CN109061591A - A kind of time-frequency line-spectrum detection method based on sequential cluster - Google Patents
A kind of time-frequency line-spectrum detection method based on sequential cluster Download PDFInfo
- Publication number
- CN109061591A CN109061591A CN201810811825.XA CN201810811825A CN109061591A CN 109061591 A CN109061591 A CN 109061591A CN 201810811825 A CN201810811825 A CN 201810811825A CN 109061591 A CN109061591 A CN 109061591A
- Authority
- CN
- China
- Prior art keywords
- frame
- line
- frequency
- spectrum
- serial number
- 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
Links
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01S—RADIO DIRECTION-FINDING; RADIO NAVIGATION; DETERMINING DISTANCE OR VELOCITY BY USE OF RADIO WAVES; LOCATING OR PRESENCE-DETECTING BY USE OF THE REFLECTION OR RERADIATION OF RADIO WAVES; ANALOGOUS ARRANGEMENTS USING OTHER WAVES
- G01S7/00—Details of systems according to groups G01S13/00, G01S15/00, G01S17/00
- G01S7/02—Details of systems according to groups G01S13/00, G01S15/00, G01S17/00 of systems according to group G01S13/00
- G01S7/41—Details of systems according to groups G01S13/00, G01S15/00, G01S17/00 of systems according to group G01S13/00 using analysis of echo signal for target characterisation; Target signature; Target cross-section
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01S—RADIO DIRECTION-FINDING; RADIO NAVIGATION; DETERMINING DISTANCE OR VELOCITY BY USE OF RADIO WAVES; LOCATING OR PRESENCE-DETECTING BY USE OF THE REFLECTION OR RERADIATION OF RADIO WAVES; ANALOGOUS ARRANGEMENTS USING OTHER WAVES
- G01S7/00—Details of systems according to groups G01S13/00, G01S15/00, G01S17/00
- G01S7/52—Details of systems according to groups G01S13/00, G01S15/00, G01S17/00 of systems according to group G01S15/00
- G01S7/539—Details of systems according to groups G01S13/00, G01S15/00, G01S17/00 of systems according to group G01S15/00 using analysis of echo signal for target characterisation; Target signature; Target cross-section
Landscapes
- Engineering & Computer Science (AREA)
- Computer Networks & Wireless Communication (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Radar, Positioning & Navigation (AREA)
- Remote Sensing (AREA)
- Measurement Of Mechanical Vibrations Or Ultrasonic Waves (AREA)
Abstract
The time-frequency line-spectrum detection method based on sequential cluster that the invention discloses a kind of, this method comprises the following steps: step 1: obtaining the sample data sequence of signal to be detected;Step 2: carrying out sub-frame processing to sample data sequence, downlink data upon handover sequence is obtained;Step 3: parameter initialization;Step 4: calculating frame data sequence power spectrum;Step 5: extracting single frames line spectral frequencies serial number, single frames line spectral frequencies serial number set is obtained;Step 6: repeating the 4th step to the 6th step, calculating whole frame data sequence power spectrum and extracting line spectral frequencies serial number set, the line spectrum serial number set of all frame data is obtained;Step 7: carrying out sequential cluster frame by frame to obtained all main feed line spectral frequency serial number set;Step 8: extracting time-frequency line spectrum data from cluster result.The correlation that time-frequency domain line spectrum occurs over time and frequency is utilized in detection method of the invention, and using the method for clustering tracking frame by frame, operand is small, practical, is suitble to handle signal in real time.
Description
Technical field
The invention belongs to signal processing technology field more particularly to a kind of time-frequency line-spectrum detection sides based on sequential cluster
Method.
Background technique
Time-frequency line-spectrum detection be the narrow band signal with certain duration is extracted in noise signal, and in the time and
The narrow band signal is described in two dimensional parameters of frequency.Time-frequency line-spectrum detection has in noise signal analysis widely answers
With from the time-frequency line spectrum obtained in noise the result is that target detection, target's feature-extraction and target are divided in radar and sonar system
The important information source of the processing such as class identification.
Domestic and foreign scholars propose many time-frequency line-spectrum detection methods at present, mainly have: (1) image processing method.Image
Processing method is related mature theoretical using field of image processing, such as edge detection algorithm, can extract the line in time-frequency figure
Track is composed, but it is suitable for strong edge signals, and fail to make full use of to the correlation of time-frequency line spectral frequencies in time, clock synchronization
Frequency line-spectrum detection performance is poor.(2) neural network method.For neural network method, the sufficiently large and reliable instruction by quantity
Practice sample, the model of line spectrum can be trained, and then extract time-frequency line spectrum feature.This method major defect is to need benefit in advance
It is trained with a large amount of sample data, and obtains model to unknown situation Generalization Ability deficiency, algorithm performance is largely
The upper quality depending on training sample.(3) statistical model method.It is main that there is maximum by calculating for statistical model method
The path of possibility determines line spectrum track, and typical algorithm is such as based on the line-spectrum detection method of Hidden Markov Model (HMM), should
Method performance accredited signal-to-noise ratio condition is affected, and the time-frequency line spectrum of a large amount of falsenesses is readily detected under Low SNR.
