2).
The above-mentioned obtaining T and
Figure BDA00009327834000000716
in the method of (a) to (b),and the nonlinear constraint condition is converted into a linear constraint condition, so that the complexity of solution can be reduced.
In obtaining T and
Figure BDA00009327834000000717
the above constraints can also be processed using branch-and-bound methods for each row, so as to obtain the respective unknowns in each row.
Separately acquire
Figure BDA00009327834000000718
After each row of (A), a
Figure BDA00009327834000000719
Finally, the Boolean matrix is reused
Figure BDA00009327834000000720
Sampling, i.e. computing, the signal x
Figure BDA0000932783400000081
Thereby obtaining a sample value.
In the above description, how to obtain a data-dependent boolean matrix from a data-dependent real sampling matrix is described, which is only a subprocess in signal sampling, and a signal sampling method according to an embodiment of the present invention is described below with reference to fig. 1.
S101, obtaining a real number matrix psi of m rows and n columns related to application data according to the application data corresponding to the target signal, wherein m and n are integers greater than or equal to 1.
S102, acquiring a Boolean sampling matrix satisfying the condition (1) according to psi
Figure BDA0000932783400000082
Figure BDA0000932783400000083
Wherein, T is a matrix with m rows and m columns, D is a diagonal matrix with m rows and m columns, namely T is an orthogonal matrix.
S103, according to
Figure BDA0000932783400000084
And sampling the target signal to obtain a sampling value.
In the embodiment of the invention, because the target signal is sampled by using the data-related binary Boolean matrix, the matrix multiplication can be realized by using hardware with simple structure and low power consumption, namely the sampling value related to the target signal can be obtained by using the hardware with low power consumption.
In the embodiment of the present invention, constraint T in condition (1)T×T＝D2Can be further constrained to TTI is an identity matrix, I is an orthonormal matrix.
Obtaining conditions (1) according to Ψ
Figure BDA0000932783400000085
First row T of T may be initialized randomly1And
Figure BDA0000932783400000086
first row of
Figure BDA0000932783400000087
Then, the ith row of T is obtained according to the first i-1 row of T, and the first i-1 row of T and the ith row of T meet the condition (2):
Figure BDA0000932783400000088
wherein, tiThe ith row of the representation T is,
Figure BDA0000932783400000089
to represent
Figure BDA00009327834000000810
I ranges from 2 to m, m being the number of rows of T. The method can be obtained when the ith row of the T is obtained according to the first i-1 row of the T
Figure BDA00009327834000000811
The number of the ith row of (a),
the above obtaining satisfying the condition (1) according to Ψ
Figure BDA0000932783400000091
Is shown in fig. 2.
S201, randomly initializing the first row of T and
Figure BDA0000932783400000092
the first row of (1), namely to T and
Figure BDA0000932783400000093
is randomly assigned.
S202, starting from the second line of T, acquiring the ith line of T meeting the formula (2) according to the first i-1 line of T. When the ith row of T is acquired, the method can acquire
Figure BDA0000932783400000094
Row i of (2). Specifically, the second line of T satisfying formula (2) is obtained according to the first line of T, and the second line of T is obtained
Figure BDA0000932783400000095
Then according to the first and second lines of T, obtaining the third line of T satisfying the formula (2), and obtaining the second line of T
Figure BDA0000932783400000096
The third row of (2) is analogized in turn until the mth row of T meeting the formula (2) is obtained from the 1 st row to the m-1 st row of T, namely the
Figure BDA0000932783400000097
Row m.
S203, every time T sum is obtained
Figure BDA0000932783400000098
When the value of a row is not equal to the sum of T
Figure BDA0000932783400000099
In the corresponding row.
And S204, judging whether the row is the last row of the T or not when the value of one row of the T is acquired. If the last line of T is acquired, the T is completely acquired and simultaneously represents that
Figure BDA00009327834000000910
Has been completely acquired; if the row is not the last row of T, it indicates that the next row needs to be acquired.
S205, obtaining
Figure BDA00009327834000000911
Thereafter, the next step can be performed, i.e. according to
Figure BDA00009327834000000912
The signal is sampled.
Obtaining T to obtain
Figure BDA00009327834000000913
The method converts the nonlinear constraint into the linear constraint, and can reduce the computational complexity.
A signal sampling system 300 according to an embodiment of the present invention is described below with reference to fig. 3. Signal sampling system 300 may be the subject of execution of the signal sampling method shown in fig. 1.
The signal sampling system 300 includes a real sampling matrix acquisition module 301, a boolean sampling matrix acquisition module 302, and a sampling module 303.
The real sampling matrix obtaining module 301 is configured to obtain a sampling matrix Ψ related to application data according to the application data corresponding to the target signal, Ψ is a real matrix of m rows and n columns, and m and n are integers greater than or equal to 1.
The boolean sampling matrix acquisition module 302 is configured to acquire a boolean sampling matrix satisfying the following from Ψ
Figure BDA00009327834000000914
Figure BDA0000932783400000101
Wherein T is a matrix with m rows and m columns, and D is a diagonal matrix with m rows and m columns.
The sampling module 303 is used for
Figure BDA0000932783400000102
And sampling the target signal to obtain a sampling value.
The signal sampling system converts a real number sampling matrix related to application data into a Boolean sampling matrix related to the application data, and then samples a target signal according to the Boolean sampling matrix. The target signal is sampled according to the Boolean sampling matrix related to the application data, so that the information of the data can be better utilized, and the complexity and the power consumption of hardware used in sampling can be reduced.
Because signal sampling system 300 may be the subject of an implementation of the signal sampling method shown in fig. 1, the above-described and other operations or/and functions of signal sampling system 300 are the same as or similar to, respectively, the operations or/functions in the signal sampling method shown in fig. 1. For brevity, no further description is provided herein.
The sampling matrix obtaining module 301, the boolean sampling matrix obtaining module 302, and the sampling module 303 in the embodiment of the present invention may be deployed on the same hardware device, may also be deployed on different hardware devices, or may be deployed on a part of modules on the same hardware device, which is not limited in the present invention.
When the sampling matrix acquisition module and the boolean sampling matrix acquisition module are deployed on the same device (such as a server) and the sampling module is deployed on other devices (such as wearable devices), the sampling matrix acquisition module acquires a real number sampling matrix and the boolean sampling matrix acquisition module acquires and acquires the boolean sampling matrix according to the real number sampling matrix, and then the boolean sampling matrix is sent to the device where the sampling module is located through network transmission or other modes, and then the device samples a target signal according to the boolean sampling matrix.
