CN110826703A - Communication system signal sequence detection method based on cooperative time-varying bidirectional cyclic neural network - Google Patents

Communication system signal sequence detection method based on cooperative time-varying bidirectional cyclic neural network Download PDF

Info

Publication number
CN110826703A
CN110826703A CN201911265807.7A CN201911265807A CN110826703A CN 110826703 A CN110826703 A CN 110826703A CN 201911265807 A CN201911265807 A CN 201911265807A CN 110826703 A CN110826703 A CN 110826703A
Authority
CN
China
Prior art keywords
neural network
time
layer
sequence
network
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.)
Granted
Application number
CN201911265807.7A
Other languages
Chinese (zh)
Other versions
CN110826703B (en
Inventor
孙黎
王宇威
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.)
Xian Jiaotong University
Original Assignee
Xian Jiaotong 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 Xian Jiaotong University filed Critical Xian Jiaotong University
Publication of CN110826703A publication Critical patent/CN110826703A/en
Application granted granted Critical
Publication of CN110826703B publication Critical patent/CN110826703B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/049—Temporal neural networks, e.g. delay elements, oscillating neurons or pulsed inputs
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L5/00—Arrangements affording multiple use of the transmission path
    • H04L5/003—Arrangements for allocating sub-channels of the transmission path

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Health & Medical Sciences (AREA)
  • Computing Systems (AREA)
  • Biomedical Technology (AREA)
  • Biophysics (AREA)
  • Computational Linguistics (AREA)
  • Data Mining & Analysis (AREA)
  • Evolutionary Computation (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Molecular Biology (AREA)
  • Artificial Intelligence (AREA)
  • General Engineering & Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Mathematical Physics (AREA)
  • Software Systems (AREA)
  • Health & Medical Sciences (AREA)
  • Signal Processing (AREA)
  • Computer Networks & Wireless Communication (AREA)
  • Data Exchanges In Wide-Area Networks (AREA)

Abstract

The invention discloses a communication system signal sequence detection method based on a cooperative time-varying bidirectional recurrent neural network, which comprises the following steps: preprocessing a signal sequence received by a receiving end of a communication system, inputting the preprocessed data sequence into a cooperative time-varying bidirectional cyclic neural network, and performing soft decision through network output. The preceding and following layers of the hidden layer of the cooperative time-varying bidirectional cyclic neural network are in a cooperative structure, and time-varying weights are used for combination when the preceding and following networks of the last layer of the neural network are combined to obtain the final output of the neural network.