Summary of the invention
Goal of the invention: for problem and shortage existing for above-mentioned existing method, the present invention provides one kind based on sequential poly-
The time-frequency line-spectrum detection method of class, this method provide training sample without early period, time-frequency line spectrum energy are utilized in time and frequency
The regularity of distribution of rate dimension is handled the processing mode combined with Sequential processing using frame, enormously simplifies time-frequency line-spectrum detection
The complexity of processing can also obtain preferable time-frequency line-spectrum detection as a result, can satisfy radar under the conditions of lower signal-to-noise ratio
With the real time signal processing demand of sonar, there is very high practical value.
Technical solution: in order to achieve the above-mentioned object of the invention, the invention adopts the following technical scheme: a kind of be based on sequential cluster
Time-frequency line-spectrum detection method, include the following steps:
(1) the sample data sequence s (n), n=0,1 ..., N-1 of signal to be detected are obtained, the n is sampled point sequence
Number, N is number of sampling points;
(2) sub-frame processing is carried out to sample data sequence s (n), obtains downlink data upon handover sequence xi(m), i=0,1 ..., P-
1, m=0,1 ..., M-1, i be frame number, P is frame number, and m is the sampling sequence number of each frame data sequence, and M is that frame data sequence is long
Degree;
(3) parameter initialization, and currently processed frame number j=0 is set;
(4) jth frame data sequence x is calculatedj(m) power spectrum Pj(k), k be frequency serial number, k=0,1,2 ..., K-1,
K is less thanPositive integer, floor () be downward rounding operation;
(5) to power spectrum Pj(k) line-spectrum detection is carried out, line spectrum respective frequencies serial number is extracted, obtains jth main feed line spectral frequency sequence
Number set Rj;
(6) it is incremented by j, repeats step (4)~(6), the power spectrum P until calculating whole P frame data sequencesj(k), it and obtains
To P line spectral frequencies serial number set Rj, j=0,1,2 ... P-1;
(7) to obtained P frame data line spectral frequencies serial number set RjSequential cluster is carried out frame by frame;
(8) time-frequency line spectrum data are extracted from cluster result.
Wherein, in step (1), signal sample data sequence s (n) to be detected is obtained with the following method: from sensor
The acquisition data of N number of sampled point are received as data sequence s (n) to be detected;Or N number of sample point data is extracted from memory
As data sequence s (n) to be detected.
Wherein, in step (2), sub-frame processing is carried out to sample data sequence s (n), taking frame length is M, and sliding step is
B, sliding step B are the differences of the adjacent first data sequence number for taking frame data in sample data sequence s (n) twice, then can obtain
To downlink data upon handover sequence xi(m), the xi(m)=s (iB+m), i are frame number, and m is each frame data sequential sampling point serial number, P
For data frame number, N, B, the relationship of P, M meet N=(P-1) B+M.
Wherein, in step (3), line spectrum judgement window length, line-spectrum detection amplitude threshold, line spectrum are clustered with the following method
Thresholding, line-spectrum detection frame number thresholding are initialized: setting initialization line spectrum adjudicates the long W of windowl, line-spectrum detection amplitude threshold Gl, line
Spectral clustering thresholding Gc, line-spectrum detection frame number thresholding Gt, wherein WlIt is greater than 8 odd number, G for valuelIt is greater than 3 real number for value,
GcIt is greater than 3 integer, G for valuetIt is greater than 3 integer for value.
Wherein, in step (4), with the following method according to the data sequence xj(m) power spectrum signal P is calculatedj(k):
X is done to the data sequencej(m) the frequency spectrum X of data sequence is calculated in discrete Fourier transformj(k) and power spectrum Pj(k),
Include the following steps:
(4-1) calculates xj(m) leaf transformation in M point discrete Fourier, and the preceding K point of transformation results is taken, obtain data sequence xj
(m) frequency spectrum
Xj(k), k=0,1,2 ..., K-1
Wherein, K is less thanPositive integer, floor () be downward rounding operation;
(4-2) is according to Xj(k) data sequence x is calculatedj(m) power spectrum
| | represent modulus value operation.
Wherein, in step (5), using the method for the local anomaly value detection to maximum point to power spectrum Pj(k) into
Row line-spectrum detection, comprises the following steps that
(5-1) is in (wl-1)/2≤k≤K-1-(wl- 1)/2 P is found out in rangej(k) the corresponding frequency of all maximum points
Rate serial number km, meet Pj(km) > Pj(km- 1) and Pj(km)≥Pj(km+1);
(5-2) is to each kmIf Pj(km)≥Gl·std({Pj(km-(wl- 1)/2), Pj(km-(wl-1)/2+
..., P 1)j(km+(wl- 1)/2) }), then determine the frequency point for power spectrum Pj(k) line spectrum frequency point, by the frequency sequence of the frequency point
The line spectral frequencies serial number set R of current data frame j frame number is recordedjIn, it obtains:
Rj={ km|Pj(km) > Pj(km-1)&Pj(km)≥Pj(km+1)&Pj(km)≥Gl·std({Pj(km-(wl-1)/
2), Pj(km-(wl- 1)/2+1) ..., Pj(km+(wl- 1) logical AND relationship/2) } }, is indicated, std () is to take data acquisition system equal
The operation of root, WlLong, the G for line spectrum judgement windowlFor line-spectrum detection amplitude threshold.
Wherein, in step (7), to obtained line spectral frequencies serial number set RjSequential cluster, specific steps are carried out frame by frame
It is as follows:
(7-1) remembers R0Element number r0, definition tracker is an ordered set T={ (frame number, frequency serial number) }, wound
Build r0A tracker Ta, a=1,2..., r0, the line spectrum point set R that is detected using the 0th frame data0In r0A frequency serial number km,
Construct element (0, km) respectively as r0The first element of a tracker;
(7-2) is by frame number frame by frame to line spectral frequencies serial number set R1, R2..., RP-1It is clustered, step includes:
(7-2-1) enables currently processed frame number j=1;
(7-2-2) calculates the line spectral frequencies serial number set R of the jth frame in line spectral frequencies serial number setjMiddle each element is with before
One main feed line spectral frequency serial number set Rj-1The absolute value of each element frequency point difference defines jth main feed line spectral frequency serial number set RjQ
A frequency serial number Rj(q) with -1 main feed line spectral frequency serial number set R of jthj-1P-th of frequency serial number Rj-1(p) distance are as follows:
D(j)pq=| Rj-1(p)-Rj(q) |, 1≤p≤rj-1, 1≤q≤rj
Wherein, rj-1And rjRespectively indicate line spectral frequencies serial number set Rj-1And RjIn frequency serial number number;
(7-2-3) is in D (j)pqFirst minimum value of middle search, the minimum D (j) gotpq=D0It is worth corresponding frequency serial number
Rj-1(p0) and Rj(q0), if D0≤Gc, GcThresholding is clustered for line spectrum, then by Rj-1(p0) and Rj(q0) pairing, by Rj(q0) information
Addition includes (j-1, Rj-1(p0)) tracker of element, i.e., increase element (j, R in corresponding trackerj(q0));Otherwise it creates
One tracker, by (j, Rj(q0)) first element as new tracker;
(7-2-4) is in D (j)pqIn, exclude p=p0Or q=q0Corresponding D (j)pqValue, in remaining D (j)pqIn hold again
It goes (7-2-3), wherein p ≠ p0, q ≠ q0, until executing min (rj-1, rj) secondary, min () is that the minimum value in several numbers is taken to transport
It calculates;
(7-2-5) is if rj-1< rj, to line spectral frequencies serial number set RjIn still unpaired line spectral frequencies serial number build one by one
New tracker is found, and using current frame number and the frequency serial number as the first element for newly establishing tracker;
(7-3) j=j+1 returns to (7-2-2) and handles line spectral frequencies serial number set R frame by framej, until having handled all P frame numbers
According to line spectral frequencies serial number set.
Wherein, in step (8), time-frequency line spectrum data are extracted from tracker result, the specific steps are as follows: according to setting
Line spectrum curve frame number thresholding Gt, element number is filtered out from tracker more than threshold value GtTracker, screening obtain with
The data of track device are the time-frequency line spectrum data detected.
The utility model has the advantages that compared with prior art, technical solution of the present invention has following advantageous effects:
(1) present invention takes full advantage of time-frequency line spectrum energy in the regularity of distribution of time and frequency dimension, in lower letter
It makes an uproar than under the conditions of, can also obtain preferable time-frequency line-spectrum detection result;
(2) present invention is directly obtained from data by way of Sequential processing frame by frame without providing training sample early period
The judgment condition of line spectrum;
(3) it present invention employs the processing mode that frame processing and Sequential processing combine, simplifies at time-frequency line-spectrum detection
The complexity of reason can satisfy the real time signal processing demand of radar and sonar.
Detailed description of the invention
Fig. 1 is the flow diagram of the method for the present invention;
Fig. 2 is 1 frame data sequence power spectrogram of embodiment;
Fig. 3 is that 1 frame data sequence power of embodiment composes line-spectrum detection figure;
Fig. 4 is each frame power spectrum line spectrum serial number set figure of embodiment 1;
Fig. 5 is 1 time-frequency line-spectrum detection result figure of embodiment.
Specific embodiment
The present invention is described further with reference to the accompanying drawings and examples:
As shown in Figure 1, a kind of time-frequency line-spectrum detection method based on sequential cluster, includes the following steps:
(1) the sample data sequence s (n), n=0,1 ..., N-1 of signal to be detected are obtained, the n is sampled point sequence
Number, N is number of sampling points;
(2) sub-frame processing is carried out to sample data sequence s (n), obtains downlink data upon handover sequence xi(m), i=0,1 ..., P-
1, m=0,1 ..., M-1, i be frame number, P is frame number, and m is the sampling sequence number of each frame data sequence, and M is that frame data sequence is long
Degree;
(3) parameter initialization, and currently processed frame number j=0 is set;
(4) jth frame data sequence x is calculatedj(m) power spectrum Pj(k), k be frequency serial number, k=0,1,2 ..., K-1,
K is less thanPositive integer, floor () be downward rounding operation;
(5) to power spectrum Pj(k) line-spectrum detection is carried out, line spectrum respective frequencies serial number is extracted, obtains jth main feed line spectral frequency sequence
Number set Ri;
(6) it is incremented by j, repeats step (4)~(6), the power spectrum P until calculating whole P frame data sequencesj(k), it and obtains
To P line spectral frequencies serial number set Rj, j=0,1,2 ... P-1;
(7) to obtained P frame data line spectral frequencies serial number set RiSequential cluster is carried out frame by frame;
(8) time-frequency line spectrum data are extracted from cluster result.
Wherein, in step (1), signal sample data sequence s (n) to be detected is obtained with the following method: from sensor
The acquisition data of N number of sampled point are received as data sequence s (n) to be detected;Or N number of sample point data is extracted from memory
As data sequence s (n) to be detected.
Wherein, in step (2), sub-frame processing is carried out to sample data sequence s (n), taking frame length is M, and sliding step is
B, sliding step B are the differences of the adjacent first data sequence number for taking frame data in sample data sequence s (n) twice, then can obtain
To downlink data upon handover sequence xi(m), the xi(m)=s (iB+m), i are frame number, and m is each frame data sequential sampling point serial number, P
For data frame number, N, B, the relationship of P, M meet N=(P-1) B+M.
Wherein, in step (3), line spectrum judgement window length, line-spectrum detection amplitude threshold, line spectrum are clustered with the following method
Thresholding, line-spectrum detection frame number thresholding are initialized: setting initialization line spectrum adjudicates the long W of windowl, line-spectrum detection amplitude threshold Gl, line
Spectral clustering thresholding Gc, line-spectrum detection frame number thresholding Gt, wherein WlIt is greater than 8 odd number, G for valuelIt is greater than 3 real number for value,
GcIt is greater than 3 integer, G for valuetIt is greater than 3 integer for value.
Wherein, in step (4), with the following method according to the data sequence xj(m) power spectrum signal P is calculatedj(k):
X is done to the data sequencej(m) the frequency spectrum X of data sequence is calculated in discrete Fourier transformj(k) and power spectrum Pj(k),
Include the following steps:
(4-1) calculates xj(m) leaf transformation in M point discrete Fourier, and the preceding K point of transformation results is taken, obtain data sequence xj
(m) frequency spectrum
Xj(k), k=0,1,2 ..., K-1
Wherein, K is less thanPositive integer, floor () be downward rounding operation;
(4-2) is according to Xj(k) data sequence x is calculatedj(m) power spectrum
| | represent modulus value operation.
Wherein, in step (5), using the method for the local anomaly value detection to maximum point to power spectrum Pj(k) into
Row line-spectrum detection, comprises the following steps that
(5-1) is in (wl-1)/2≤k≤K-1-(wl- 1)/2 P is found out in rangej(k) the corresponding frequency of all maximum points
Rate serial number km, meet Pj(km) > Pj(km- 1) and Pj(km)≥Pj(km+1);
(5-2) is to each kmIf Pj(km)≥Gl·std({Pj(km-(wl- 1)/2), Pj(km-(wl-1)/2+
..., P 1)j(km+(wl- 1)/2) }), then determine the frequency point for power spectrum Pj(k) line spectrum frequency point, by the frequency sequence of the frequency point
The line spectral frequencies serial number set R of current data frame j frame number is recordedjIn, it obtains:
Rj={ km|Pj(km) > Pj(km-1)&Pj(km)≥Pj(km+1)&Pj(km)≥Gl·std({Pj(km-(wl-1)/
2), Pj(km-(wl- 1)/2+1) ..., Pj(km+(wl- 1) logical AND relationship/2) } }, is indicated, std () is to take data acquisition system equal
The operation of root, WlLong, the G for line spectrum judgement windowlFor line-spectrum detection amplitude threshold.
Wherein, in step (7), to obtained line spectral frequencies serial number set RjSequential cluster, specific steps are carried out frame by frame
It is as follows:
(7-1) remembers R0Element number r0, definition tracker is an ordered set T={ (frame number, frequency serial number) }, wound
Build r0A tracker Ta, a=1,2..., r0, the line spectrum point set R that is detected using the 0th frame data0In r0A frequency serial number km,
Construct element (0, km) respectively as r0The first element of a tracker;
(7-2) is by frame number frame by frame to line spectral frequencies serial number set R1, R2..., RP-1It is clustered, step includes:
(7-2-1) enables currently processed frame number j=1;
(7-2-2) calculates the line spectral frequencies serial number set R of the jth frame in line spectral frequencies serial number setjMiddle each element is with before
One main feed line spectral frequency serial number set Rj-1The absolute value of each element frequency point difference defines jth main feed line spectral frequency serial number set RjQ
A frequency serial number Rj(q) with -1 main feed line spectral frequency serial number set R of jthj-1P-th of frequency serial number Rj-1(p) distance are as follows:
D(j)pq=| Rj-1(p)-Rj(q) |, 1≤p≤rj-1, 1≤q≤rj
Wherein, rj-1And rjRespectively indicate line spectral frequencies serial number set Rj-1And RjIn frequency serial number number;
(7-2-3) is in D (j)pqFirst minimum value of middle search, the minimum D (j) gotpq=D0It is worth corresponding frequency serial number
Rj-1(p0) and Rj(q0), if D0≤Gc, GcThresholding is clustered for line spectrum, then by Rj-1(p0) and Rj(q0) pairing, by Rj(q0) information
Addition includes (j-1, Rj-1(p0)) tracker of element, i.e., increase element (j, R in corresponding trackerj(q0));Otherwise it creates
One tracker, by (j, Rj(q0)) first element as new tracker;
(7-2-4) is in D (j)pqIn, exclude p=p0Or q=q0Corresponding D (j)pqValue, in remaining D (j)pqIn hold again
It goes (7-2-3), wherein p ≠ p0, q ≠ q0, until executing min (rj-1, rj) secondary, min () is that the minimum value in several numbers is taken to transport
It calculates;
(7-2-5) is if rj-1< rj, to line spectral frequencies serial number set RjIn still unpaired line spectral frequencies serial number build one by one
New tracker is found, and using current frame number and the frequency serial number as the first element for newly establishing tracker;
(7-3) j=j+1 returns to (7-2-2) and handles line spectral frequencies serial number set R frame by framej, until having handled all P frame numbers
According to line spectral frequencies serial number set.
Wherein, in step (8), time-frequency line spectrum data are extracted from tracker result, the specific steps are as follows: according to setting
Line spectrum curve frame number thresholding Gt, element number is filtered out from tracker more than threshold value GtTracker, screening obtain with
The data of track device are the time-frequency line spectrum data detected.
Embodiment 1:
Emulation signal parameter is respectively set are as follows: set sample frequency FS as 5000Hz, ambient noise be mean value be 0, variance 1
White Gaussian noise;3 time-frequency line spectrums are set, and wherein time-frequency line spectrum 1 is simple signal, is always existed in time, frequency
100Hz, amplitude 0.15;Time-frequency line spectrum 2 is linear FM signal, is 0s-10s, initial frequency 200Hz there are the time, terminates frequency
Rate 210Hz, amplitude 0.2;Time-frequency line spectrum 3 is linear FM signal, is 15s-25s, initial frequency 200Hz there are the time, terminates
Frequency 210Hz, amplitude 0.2.The data sampling of 30s duration is carried out to signal, is obtained sample data sequence s (n).
According to (2) step, sub-frame processing is carried out to sample data sequence s (n), taking frame length M is 5000, and sliding step is
5000, then downlink data upon handover x can be obtainedi(m), i=0,1 ..., 29, m=0,1 ..., 4999.
According to (3) step, setting initialization line spectrum adjudicates the long W of windowl=31, line-spectrum detection amplitude threshold Gl=5, line spectrum is poly-
Class thresholding Gc=4, line-spectrum detection frame number thresholding Gt=4, and currently processed frame number j=0 is set.
According to (4) step, the power spectrum P of single frames sample data sequence is calculatedj(k), as shown in Figure 2.
According to (5) step, to power spectrum Pj(k) line spectral frequencies serial number is extracted, as shown in Figure 3.
According to (6) step, it is incremented by j, repeats function of (4) step to (6) step, until calculating whole P frame data sequences
Rate composes Pj(k), and P line spectral frequencies serial number set R is obtainedl, as shown in Figure 4.
According to (7) step, to obtained P frame data line spectral frequencies serial number set RjSequential cluster is carried out frame by frame.
According to (8) step, time-frequency line spectrum data are extracted from cluster result, as shown in figure 5, from testing result as it can be seen that
Three time-frequency line spectrum curves are accurately detected, identical as preset condition.