When the sampling matrix acquisition module and the sampling module are arranged on the same device (such as a wearable device), and the Boolean sampling matrix acquisition module can be arranged on other devices (such as a server), the real number sampling matrix can be sent to the server after the sampling matrix acquisition module on the wearable device acquires the real number sampling matrix, the Boolean sampling matrix acquisition module on the server acquires the Boolean sampling matrix according to the real number sampling matrix, the server sends the Boolean sampling matrix to the wearable device, and finally the wearable device samples a target signal according to the Boolean sampling matrix.
When the sampling matrix acquisition module, the boolean sampling matrix acquisition module, and the sampling module are deployed on the same device or apparatus (e.g., on the same wearable device), a schematic structural diagram of the signal sampling apparatus is shown in fig. 4.
The functions of the sampling matrix obtaining module 401, the boolean sampling matrix obtaining module 402, and the sampling module 403 included in the signal sampling apparatus 400 in fig. 4 are respectively the same as or similar to the functions of the sampling matrix obtaining module 301, the boolean sampling matrix obtaining module 302, and the sampling module 303 in the signal sampling system 300 in fig. 3, and therefore, for brevity, no further description is provided here.
Fig. 5 is a schematic structural diagram of a signal sampling apparatus 500 according to an embodiment of the present invention. It should be understood that the signal sampling apparatus 500 of fig. 5 is capable of implementing the various steps of the signal sampling method of fig. 1.
The signal sampling apparatus 500 includes a memory 501 and a processor 502. The memory 501 is used to store programs; the processor 502 is configured to execute programs in the memory 501, and when executed, the processor 502 is configured to: obtaining a sampling matrix psi related to application data according to the application data corresponding to the target signal, psi is a real matrix with m rows and n columns, and m and n are integers greater than or equal to 1; obtaining a Boolean sampling matrix from Ψ that satisfies the following
Figure BDA0000932783400000111
Figure BDA0000932783400000112
Wherein T is the moment of m rows and m columnsD is a diagonal matrix with m rows and m columns; according to
Figure BDA0000932783400000113
And sampling the target signal to obtain a sampling value.
Optionally, as an embodiment, the processor 502 may be specifically configured to: first row T of random initialization T1(ii) a Random initialization
Figure BDA0000932783400000114
First row of
Figure BDA0000932783400000115
T according to Ψ and T in sequence1To line i-1 ti-1Obtaining
Figure BDA0000932783400000116
Is satisfied with the following conditions
Figure BDA0000932783400000117
Figure BDA0000932783400000118
Wherein i is an integer greater than 1 and less than or equal to m.
Optionally, as an embodiment, the processor 502 may be further configured to obtain a boolean sampling matrix satisfying the following from Ψ
Figure BDA0000932783400000121
Figure BDA0000932783400000122
When the data is sampled using a boolean matrix related to the data, it may be implemented using a Complementary Metal Oxide Semiconductor (CMOS) or a Resistive Random Access Memory (RRAM).
Generally, the sampling energy consumption and area of the CMOS is smaller than that of the RRAM, and the sampling time of the RRAM is smaller than that of the CMOS.
Those of ordinary skill in the art will appreciate that the various illustrative elements and algorithm steps described in connection with the embodiments disclosed herein may be implemented as electronic hardware or combinations of computer software and electronic hardware. Whether such functionality is implemented as hardware or software depends upon the particular application and design constraints imposed on the implementation. Skilled artisans may implement the described functionality in varying ways for each particular application, but such implementation decisions should not be interpreted as causing a departure from the scope of the present invention.
It is clear to those skilled in the art that, for convenience and brevity of description, the specific working processes of the above-described systems, apparatuses and units may refer to the corresponding processes in the foregoing method embodiments, and are not described herein again.
In the several embodiments provided in the present application, it should be understood that the disclosed system, apparatus and method may be implemented in other ways. For example, the above-described apparatus embodiments are merely illustrative, and for example, the division of the units is only one logical division, and other divisions may be realized in practice, for example, a plurality of units or components may be combined or integrated into another system, or some features may be omitted, or not executed. In addition, the shown or discussed mutual coupling or direct coupling or communication connection may be an indirect coupling or communication connection through some interfaces, devices or units, and may be in an electrical, mechanical or other form.
The units described as separate parts may or may not be physically separate, and parts displayed as units may or may not be physical units, may be located in one place, or may be distributed on a plurality of network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of the embodiment.
In addition, functional units in the embodiments of the present invention may be integrated into one processing unit, or each unit may exist alone physically, or two or more units are integrated into one unit.
The functions, if implemented in the form of software functional units and sold or used as a stand-alone product, may be stored in a computer readable storage medium. Based on such understanding, the technical solution of the present invention may be embodied in the form of a software product, which is stored in a storage medium and includes instructions for causing a computer device (which may be a personal computer, a server, or a network device) to execute all or part of the steps of the method according to the embodiments of the present invention. And the aforementioned storage medium includes: a U-disk, a removable hard disk, a Read-Only Memory (ROM), a Random Access Memory (RAM), a magnetic disk or an optical disk, and other various media capable of storing program codes.
The above description is only for the specific embodiments of the present invention, but the scope of the present invention is not limited thereto, and any person skilled in the art can easily conceive of the changes or substitutions within the technical scope of the present invention, and all the changes or substitutions should be covered within the scope of the present invention. Therefore, the protection scope of the present invention shall be subject to the protection scope of the claims.