Description

Communication system signal sequence detection method based on cooperative time-varying bidirectional cyclic neural network
Technical Field
The invention relates to the technical field of communication system signal detection, in particular to a method for detecting a communication system signal sequence based on a cooperative time-varying bidirectional recurrent neural network.
Background
In modern digital communication systems, signal detection is an important component. The signal is coded and modulated at the transmitting end, then transmitted out, and comes to the receiving end after passing through the channel, and at this time, the signal received by the receiving end is a signal subjected to noise interference or intersymbol interference, so that the received signal needs to be judged by using signal detection. In the traditional wireless communication, electromagnetic waves are used for transmitting information, and a propagation mechanism can be described by Maxwell equations, so that a probability statistical model of a channel can be established by using a mathematical formula, and a signal detection algorithm can be designed according to the probability statistical model of the channel. However, in some new communication systems, such as underwater acoustic communication, molecular communication, etc., or in conventional systems where the channel conditions are relatively complex, it is difficult to effectively model the propagation of signals. It is of practical importance in these cases to design the signal detection method using a method that does not rely on a probabilistic statistical model of the channel.
Deep learning among machine learning is more suitable for a signal detection problem in a communication system in which channel modeling is difficult. The advantage of deep learning is that its hidden layer can fit any function within a certain error range, with great flexibility. The existing neural network structure is not completely suitable for the problem of signal detection of a communication system, so that the method for detecting the signal of the communication system based on the cooperative time-varying bidirectional cyclic neural network has important significance.
Disclosure of Invention
In order to overcome the defects of the prior art, the invention aims to provide a communication system signal sequence detection method based on a cooperative time-varying bidirectional recurrent neural network, which can improve the accuracy of signal detection.
In order to achieve the purpose, the invention adopts the technical scheme that:
communication system signal sequence based on cooperative time-varying bidirectional cyclic neural networkA column detection method, wherein a finite set of transmission signals of the communication system is Γ ═ s1,s2,...,smIs of s1,s2,...,smM symbols are total, and the transmitting symbol of a transmitting end of the communication system at the time k is xkDefining the transmitted signal at that time as pk＝[1(xk＝s1),1(xk＝s2),...,1(xk＝sm)]Where 1 (-) represents the indicator function, at pkIn the vector, only one element is 1, the other elements are all zero, the length of the transmission sequence is K, and the transmission sequence is PK＝[p1,p2,...,pK]TThe receiving end of the communication system receives the received signal at the time k as yk＝[uk1,uk2,...,ukl]The signal is a vector of length l, and the signal sequence is YK＝[y1,y2,...,yK]TThe signal sequence detection method is characterized by comprising the following steps:
1) preprocessing the mean value of the sequence received by a receiving end, firstly calculating the mean value of the signal sequence:
then, the average value is subtracted from each data in the received sequence to obtain the data sequence IK：
Figure BDA0002312804800000022
Wherein IK＝[i1,i2,...,iK]T，
Figure BDA0002312804800000023
2) Data sequence IKInputting into a N-layer neural network using a cooperative time-varying bidirectional recurrent neural network structure, the number of neurons in the output layer of the network being mThe neuron is one of RNN, GRU, LSTM or other recurrent neural network neurons, and the output of the neural network at the time k is ok＝[zk1,zk2,...,zkm]Wherein z iskiTypically, the transmitted signal at time k is siThe probability of (c) is then the sequence of the neural network output is OK＝[o1,o2,...,oK]T；
3) Output sequence O of neural networkKMaking soft decisions, i.e. outputting o for time kk＝[zk1,zk2,...,zkm]To say, first, find out
Figure BDA0002312804800000031
The transmission signal is determined as
Figure BDA0002312804800000032
The decision to transmit a signal sequence
Figure BDA0002312804800000033
And completing sequence detection.
The cooperative time-varying bidirectional recurrent neural network structure in the step 2) comprises the following parts:
1) the input sequence of the neural network is IK＝[i1,i2,...,iK]TWherein for time k, the input to the neural network isThe number of the neurons of the input layer of the neural network is l;
2) the hidden layer of the neural network is N, a cooperative structure is formed between the front layer and the rear layer of the hidden layer of the neural network, and at the kth moment, the mode of the propagation from the nth layer of the network to the N +1 layer is as follows:
Figure BDA0002312804800000035
Figure BDA0002312804800000036
wherein
Figure BDA0002312804800000037
Is the input to the forward network of layer n +1 at time k,
Figure BDA0002312804800000038
for the input to the backward network of layer n +1 at time k,
Figure BDA0002312804800000039
is the output of the forward network of the nth layer at time k,
Figure BDA00023128048000000310
the output of the backward network of the nth layer at the kth time,
Figure BDA00023128048000000311
in order to combine the outputs of the nth layer, m is the cooperation length, which represents that the outputs of the current time and the previous m-1 times of the nth layer are used for cooperation when the nth layer is propagated to the n +1 layer;
3) merging the backward and forward networks of the last layer of the neural network by using a time-varying weight to obtain the final output of the neural network, and outputting o of the neural network at the kth momentkComprises the following steps:
Figure BDA00023128048000000312
wherein
Figure BDA00023128048000000313
Represents the output of the forward network of the last layer,
Figure BDA00023128048000000314
output of backward network representing the last layer, [, ]]Denotes the concatenation of the two outputs, w (k) denotes the time-varying weights;
the method for acquiring the time-varying weight comprises the following steps of constructing a fully-connected neural network, taking time k as input, and outputting the time-varying weight W (k):
W(k)＝NNfully(k) (4)
wherein NNfully(. cndot.) is a function represented by a fully-connected neural network.
The invention has the beneficial effects that:
the invention relates to a communication system signal sequence detection method based on a cooperative time-varying bidirectional recurrent neural network, which inputs a cooperative time-varying bidirectional recurrent neural network after preprocessing a received signal, adopts a cooperative structure between the front layer and the rear layer of the neural network, ensures that the front layer and the rear layer cooperate with each other in the process of propagating to the rear layer, considers the time-varying property of the judgment reliability of the front layer and the rear layer when the front layer and the rear layer of the neural network are combined, uses the time-varying weight to combine, and improves the accuracy of signal detection.
Drawings
Fig. 1 is a schematic structural diagram of a cooperative time-varying bidirectional recurrent neural network.
Detailed Description
The present invention will be described in further detail with reference to the accompanying drawings.
The invention discloses a communication system signal sequence detection method based on a cooperative time-varying bidirectional cyclic neural network, wherein the finite set of the transmitted signals of the communication system is gamma(s)1,s2,...,smThe sending signal of the sending end of the communication system at the time k is pk＝[1(xk＝s1),1(xk＝s2),...,1(xk＝sm)]Where 1 (-) represents the indicator function, at pkIn the vector, only one element is 1, the other elements are all zero, the length of the transmission sequence is K, and the transmission sequence is PK＝[p1,p2,...,pK]T. The receiving end of the communication system receives the received signal at the time k as yk＝[uk1,uk2,...,ukl]Then the signal sequence is YK＝[y1,y2,...,yK]TThe signal sequence detection method is characterized by comprising the following steps:
1) preprocessing the mean value of the sequence received by a receiving end, firstly calculating the mean value of the signal sequence:
Figure BDA0002312804800000051
then, the average value is subtracted from each data in the received sequence to obtain the data sequence IK：
Figure BDA0002312804800000052
Wherein IK＝[i1,i2,...,iK]T，
Figure BDA0002312804800000053
2) Data sequence IKInputting the data into a N-layer neural network using a cooperative time-varying bidirectional recurrent neural network structure, wherein the neurons in the network can be RNN, GRU, LSTM or the rest of recurrent neural network neurons. The output of the neural network at time k is ok＝[zk1,zk2,...,zkm]Wherein z iskiTypically, the transmitted signal at time k is siThe probability of (c) is then the sequence of the neural network output is OK＝[o1,o2,...,oK]T。
3) Output sequence O of neural networkKMaking soft decisions, i.e. outputting o for time kk＝[zk1,zk2,...,zkm]To say, first, find outThe transmission signal is determined as
Figure BDA0002312804800000055
The decision to transmit a signal sequence
Figure BDA0002312804800000056
And completing sequence detection.
The cooperative time-varying bidirectional cyclic neural network structure in the step 2) comprises the following parts:
1) the input sequence of the neural network is IK＝[i1,i2,...,iK]TWherein for time k, the input to the neural network is
Figure BDA0002312804800000057
The number of the neurons of the input layer of the neural network is l;
2) the hidden layer of the neural network is N, a cooperative structure is formed between the front layer and the rear layer of the hidden layer of the neural network, and at the kth moment, the mode of the propagation from the nth layer of the network to the N +1 layer is as follows:
Figure BDA0002312804800000058
Figure BDA0002312804800000059
wherein
Figure BDA00023128048000000510
Is the input to the forward network of layer n +1 at time k,
Figure BDA00023128048000000511
for the input to the backward network of layer n +1 at time k,
Figure BDA00023128048000000512
is the output of the forward network of the nth layer at time k,
Figure BDA00023128048000000513
the output of the backward network of the nth layer at the kth time,
Figure BDA00023128048000000514
in order to combine the outputs of the nth layer, m is the cooperation length, which represents that the outputs of the current time and the previous m-1 times of the nth layer are used for cooperation when the nth layer is propagated to the n +1 layer;
3) merging the backward and forward networks of the last layer of the neural network by using a time-varying weight to obtain the final output of the neural network, and outputting o of the neural network at the kth momentkComprises the following steps:
Figure BDA0002312804800000061
wherein
Figure BDA0002312804800000062
Represents the output of the forward network of the last layer,
Figure BDA0002312804800000063
output of backward network representing the last layer, [, ]]Denotes the concatenation of the two outputs, w (k) denotes the time-varying weights;
the method for acquiring the time-varying weight value comprises the following steps of constructing a full-connection neural network, taking time k as input, and outputting the time-varying weight value W (k):
W(k)＝NNfully(k) (4)
wherein NNfully(. cndot.) is a function represented by a fully-connected neural network.

Claims (3)

1. A communication system signal sequence detection method based on a cooperative time-varying bidirectional cyclic neural network is characterized in that a finite set of transmission signals of the communication system is gamma { s ═ s1,s2,...,smIs of s1,s2,...,smM symbols are total, and the transmitting symbol of a transmitting end of the communication system at the time k is xkDefining the transmitted signal at that time as pk＝[1(xk＝s1),1(xk＝s2),...,1(xk＝sm)]Where 1 (-) represents the indicator function, at pkIn a vector, there is only one elementThe element is 1, the rest elements are zero, the length of the transmission sequence is K, and the transmission sequence is PK＝[p1,p2,...,pK]TThe receiving end of the communication system receives the received signal at the time k as yk＝[uk1,uk2,...,ukl]The signal is a vector of length l, and the signal sequence is YK＝[y1,y2,...,yK]TThe signal sequence detection method is characterized by comprising the following steps:
1) preprocessing the mean value of the sequence received by a receiving end, firstly calculating the mean value of the signal sequence:
then, the average value is subtracted from each data in the received sequence to obtain the data sequence IK：
Figure FDA0002312804790000012
Wherein IK＝[i1,i2,...,iK]T，
Figure FDA0002312804790000013
2) Data sequence IKInputting into a N-layer neural network using cooperative time-varying bidirectional recurrent neural network structure, the number of neurons in the output layer of the network is m, the neuron in the network is one of RNN, GRU, LSTM or other recurrent neural network neurons, and the output of the neural network at time k is ok＝[zk1,zk2,...,zkm]Wherein z iskiTypically, the transmitted signal at time k is siThe probability of (c) is then the sequence of the neural network output is OK＝[o1,o2,...,oK]T；
3) Output sequence O of neural networkKMaking soft decisions, i.e. outputting o for time kk＝[zk1,zk2,...,zkm]To say, first, find out
Figure FDA0002312804790000014
The transmission signal is determined as
Figure FDA0002312804790000015
The decision to transmit a signal sequence
Figure FDA0002312804790000021
And completing sequence detection.
2. The method for detecting the signal sequence of the communication system based on the cooperative time-varying bidirectional cyclic neural network as claimed in claim 1, wherein the cooperative time-varying bidirectional cyclic neural network structure in step 2) comprises the following parts:
1) the input sequence of the neural network is IK＝[i1,i2,...,iK]TWherein for time k, the input to the neural network isThe number of the neurons of the input layer of the neural network is l;
2) the hidden layer of the neural network is N, a cooperative structure is formed between the front layer and the rear layer of the hidden layer of the neural network, and at the kth moment, the mode of the propagation from the nth layer of the network to the N +1 layer is as follows:
Figure FDA0002312804790000023
Figure FDA0002312804790000024
wherein
Figure FDA0002312804790000025
Is the input to the forward network of layer n +1 at time k,
Figure FDA0002312804790000026
for the input to the backward network of layer n +1 at time k,
Figure FDA0002312804790000027
is the output of the forward network of the nth layer at time k,
Figure FDA0002312804790000028
the output of the backward network of the nth layer at the kth time,in order to combine the outputs of the nth layer, m is the cooperation length, which represents that the outputs of the current time and the previous m-1 times of the nth layer are used for cooperation when the nth layer is propagated to the n +1 layer;
3) merging the backward and forward networks of the last layer of the neural network by using a time-varying weight to obtain the final output of the neural network, and outputting o of the neural network at the kth momentkComprises the following steps:
Figure FDA00023128047900000210
wherein
Figure FDA00023128047900000211
Represents the output of the forward network of the last layer,output of backward network representing the last layer, [, ]]Indicating that the two outputs are spliced together, w (k) represents a time-varying weight.
3. The method for detecting the signal sequence of the communication system based on the cooperative time-varying bidirectional recurrent neural network as claimed in claim 1, wherein the time-varying weight is obtained by constructing a fully-connected neural network, inputting time k, and outputting a time-varying weight w (k):
W(k)＝NNfully(k) (4)
wherein NNfully(. cndot.) is a function represented by a fully-connected neural network.
CN201911265807.7A 2019-06-06 2019-12-11 Communication signal detection method based on cooperative time-varying bidirectional recurrent neural network Active CN110826703B (en)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN201910492246 2019-06-06
CN2019104922468 2019-06-06

Publications (2)

Publication Number Publication Date
CN110826703A true CN110826703A (en) 2020-02-21
CN110826703B CN110826703B (en) 2022-06-07

Family

ID=69544670

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201911265807.7A Active CN110826703B (en) 2019-06-06 2019-12-11 Communication signal detection method based on cooperative time-varying bidirectional recurrent neural network

Country Status (1)

Country Link
CN (1) CN110826703B (en)

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111711585A (en) * 2020-06-11 2020-09-25 西安交通大学 Real-time signal sequence detection method based on deep learning
CN113300904A (en) * 2021-04-14 2021-08-24 浙江工业大学 Mobile molecular communication bidirectional network performance analysis method based on amplification forwarding and analog network coding strategies

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN106647808A (en) * 2017-01-05 2017-05-10 南宁市健佳网络科技有限公司 Method for searching AUVs and allocating and controlling capturing tasks based on fuzzy control algorithm
CN109525369A (en) * 2018-11-28 2019-03-26 电子科技大学 A kind of channel coding type blind-identification method based on Recognition with Recurrent Neural Network
CN109670580A (en) * 2018-12-21 2019-04-23 浙江工业大学 A kind of data recovery method based on time series
US20190138887A1 (en) * 2017-11-01 2019-05-09 Board Of Trustees Of Michigan State University Systems, methods, and media for gated recurrent neural networks with reduced parameter gating signals and/or memory-cell units

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN106647808A (en) * 2017-01-05 2017-05-10 南宁市健佳网络科技有限公司 Method for searching AUVs and allocating and controlling capturing tasks based on fuzzy control algorithm
US20190138887A1 (en) * 2017-11-01 2019-05-09 Board Of Trustees Of Michigan State University Systems, methods, and media for gated recurrent neural networks with reduced parameter gating signals and/or memory-cell units
CN109525369A (en) * 2018-11-28 2019-03-26 电子科技大学 A kind of channel coding type blind-identification method based on Recognition with Recurrent Neural Network
CN109670580A (en) * 2018-12-21 2019-04-23 浙江工业大学 A kind of data recovery method based on time series

Non-Patent Citations (2)

* Cited by examiner, † Cited by third party
Title
MARK EISEN 等: "Learning Optimal Resource Allocations in Wireless Systems", 《IEEE TRANSACTIONS ON SIGNAL PROCESSING》 *
杨文臣 等: "多智能体强化学习在城市交通网络信号控制方法中的应用综述", 《计算机应用研究》 *

Cited By (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111711585A (en) * 2020-06-11 2020-09-25 西安交通大学 Real-time signal sequence detection method based on deep learning
CN111711585B (en) * 2020-06-11 2021-06-22 西安交通大学 Real-time signal sequence detection method based on deep learning
CN113300904A (en) * 2021-04-14 2021-08-24 浙江工业大学 Mobile molecular communication bidirectional network performance analysis method based on amplification forwarding and analog network coding strategies

Also Published As

Publication number Publication date
CN110826703B (en) 2022-06-07

Similar Documents

Publication Publication Date Title
CN110826703B (en) Communication signal detection method based on cooperative time-varying bidirectional recurrent neural network
CN108566257B (en) Signal recovery method based on back propagation neural network
CN104714920B (en) A kind of signal acceptance method of adaptive baud rate
CN110932809B (en) Fiber channel model simulation method, device, electronic equipment and storage medium
CN109525254B (en) Convolutional code soft decision decoding method based on deep learning
CN110061946B (en) Depth signal detection method for high-speed rail
CN102638282A (en) Impedance and gain compensation device and method for transmission interface
CN114143040B (en) Antagonistic signal detection method based on multichannel characteristic reconstruction
CN109525253B (en) Convolutional code decoding method based on deep learning and integration method
US20220321319A1 (en) Skew detection and correction for orthogonal differential vector signaling codes
CN112468230A (en) Wireless ultraviolet light scattering channel estimation method based on deep learning
CN112422208B (en) Signal detection method based on antagonistic learning under unknown channel model
CN114936976A (en) Restoration method for generating anti-network haze image based on memory perception module
CN111711585B (en) Real-time signal sequence detection method based on deep learning
CN111209999B (en) Contact network performance degradation prediction method based on recurrent neural network
CN113890633B (en) Underwater acoustic communication system self-adaptive selection method based on deep neural network
CN113608264B (en) Joint pre-stack elastic inversion parameter and depth network target inversion method
CN116298657A (en) Multi-branch power distribution network fault positioning method under weak fault traveling wave
CN113645166B (en) Molecular communication method and receiver for molecular communication
CN112821971A (en) Time-varying channel signal detection method based on countermeasure learning
CN113516170A (en) Image classification method based on Bayesian neural network random addition saturation structure
CN112329523A (en) Underwater acoustic signal type identification method, system and equipment
CN111797750A (en) Elevator car sill and well inner surface distance measuring method based on BP neural network
CN111883173A (en) Audio packet loss repairing method, device and system based on neural network
CN114157544B (en) Frame synchronization method, device and medium based on convolutional neural network

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