Claims (8)
1. a kind of time-frequency line-spectrum detection method based on sequential cluster, which comprises the steps of:
(1) the sample data sequence s (n), n=0,1 ..., N-1 of signal to be detected are obtained, the n is sampled point serial number, N
For number of sampling points;
(2) sub-frame processing is carried out to sample data sequence s (n), obtains downlink data upon handover sequence xi(m), i=0,1 ..., P-1, m=
0,1 ..., M-1, i are frame number, and P is frame number, and m is the sampling sequence number of each frame data sequence, and M is frame data sequence length;
(3) parameter initialization, and currently processed frame number j=0 is set;
(4) jth frame data sequence x is calculatedj(m) power spectrum Pj(k), k is frequency serial number, and k=0,1,2 ..., K-1, K are
It is less thanPositive integer, floor () be downward rounding operation;
(5) to power spectrum Pj(k) line-spectrum detection is carried out, line spectrum respective frequencies serial number is extracted, obtains jth main feed line spectral frequency serial number collection
Close Rj;
(6) it is incremented by j, repeats step (4)~(6), the power spectrum P until calculating whole P frame data sequencesj(k), and P are obtained
Line spectral frequencies serial number set Rj, j=0,1,2 ... P-1;
(7) to obtained P frame data line spectral frequencies serial number set RjSequential cluster is carried out frame by frame;
(8) time-frequency line spectrum data are extracted from cluster result.
2. a kind of time-frequency line-spectrum detection method based on sequential cluster according to claim 1, which is characterized in that in step
(1) in, signal sample data sequence s (n) to be detected is obtained with the following method: the acquisition of N number of sampled point is received from sensor
Data are as data sequence s (n) to be detected;Or N number of sample point data is extracted from memory as data sequence to be detected
It arranges s (n).
3. a kind of time-frequency line-spectrum detection method based on sequential cluster according to claim 1, which is characterized in that in step
(2) in, sub-frame processing is carried out to sample data sequence s (n), taking frame length is M, and sliding step B, sliding step B are adjacent two
The difference of the secondary first data sequence number that frame data are taken in sample data sequence s (n), then can be obtained downlink data upon handover sequence xi(m),
The xi(m)=s (iB+m), i are frame number, and m is each frame data sequential sampling point serial number, and P is data frame number, N, B, P, M
Relationship meet N=(P-1) B+M.
4. a kind of time-frequency line-spectrum detection method based on sequential cluster according to claim 1, which is characterized in that in step
(3) in, thresholding, line-spectrum detection frame number door are clustered to line spectrum judgement window length, line-spectrum detection amplitude threshold, line spectrum with the following method
Limit is initialized: the setting initialization line spectrum judgement long W of windowl, line-spectrum detection amplitude threshold Gl, line spectrum cluster thresholding Gc, line spectrum inspection
Survey frame number thresholding Gt, wherein WlIt is greater than 8 odd number, G for valuelIt is greater than 3 real number, G for valuecIt is greater than 3 integer for value,
GtIt is greater than 3 integer for value.
5. a kind of time-frequency line-spectrum detection method based on sequential cluster according to claim 1, which is characterized in that in step
(4) in, with the following method according to the data sequence xj(m) power spectrum signal P is calculatedj(k): x is done to the data sequencej
(m) the frequency spectrum X of data sequence is calculated in discrete Fourier transformj(k) and power spectrum Pj(k), include the following steps:
(4-1) calculates xj(m) leaf transformation in M point discrete Fourier, and the preceding K point of transformation results is taken, obtain data sequence xj(m)
Frequency spectrum
Xj(k), k=0,1,2 ..., K-1
Wherein, K is less thanPositive integer, floor () be downward rounding operation;
(4-2) is according to Xj(k) data sequence x is calculatedj(m) power spectrum
| | represent modulus value operation.
6. a kind of time-frequency line-spectrum detection method based on sequential cluster according to claim 5, which is characterized in that in step
(5) in, using the method for the local anomaly value detection to maximum point to power spectrum Pj(k) line-spectrum detection, including step are carried out
It is as follows:
(5-1) is in (wl-1)/2≤k≤K-1-(wl- 1)/2 P is found out in rangej(k) the corresponding frequency sequence of all maximum points
Number km, meet Pj(km) > Pj(km- 1) and Pj(km)≥Pj(km+1);
(5-2) is to each kmIf Pj(km)≥Gl·std({Pj(km-(wl- 1)/2), Pj(km-(wl- 1)/2+1) ..., Pj
(km+(wl- 1)/2) }), then determine the frequency point for power spectrum Pj(k) the frequency serial number of the frequency point is recorded and works as by line spectrum frequency point
The line spectral frequencies serial number set R of preceding data frame j framejIn, it obtains:
Rj={ km|Pj(km) > Pj(km-1)&Pj(km)≥Pj(km+1)&Pj(km)≥Gl·std({Pj(km-(wl- 1)/2), Pj
(km-(wl- 1)/2+1) ..., Pj(km+(wl- 1) logical AND relationship/2) } }, is indicated, std () is to take data acquisition system root mean square
Operation, WlLong, the G for line spectrum judgement windowlFor line-spectrum detection amplitude threshold.
7. a kind of time-frequency line-spectrum detection method based on sequential cluster according to claim 1, which is characterized in that in step
(7) in, to obtained line spectral frequencies serial number set RjSequential cluster is carried out frame by frame, the specific steps are as follows:
(7-1) remembers R0Element number r0, definition tracker is an ordered set T={ (frame number, frequency serial number) }, creates r0
A tracker Ta, a=1,2..., r0, the line spectrum point set R that is detected using the 0th frame data0In r0A frequency serial number km, building
Element (0, km) respectively as r0The first element of a tracker;
(7-2) is by frame number frame by frame to line spectral frequencies serial number set R1, R2..., RP-1It is clustered, step includes:
(7-2-1) enables currently processed frame number j=1;
(7-2-2) calculates the line spectral frequencies serial number set R of the jth frame in line spectral frequencies serial number setjMiddle each element and previous main feed line
Spectral frequency serial number set Rj-1The absolute value of each element frequency point difference defines jth main feed line spectral frequency serial number set RjQ-th of frequency
Serial number Rj(q) with -1 main feed line spectral frequency serial number set R of jthj-1P-th of frequency serial number Rj-1(p) distance are as follows:
D(j)pq=| Rj-1(p)-Rj(q) |, 1≤p≤rj-1, 1≤q≤rj
Wherein, rj-1And rjRespectively indicate line spectral frequencies serial number set Rj-1And RjIn frequency serial number number;
(7-2-3) is in D (j)pqFirst minimum value of middle search, the minimum D (j) gotpq=D0It is worth corresponding frequency serial number Rj-1
(p0) and Rj(q0), if D0≤Gc, GcThresholding is clustered for line spectrum, then by Rj-1(p0) and Rj(q0) pairing, by Rj(q0) information addition
It include (j-1, Rj-1(p0)) tracker of element, i.e., increase element (j, R in corresponding trackerj(q0));Otherwise one is created
Tracker, by (j, Rj(q0)) first element as new tracker;
(7-2-4) is in D (j)pqIn, exclude p=p0Or q=q0Corresponding D (j)pqValue, in remaining D (j)pqIn re-execute (7-
2-3), wherein p ≠ p0, q ≠ q0, until executing min (rj-1, rj) secondary, min () is the minimum operation taken in several numbers;
(7-2-5) is if rj-1< rj, to line spectral frequencies serial number set RjIn still unpaired line spectral frequencies serial number establish one by one newly
Tracker, and using current frame number and the frequency serial number as the first element for newly establishing tracker;
(7-3) j=j+1 returns to (7-2-2) and handles line spectral frequencies serial number set R frame by framej, until having handled all P frame datas
Line spectral frequencies serial number set.
8. a kind of time-frequency line-spectrum detection method based on sequential cluster according to claim 1, which is characterized in that in step
(8) in, time-frequency line spectrum data are extracted from tracker result, the specific steps are as follows: according to the frame number door of the line spectrum curve of setting
Limit Gt, element number is filtered out from tracker more than threshold value GtTracker, screening obtain tracker data be detect
The time-frequency line spectrum data arrived.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810811825.XA CN109061591B (en) | 2018-07-23 | 2018-07-23 | Time-frequency line spectrum detection method based on sequential clustering |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810811825.XA CN109061591B (en) | 2018-07-23 | 2018-07-23 | Time-frequency line spectrum detection method based on sequential clustering |
Publications (2)
Publication Number | Publication Date |
---|---|
CN109061591A true CN109061591A (en) | 2018-12-21 |
CN109061591B CN109061591B (en) | 2022-05-10 |
Family
ID=64835302
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201810811825.XA Active CN109061591B (en) | 2018-07-23 | 2018-07-23 | Time-frequency line spectrum detection method based on sequential clustering |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN109061591B (en) |
Cited By (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110376436A (en) * | 2019-06-27 | 2019-10-25 | 东南大学 | A kind of multiple dimensioned line-spectrum detection method for noise power spectra |
CN111736158A (en) * | 2020-08-25 | 2020-10-02 | 东南大学 | Target line spectrum feature identification method based on distributed multi-buoy matching |
CN111929666A (en) * | 2020-09-09 | 2020-11-13 | 东南大学 | Weak underwater sound target line spectrum autonomous extraction method based on sequential environment learning |
US20220343898A1 (en) * | 2019-09-24 | 2022-10-27 | Jingdong Technology Holding Co., Ltd. | Speech recognition method and apparatus, and computer-readable storage medium |
Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
FR2566918A1 (en) * | 1984-06-28 | 1986-01-03 | Labo Cent Telecommunicat | System for obtaining summarised information on the line spectrum provided by the spectrum analyser of a Doppler radar |
CN101938317A (en) * | 2010-09-10 | 2011-01-05 | 东南大学 | Line-spectrum detection method for noise power spectra |
CN102279399A (en) * | 2009-12-25 | 2011-12-14 | 中国科学院声学研究所 | Dim target frequency spectrum tracking method based on dynamic programming |
CN102353952A (en) * | 2011-06-03 | 2012-02-15 | 哈尔滨工程大学 | Line spectrum detection method by coherent accumulation of frequency domains |
CN105223576A (en) * | 2015-09-24 | 2016-01-06 | 哈尔滨工程大学 | A kind of line spectrum signal target automatic testing method based on single vector subsurface buoy |
-
2018
- 2018-07-23 CN CN201810811825.XA patent/CN109061591B/en active Active
Patent Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
FR2566918A1 (en) * | 1984-06-28 | 1986-01-03 | Labo Cent Telecommunicat | System for obtaining summarised information on the line spectrum provided by the spectrum analyser of a Doppler radar |
CN102279399A (en) * | 2009-12-25 | 2011-12-14 | 中国科学院声学研究所 | Dim target frequency spectrum tracking method based on dynamic programming |
CN101938317A (en) * | 2010-09-10 | 2011-01-05 | 东南大学 | Line-spectrum detection method for noise power spectra |
CN102353952A (en) * | 2011-06-03 | 2012-02-15 | 哈尔滨工程大学 | Line spectrum detection method by coherent accumulation of frequency domains |
CN105223576A (en) * | 2015-09-24 | 2016-01-06 | 哈尔滨工程大学 | A kind of line spectrum signal target automatic testing method based on single vector subsurface buoy |
Non-Patent Citations (1)
Title |
---|
张士忠 等: "基于对称相关函数法的线谱检测", 《科学技术与工程》 * |
Cited By (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110376436A (en) * | 2019-06-27 | 2019-10-25 | 东南大学 | A kind of multiple dimensioned line-spectrum detection method for noise power spectra |
CN110376436B (en) * | 2019-06-27 | 2021-06-01 | 东南大学 | Multi-scale noise power spectral line spectrum detection method |
US20220343898A1 (en) * | 2019-09-24 | 2022-10-27 | Jingdong Technology Holding Co., Ltd. | Speech recognition method and apparatus, and computer-readable storage medium |
CN111736158A (en) * | 2020-08-25 | 2020-10-02 | 东南大学 | Target line spectrum feature identification method based on distributed multi-buoy matching |
CN111929666A (en) * | 2020-09-09 | 2020-11-13 | 东南大学 | Weak underwater sound target line spectrum autonomous extraction method based on sequential environment learning |
Also Published As
Publication number | Publication date |
---|---|
CN109061591B (en) | 2022-05-10 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN109061591A (en) | A kind of time-frequency line-spectrum detection method based on sequential cluster | |
CN107576943B (en) | Adaptive Time and Frequency Synchronization compression method based on Rayleigh entropy | |
CN108600135A (en) | A kind of recognition methods of signal modulation mode | |
CN107393555B (en) | Detection system and detection method for abnormal sound signal with low signal-to-noise ratio | |
CN106772331B (en) | Target identification method and Target Identification Unit | |
CN109243470A (en) | Broiler chicken cough monitoring method based on Audiotechnica | |
CN103854662A (en) | Self-adaptation voice detection method based on multi-domain joint estimation | |
CN104217722A (en) | Dolphin whistle signal spectrum contour extraction method | |
CN111597991A (en) | Rehabilitation detection method based on channel state information and BilSTM-Attention | |
CN114469124B (en) | Method for identifying abnormal electrocardiosignals in movement process | |
CN113567969B (en) | Illegal sand dredger automatic monitoring method and system based on underwater acoustic signals | |
CN110120230A (en) | A kind of acoustic events detection method and device | |
CN108198558B (en) | Voice recognition method based on CSI data | |
CN110176250A (en) | It is a kind of based on the robust acoustics scene recognition method locally learnt | |
KR20210001575A (en) | Real-time cardiac rate detection apparatus in noisy environment and method thereof | |
CN112462355A (en) | Sea target intelligent detection method based on time-frequency three-feature extraction | |
CN112364296A (en) | P wave arrival time automatic picking method based on deep learning | |
CN114545342B (en) | Radar pulse signal parameter measurement method using multichannel reconnaissance receiver | |
Malik et al. | Automatic threshold optimization in nonlinear energy operator based spike detection | |
CN113780521B (en) | Radiation source individual identification method based on deep learning | |
CN111025100A (en) | Transformer ultrahigh frequency partial discharge signal mode identification method and device | |
CN113837122A (en) | Wi-Fi channel state information-based non-contact human body behavior identification method and system | |
RU2351005C1 (en) | Method of evaluating signal parameters and device to this end (versions) | |
CN110443276A (en) | Time series classification method based on depth convolutional network Yu the map analysis of gray scale recurrence | |
CN115500790A (en) | Method for predicting and analyzing epileptic seizure based on BI-LSTM 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 |