Claims (4)

1. A method of sampling a signal, comprising:
acquiring a sampling matrix psi related to application data according to the application data corresponding to a target signal, wherein psi is a real matrix of m rows and n columns, and m and n are integers greater than or equal to 1;
obtaining a Boolean sampling matrix satisfying condition (1) from the Ψ
Figure FDA0002557570140000011
Figure FDA0002557570140000012
Wherein T is a matrix with m rows and m columns, and D is a diagonal matrix with m rows and m columns;
the obtaining a Boolean sampling matrix satisfying condition (1) according to Ψ
Figure FDA0002557570140000013
The method comprises the following steps:
randomly initializing the first row T of said T1；
Randomly initializing the
Figure FDA0002557570140000014
First row of
Figure FDA0002557570140000015
According to the psi and T of the T in sequence1To line i-1 ti-1Obtaining the
Figure FDA0002557570140000016
Row i of (2)
Figure FDA0002557570140000017
Figure FDA0002557570140000018
Wherein i is an integer greater than 1 and less than or equal to m;
according to the above
Figure FDA0002557570140000019
And sampling the target signal to obtain a sampling value.
2. Method for sampling a signal according to claim 1, characterized in that said T isT×T＝D2The method comprises the following steps: t isTAnd x T ═ I, where I is the identity matrix.
3. A signal sampling system, comprising:
a real number sampling matrix obtaining module, configured to obtain a sampling matrix Ψ related to application data according to the application data corresponding to a target signal, where Ψ is a real number matrix of m rows and n columns, and m and n are integers greater than or equal to 1;
a Boolean sampling matrix acquisition module for acquiring a Boolean sampling matrix satisfying the condition (1) according to the Ψ
Figure FDA00025575701400000110
Figure FDA00025575701400000111
Wherein T is a matrix with m rows and m columns, and D is a diagonal matrix with m rows and m columns;
the boolean sampling matrix acquisition module is specifically configured to:
randomly initializing the first row T of said T1；
Randomly initializing the
Figure FDA0002557570140000021
First row of
Figure FDA0002557570140000022
According to the psi and T of the T in sequence1To line i-1 ti-1Obtaining the
Figure FDA0002557570140000023
Row i of (2)
Figure FDA0002557570140000024
Figure FDA0002557570140000025
Wherein i is an integer greater than 1 and less than or equal to m;
a sampling module for sampling according to
Figure FDA0002557570140000026
And sampling the target signal to obtain a sampling value.
4. The signal sampling system of claim 3, wherein the T isT×T＝D2The method comprises the following steps: t isTAnd x T ═ I, where I is the identity matrix.
CN201610116285.4A 2016-03-01 2016-03-01 Signal sampling method, signal sampling system and signal sampling device Active CN107147396B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201610116285.4A CN107147396B (en) 2016-03-01 2016-03-01 Signal sampling method, signal sampling system and signal sampling device

