CN104270167A - Signal detection and estimation method based on multi-dimensional characteristic neural network - Google Patents
Signal detection and estimation method based on multi-dimensional characteristic neural network Download PDFInfo
- Publication number
- CN104270167A CN104270167A CN201310704841.6A CN201310704841A CN104270167A CN 104270167 A CN104270167 A CN 104270167A CN 201310704841 A CN201310704841 A CN 201310704841A CN 104270167 A CN104270167 A CN 104270167A
- Authority
- CN
- China
- Prior art keywords
- signal
- code
- dsss
- sequence
- cycle
- 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.)
- Pending
Links
Landscapes
- Radar Systems Or Details Thereof (AREA)
Abstract
The invention discloses a signal detection and estimation method based on a multi-dimensional characteristic neural network, and relates to the technical fields of artificial neural networks and direct sequence spread spectrum (DSSS) communication. According to the method, the presence of a DSSS signal is detected by using a correlation function accumulation method, and the estimation of a PN (Pseudo-Noise) code is finished, namely, a direct spread signal spread spectrum code is detected under the situation of low signal-to-noise ratio based on a multi-dimensional characteristic neural network method. The method comprises the following steps: extracting the multi-dimensional characteristics of a signal (comprising the period of the PN code, the width of a code element, the starting and stopping moments of the synchronization of the PN code and an information code, and the like); periodically segmenting the signal; inputting the segmented signals into the neural network in batches for training; and training and converging the segmented signals through the neural network to detect a spread spectrum sequence. Through adoption of the method, a signal can be detected from noise and a PN sequence of the signal can be estimated in non-cooperative communication through an artificial neural network; and moreover, the network can quickly converge to a sequence or a reverse sequence thereof under the situation of low signal-to-noise ratio.
Description
Technical field
What the present invention relates to is artificial neural net and Direct Sequence Spread Spectrum Communication technical field, is specifically related to a kind of input based on multidimensional characteristic neural net and method of estimation.
Background technology
Direct sequence spread spectrum (Direct Sequence Spread Spectrum, DSSS) signal of communication utilizes the frequency expansion sequence of two-forty to be multiplied with information code sequence to obtain, make signal spectrum broadened and reduce power spectrum density, negative signal to noise ratio environment can be worked in and signal is submerged among noise, there is strong anti-interference, the advantage such as anti-multipath, low probability of intercept, multiple access multiplexing, be widely used in the military and civilian communications field.First DSSS communication come across during the Second World War, and along with the appearance of wireless communication technology and new unit, obtain a wide range of applications, comprise the every field such as military communication, satellite communication, mobile communication, navigation tracking, range finding, remote control from after the war.
For cooperation DSSS communication source, known frequency expansion sequence can be utilized to carry out to received signal detecting to extract transmission information, the frequency spectrum of interchannel noise and narrow-band interference signal then due to broadened, be easy to suppressed fall; But for non-cooperation recipient, what be difficult to realize transmission information due to frequency expansion sequence the unknown has efficient recovery, and need the estimation completing frequency expansion sequence in advance, therefore how carrying out effectively detecting to direct sequence signal is the important topic that communication countermeasures faces.
In recent years along with the development of nonlinear science, neural net becomes more and more extensive in the application in signal transacting field.Research shows have some artificial neural network systems to have extremely sensitivity characteristic to the small-signal under low signal-to-noise ratio, and have immunity to noise again, this makes it in input, very have potentiality simultaneously.That is, in detection of DS, incorporate the method based on chaology such as neural net detection method, the some shortcomings of common detection methods can be made up, and be expected more effectively to detect direct sequence signal under low signal-to-noise ratio.The present invention, for the direct sequence spread spectrum of binary phase shift keying (DSSS/BPSK) signal, have studied the detection of DS algorithm based on multidimensional characteristic and neural net and parameter Estimation.
To input and the research of estimation come from communicate in the needs of the detection to cooperation signal of communication and the detection to radar signal, to make under certain conditions, obtain best detection perform.The research of this respect long ago just starts, and has shown that the structure of optimal detector is correlator or matched filter.This optimum detection needs a lot of conditions: signal will know, as known in the carrier frequency of signal, baseband waveform; Time of advent of signal is known etc.But in noncooperative scouting detects, do not have these priori conditions, this optimum detection is not applicable.
Traditional receiver is that frequency domain character is to the received signal analyzed, if there is the composition of some exceptions on frequency domain, just illustrates and has signal, otherwise illustrate that signal does not exist.The basis of this method is also come from correlator.Because signal is unknown, going with regard to there being a lot of correlators to mate to received signal, if found to wherein a kind of signal is very relevant, then illustrating at this section of time memory at signal; Otherwise, then signal is not had.When the form of unknown signaling, we are sinusoidal wave forms with regard to putative signal, but this method Direct Sequence Spread Spectrum Signal very large to spreading gain seems helpless when signal to noise ratio is very low.
Summary of the invention
For the deficiency that prior art exists, the present invention seeks to be to provide a kind of input based on multidimensional characteristic neural net and method of estimation, pass through artificial neural net, can in non-cooperative communication, from noise, detect signal and estimate its PN sequence, and when low signal-to-noise ratio, what network also can be very fast converge to sequence or it is anti-.
To achieve these goals, the present invention realizes by the following technical solutions: a kind of input based on multidimensional characteristic neural net and method of estimation, the steps include: that the data that DSSS signal observes are divided into K section by (1), calculate the auto-correlation function R of kth segment data
k(τ), 1≤k<K;
(2) to the auto-correlation function R of K segment data altogether
k(τ) be averaged and obtain R (τ);
(3) if | R (τ) | in have an obvious spike, show to there is DSSS signal, otherwise there is not DSSS signal, now need again to extract observation data, repeat step (1) to step (3) until detect DSSS signal;
(4) because DSSS signal is periodic signal, thus there is discrete power spectrum, again Fourier transform delivery square is done to this power spectrum, namely secondary power is asked to compose, it can obtain sharp-pointed arteries and veins at PN code cycle integral multiple place, just can be estimated the cycle of DSSS signal PN code by the distance detected between these poops.
Visible, the sharp-pointed triangular pulse sequence in the secondary power of DSSS useful signal spectrum to be spacing be PN code cycle, and the Quadratic Spectrum of noise n (t) does not have this characteristic.Therefore the cycle T of PN code can be estimated by detecting these interpulse distances
0.
(5) because DSSS useful signal ds (t) is cyclic stationary process, PN chip width is then the important periodic characteristic parameter of this process one, and the cyclic spectrum of ds (t) is:
The cyclic spectrum of DSSS useful signal ds (t) can be found out on α axle with α=± k/T by (b) formula
cdiscrete distribution, and amplitude taper, the cyclic spectrum of noise n (t) does not then have this character.If the maximum position of spectral line α=1/T of amplitude except zero-frequency can be searched along α axle
cthen can calculate the PN chip width T that will estimate
c=1/ α.
(6) because adopted DSSS useful signal ds (t) carrys out synchronous modulation information code with the PN code in a cycle, after having estimated the parameter such as code cycle, chip width of signal PN code, after needing to carry out segmentation to the received signal, be just convenient to follow-up process.Visible, also must estimate the start/stop time T that PN code is synchronous with information code
x.On average can come reliably to estimate this parameter by K time, and by average all right restraint speckle.That is:
(7) after obtaining the characteristic parameter such as cycle of DSSS signal PN code, chip width and information code and the synchronous start/stop time of PN code, just primary signal can be carried out periodic segment, again the signal after segmentation is trained by batch to send in Hebb network, and then estimate frequency expansion sequence or it is anti-.Shown in Fig. 3 is the neural net that the present invention uses, and has two hypothesis below:
(71) neuron of output layer is linear
(72) this network has m to input and l output, general requirement l<m.
If output layer neuron j is y in the output in n moment
j(n), input point set { x
i(n) | i=1,2 ..., the direct cynapse weight vector of m} and output layer neuron j is w
jin (), then have:
Cynapse weight vector carries out self-adaptative adjustment according to Hebb algorithm:
Wherein η is learning parameter.Formula (e) is generalized Hebbian algorithm, its output layer has l neuron, for the neural net using generalized Hebbian algorithm, can extract front l principal component vector of input data, the convergency value that namely a jth output layer neuronic cynapse weight vector is final is a jth principal component vector of input data.
The present invention has following beneficial effect: what the present invention adopted is the Hebbian learning process of self-organizing system, the neural network structure realized with this algorithm generally has the function of pivot analysis (PCA), also PCA neural net can be called, utilize the algorithm of neural net to obtain the PN code sequence of DSSS signal, its principle be based upon multidimensional characteristic analyze basis on, the multidimensional characteristic parameters input neural net of the signal obtained, under CHA learning algorithm, the synaptic weight of this neural net will converge on PN code sequence itself, thus PN code sequence can be gone out by these weights estimation.By artificial neural net, in non-cooperative communication, can detect signal and estimate its PN sequence from noise, and when low signal-to-noise ratio, what network also can be very fast converge to sequence or it is anti-.
Accompanying drawing explanation
The present invention is described in detail below in conjunction with the drawings and specific embodiments;
Fig. 1 is flow chart of the present invention;
Fig. 2 is the DSSS signal (autocorrelation function graph of signal and the power spectrum of signal) that the present invention detects;
Fig. 3 is the feedforward neural network that the present invention uses Hebb algorithm;
Fig. 4 is based on the network-evaluated PN code sequence chart of the PCA of Hebb algorithm;
Fig. 5 is the performance evaluation figure (error rate of SNR=-10dB) based on the multi output PCA network of Hebb algorithm;
Fig. 6 is the performance evaluation figure (error rate of SNR=-14dB) based on the multi output PCA network of Hebb algorithm.
Embodiment
The technological means realized for making the present invention, creation characteristic, reaching object and effect is easy to understand, below in conjunction with embodiment, setting forth the present invention further.
With reference to Fig. 1-6, this embodiment by the following technical solutions: 1, the DSSS information data of catching is divided into K section, calculates the auto-correlation function R of kth segment data
k(τ), 1≤k<K; Get K=100, calculate the auto-correlation function mean value R (τ) of this 100 segment data, if | R (τ) | there is no obvious spike, illustrate in the data observed not containing DSSS signal, need again to observe, until can be | R (τ) | in obvious spike detected.As shown in Figure 2.
2, according to formula (a), compose by the secondary power of DSSS signal the cycle T that interpulse distance estimates PN code
0.
3, according to formula (b), the maximum position of spectral line α=1/T of amplitude except zero-frequency is composed along the α axle search DSSS signal period
c, obtain the PN chip width T that will estimate
c=1/ α.
4, according to formula (c), the PN code start/stop time T synchronous with information code is gone out by the averaged power spectrum of 100 times
x.
5, primary signal is divided into 100 sections, trains by the neural net shown in crowd feeding Fig. 3.To the performance evaluation of network, training result as shown in Figure 4, can find out that the present invention well can detect when low signal-to-noise ratio and estimate the PN code of signal by such as Fig. 5 and Fig. 6.
What this embodiment adopted is the Hebbian learning process of self-organizing system, the neural network structure realized with this algorithm generally has the function of pivot analysis (PCA), also PCA neural net can be called, utilize the algorithm of neural net to obtain the PN code sequence of DSSS signal, its principle be based upon multidimensional characteristic analyze basis on, the multidimensional characteristic parameters input neural net of the signal obtained, under CHA learning algorithm, the synaptic weight of this neural net will converge on PN code sequence itself, thus PN code sequence can be gone out by these weights estimation.By artificial neural net, in non-cooperative communication, can detect signal and estimate its PN sequence from noise, and when low signal-to-noise ratio, what network also can be very fast converge to sequence or it is anti-.
More than show and describe general principle of the present invention and principal character and advantage of the present invention.The technical staff of the industry should understand; the present invention is not restricted to the described embodiments; what describe in above-described embodiment and specification just illustrates principle of the present invention; without departing from the spirit and scope of the present invention; the present invention also has various changes and modifications, and these changes and improvements all fall in the claimed scope of the invention.Application claims protection range is defined by appending claims and equivalent thereof.
Claims (1)
1. based on input and the method for estimation of multidimensional characteristic neural net, it is characterized in that, the data that DSSS signal observes are divided into K section by (1), calculate the auto-correlation function R of kth segment data
k(τ), 1≤k<K;
(2) to the auto-correlation function R of K segment data altogether
k(τ) be averaged and obtain R (τ);
(3) if | R (τ) | in have an obvious spike, show to there is DSSS signal, otherwise there is not DSSS signal, now need again to extract observation data, repeat step (1) to step (3) until detect DSSS signal;
(4) because DSSS signal is periodic signal, thus there is discrete power spectrum, again Fourier transform delivery square is done to this power spectrum, namely secondary power is asked to compose, it can obtain sharp-pointed arteries and veins at PN code cycle integral multiple place, just can be estimated the cycle of DSSS signal PN code by the distance detected between these poops.
Visible, the sharp-pointed triangular pulse sequence in the secondary power of DSSS useful signal spectrum to be spacing be PN code cycle, and the Quadratic Spectrum of noise n (t) does not have this characteristic.Therefore the cycle T of PN code can be estimated by detecting these interpulse distances
0.
(5) because DSSS useful signal ds (t) is cyclic stationary process, PN chip width is then the important periodic characteristic parameter of this process one, and the cyclic spectrum of ds (t) is:
The cyclic spectrum of DSSS useful signal ds (t) can be found out on α axle with α=± k/T by (b) formula
cdiscrete distribution, and amplitude taper, the cyclic spectrum of noise n (t) does not then have this character.If the maximum position of spectral line α=1/T of amplitude except zero-frequency can be searched along α axle
cthen can calculate the PN chip width T that will estimate
c=1/ α.
(6) because adopted DSSS useful signal ds (t) carrys out synchronous modulation information code with the PN code in a cycle, after having estimated the parameter such as code cycle, chip width of signal PN code, after needing to carry out segmentation to the received signal, be just convenient to follow-up process.Visible, also must estimate the start/stop time T that PN code is synchronous with information code
x.On average can come reliably to estimate this parameter by K time, and by average all right restraint speckle.That is:
(7) after obtaining the characteristic parameter such as cycle of DSSS signal PN code, chip width and information code and the synchronous start/stop time of PN code, just primary signal can be carried out periodic segment, again the signal after segmentation is trained by batch to send in Hebb network, and then estimate frequency expansion sequence or it is anti-.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201310704841.6A CN104270167A (en) | 2013-12-20 | 2013-12-20 | Signal detection and estimation method based on multi-dimensional characteristic neural network |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201310704841.6A CN104270167A (en) | 2013-12-20 | 2013-12-20 | Signal detection and estimation method based on multi-dimensional characteristic neural network |
Publications (1)
Publication Number | Publication Date |
---|---|
CN104270167A true CN104270167A (en) | 2015-01-07 |
Family
ID=52161668
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201310704841.6A Pending CN104270167A (en) | 2013-12-20 | 2013-12-20 | Signal detection and estimation method based on multi-dimensional characteristic neural network |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN104270167A (en) |
Cited By (9)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN106100693A (en) * | 2016-05-31 | 2016-11-09 | 东南大学 | A kind of direct sequence signal chip width method of estimation |
CN106849992A (en) * | 2016-12-12 | 2017-06-13 | 西安空间无线电技术研究所 | A kind of detection method of the direct sequence signal based on Generalized Quadratic power spectrum |
CN106877901A (en) * | 2017-03-30 | 2017-06-20 | 西安电子科技大学 | A kind of detection method of low noise than direct sequence signal |
CN107147603A (en) * | 2017-05-05 | 2017-09-08 | 西安电子科技大学 | DBPSK demodulation methods based on multiple neural network |
CN109543643A (en) * | 2018-11-30 | 2019-03-29 | 电子科技大学 | Carrier signal detection method based on one-dimensional full convolutional neural networks |
CN110099019A (en) * | 2019-04-24 | 2019-08-06 | 西安电子科技大学 | LoRa Modulation Signal Detection Method based on deep learning |
CN110278022A (en) * | 2019-05-23 | 2019-09-24 | 军事科学院系统工程研究院网络信息研究所 | A kind of simplification method of satellite communication radio channel resource monitoring |
CN110611627A (en) * | 2018-06-15 | 2019-12-24 | 维沃移动通信有限公司 | Signal detection method and receiving end |
CN113037411A (en) * | 2019-12-24 | 2021-06-25 | 清华大学 | Multi-user signal detection method and device based on deep learning |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20040059992A1 (en) * | 2002-06-17 | 2004-03-25 | Tan Keng Tiong | Methods of optimizing the decoding of signals based on a complete majority logic representation |
CN1731779A (en) * | 2005-08-05 | 2006-02-08 | 西安电子科技大学 | Wavelet packet multi-carrier spread-spectrum system and control method based on neural net equalizer |
CN101943324A (en) * | 2010-06-28 | 2011-01-12 | 东北大学 | Weak signal detection device and method based on wavelets and RBF neural network |
CN102868422A (en) * | 2012-09-07 | 2013-01-09 | 天津理工大学 | MMSE-BDFE (Minimum Mean Square Error-Blind Decision Feedback Equalizer) multi-user detection system based on neural network, and working method of MMSE-BDFE multi-user detection system |
-
2013
- 2013-12-20 CN CN201310704841.6A patent/CN104270167A/en active Pending
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20040059992A1 (en) * | 2002-06-17 | 2004-03-25 | Tan Keng Tiong | Methods of optimizing the decoding of signals based on a complete majority logic representation |
CN1731779A (en) * | 2005-08-05 | 2006-02-08 | 西安电子科技大学 | Wavelet packet multi-carrier spread-spectrum system and control method based on neural net equalizer |
CN101943324A (en) * | 2010-06-28 | 2011-01-12 | 东北大学 | Weak signal detection device and method based on wavelets and RBF neural network |
CN102868422A (en) * | 2012-09-07 | 2013-01-09 | 天津理工大学 | MMSE-BDFE (Minimum Mean Square Error-Blind Decision Feedback Equalizer) multi-user detection system based on neural network, and working method of MMSE-BDFE multi-user detection system |
Non-Patent Citations (1)
Title |
---|
张天骐等: "直扩信号的谱检测和神经网络估计", 《系统工程与电子技术》 * |
Cited By (17)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN106100693B (en) * | 2016-05-31 | 2018-05-08 | 东南大学 | A kind of direct sequence signal chip width method of estimation |
CN106100693A (en) * | 2016-05-31 | 2016-11-09 | 东南大学 | A kind of direct sequence signal chip width method of estimation |
CN106849992A (en) * | 2016-12-12 | 2017-06-13 | 西安空间无线电技术研究所 | A kind of detection method of the direct sequence signal based on Generalized Quadratic power spectrum |
CN106849992B (en) * | 2016-12-12 | 2019-06-18 | 西安空间无线电技术研究所 | A kind of detection method of the direct sequence signal based on Generalized Quadratic power spectrum |
CN106877901A (en) * | 2017-03-30 | 2017-06-20 | 西安电子科技大学 | A kind of detection method of low noise than direct sequence signal |
CN107147603B (en) * | 2017-05-05 | 2019-10-08 | 西安电子科技大学 | DBPSK demodulation method based on multiple neural network |
CN107147603A (en) * | 2017-05-05 | 2017-09-08 | 西安电子科技大学 | DBPSK demodulation methods based on multiple neural network |
CN110611627B (en) * | 2018-06-15 | 2021-03-19 | 维沃移动通信有限公司 | Signal detection method and receiving end |
CN110611627A (en) * | 2018-06-15 | 2019-12-24 | 维沃移动通信有限公司 | Signal detection method and receiving end |
CN109543643A (en) * | 2018-11-30 | 2019-03-29 | 电子科技大学 | Carrier signal detection method based on one-dimensional full convolutional neural networks |
CN109543643B (en) * | 2018-11-30 | 2022-07-01 | 电子科技大学 | Carrier signal detection method based on one-dimensional full convolution neural network |
CN110099019A (en) * | 2019-04-24 | 2019-08-06 | 西安电子科技大学 | LoRa Modulation Signal Detection Method based on deep learning |
CN110099019B (en) * | 2019-04-24 | 2020-04-07 | 西安电子科技大学 | LoRa modulation signal detection method based on deep learning |
CN110278022A (en) * | 2019-05-23 | 2019-09-24 | 军事科学院系统工程研究院网络信息研究所 | A kind of simplification method of satellite communication radio channel resource monitoring |
CN110278022B (en) * | 2019-05-23 | 2021-04-30 | 军事科学院系统工程研究院网络信息研究所 | Simplified method for monitoring satellite communication wireless channel resources |
CN113037411A (en) * | 2019-12-24 | 2021-06-25 | 清华大学 | Multi-user signal detection method and device based on deep learning |
CN113037411B (en) * | 2019-12-24 | 2022-02-15 | 清华大学 | Multi-user signal detection method and device based on deep learning |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN104270167A (en) | Signal detection and estimation method based on multi-dimensional characteristic neural network | |
CN104267379B (en) | A kind of active radar and passive radar based on Waveform Design works in coordination with anti-interference method | |
CN103885053B (en) | A kind of radar data moving-target based on tracking filter detects processing method | |
CN103746722B (en) | Method for estimating jump cycle and take-off time of frequency hopping signal | |
CN102662183B (en) | Method and system for global position system (GPS) signal capture | |
CN108919203B (en) | Radar active interference identification method and system | |
CN104143997B (en) | A kind of multi-access inference error resilience lock decision method based on expanding system frequently | |
CN106546949B (en) | A kind of double array element sinusoidal signal arrival bearing's estimation methods based on frequency estimation | |
CN102230961A (en) | Method for detecting spread target based on phase compensation processing | |
CN106879068A (en) | The arrival time method of estimation of signal under a kind of strong multi-path environment | |
CN110247867B (en) | Underwater acoustic Doppler estimation method and device, and underwater acoustic communication method and system | |
CN106936514A (en) | The underwater sound targeting signal detection method concentrated based on energy under condition of sparse channel | |
CN112083393A (en) | Intermittent sampling forwarding interference identification method based on spectrogram average time characteristic | |
CN106886011A (en) | A kind of MIMO radar Cramér-Rao lower bound computational methods for reflecting through wave action | |
CN108535686B (en) | Time difference positioning method for cutting off broadband linear frequency modulation signal | |
CN105553506A (en) | Fast acquisition method and device of long code spread spectrum signal | |
CN108809354A (en) | A kind of long-time accumulation method of high dynamic Doppler capture | |
CN105099499B (en) | Noise like Chirp spread-spectrum signals are designed and quick capturing method | |
CN105553635A (en) | BOC signal parameter blind estimation method based on cyclic autocorrelation | |
CN105656511B (en) | Differential correlation acquisition method suitable for environment with frequency offset and low signal-to-noise ratio | |
Hu et al. | Coherent TDOA/FDOA estimation method for frequency-hopping signal | |
CN104360360A (en) | Efficient algorithm for acquiring and judging GPS (global positioning system) satellites | |
Yi-Jia et al. | Modeling and estimation of the space-time varying channels | |
CN104199064B (en) | Correlation fluctuation based BOC (binary offset carrier) signal parameter blind estimation method | |
CN101572564B (en) | Locally optimal detector based method for capturing pseudocode under weakly dependent non-Gaussian environment |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
DD01 | Delivery of document by public notice |
Addressee: Zhang Dong Document name: Notification to Make Rectification |
|
C06 | Publication | ||
PB01 | Publication | ||
C10 | Entry into substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
RJ01 | Rejection of invention patent application after publication | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20150107 |