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201610116285.4A CN107147396B (en) 2016-03-01 2016-03-01 Signal sampling method, signal sampling system and signal sampling device

Publications (2)

Publication Number Publication Date
CN107147396A CN107147396A (en) 2017-09-08
CN107147396B true CN107147396B (en) 2020-10-23

Family

ID=59783164

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201610116285.4A Active CN107147396B (en) 2016-03-01 2016-03-01 Signal sampling method, signal sampling system and signal sampling device

Country Status (1)

Country Link
CN (1) CN107147396B (en)

Families Citing this family (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111447229B (en) * 2020-03-27 2022-03-11 青岛海信网络科技股份有限公司 Large-scale data acquisition method and device based on compressed sensing theory

Citations (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US8170311B2 (en) * 2008-05-02 2012-05-01 The Uwm Research Foundation, Inc. Toeplitz random encoding for reduced acquisition in compressed sensing magnetic resonance imaging
CN104168228A (en) * 2014-08-06 2014-11-26 哈尔滨工业大学深圳研究生院 Compressed sensing ultra-wide band channel estimation method and system based on cluster position set

Patent Citations (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US8170311B2 (en) * 2008-05-02 2012-05-01 The Uwm Research Foundation, Inc. Toeplitz random encoding for reduced acquisition in compressed sensing magnetic resonance imaging
CN104168228A (en) * 2014-08-06 2014-11-26 哈尔滨工业大学深圳研究生院 Compressed sensing ultra-wide band channel estimation method and system based on cluster position set

Non-Patent Citations (2)

* Cited by examiner, † Cited by third party
Title
Optimizing Boolean embedding matrix for compressive sensing in RRAM crossbar;Yuhao Wang et al;《 2015 IEEE/ACM International Symposium on Low Power Electronics and Design (ISLPED)》;20150924;第12-17页 *
基于行列式随机循环的压缩感知测量矩阵研究;魏从静 等;《电视技术》;20151231;第39卷(第19期);第6-9页 *

Also Published As

Publication number Publication date
CN107147396A (en) 2017-09-08

Similar Documents

Publication Publication Date Title
Shekhar et al. Analysis sparse coding models for image-based classification
JP6352968B2 (en) Image sensor
Brajović et al. Gradient‐based signal reconstruction algorithm in Hermite transform domain
Gunasheela et al. Compressed sensing for image compression: survey of algorithms
Mahalanobis et al. Object specific image reconstruction using a compressive sensing architecture for application in surveillance systems
Hegde et al. A fast approximation algorithm for tree-sparse recovery
Zhang et al. Guided image filtering using signal subspace projection
JP2019533867A (en) Structured orthogonal random features for kernel-based machine learning
Yazdi et al. Time warp invariant kSVD: Sparse coding and dictionary learning for time series under time warp
Gai et al. Sparse representation based on vector extension of reduced quaternion matrix for multiscale image denoising
Fang et al. Graph regularised sparse NMF factorisation for imagery de‐noising
Carassou et al. Inferring the photometric and size evolution of galaxies from image simulations-I. Method
Yamaç et al. Convolutional sparse support estimator network (csen): From energy-efficient support estimation to learning-aided compressive sensing
Aires et al. Dimension reduction of satellite observations for remote sensing. Part 1: A comparison of compression, channel selection and bottleneck channel approaches
Wakin et al. Compressive sensing fundamentals
CN107147396B (en) Signal sampling method, signal sampling system and signal sampling device
Ramalho et al. A review of deterministic sensing matrices
Lyu et al. Identifiability-guaranteed simplex-structured post-nonlinear mixture learning via autoencoder
Dai et al. Dictionary learning and update based on simultaneous codeword optimization (SimCO)
Dias et al. Comparative analysis of sensing matrices for compressed sensed thermal images
Srinivas et al. Performance comparison of measurement matrices in compressive sensing
López-Radcenco et al. SRoll3: A neural network approach to reduce large-scale systematic effects in the Planck High-Frequency Instrument maps
Rateb et al. A generic top-level mechanism for accelerating signal recovery in compressed sensing
Zhang et al. Randomized methods in lossless compression of hyperspectral data
Li et al. Transforming the SMV model into MMV model based on the characteristics of wavelet coefficients

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant