CN103884909A - Low-signal-to-noise ratio track frequency shift signal high accuracy detection method based on sparse decomposition - Google Patents

Abstract

The invention discloses a low-signal-to-noise-ratio track frequency shift signal high accuracy detection method based on sparse decomposition, comprising steps of constructing a first atomic library D1 according to nominal frequency parameters of the track frequency shift signals, using a matching and tracking algorithm to finish coarseness detection of the track frequency shift signal carrier frequency and low frequency in the atomic library, then constructing a second atomic library D2 by using the carrier frequency and low frequency in the coarseness detection as a reference according to the frequency deviation region and the detection accuracy requirement, screening out the optimal atoms from the D2 by utilizing the matching and tracking algorithm and realizing the fine grain detection of the ZPW-2000 track deviation signal carrier frequency and the low frequency detection. The invention is controllable in detection accuracy, good in instantaneity and can be achieved by software. The low-signal-to-noise ratio track frequency shift signal high accuracy detection method can perform high accuracy detection on the ZPW-2000 mode and the UM-71 mode track frequency shift signals, can be used for signal detection and decoding of the railway vehicle-mounted locomotive, and testing and quality inspection of the track frequency shift transceiver.

Description

Low signal-to-noise ratio track frequency-shifting signal high-precision detecting method based on Its Sparse Decomposition

Technical field

The present invention relates to a kind of low signal-to-noise ratio ZPW-2000 track frequency-shifting signal detection method, a specifically high-precision detecting method for the low signal-to-noise ratio ZPW-2000 track frequency-shifting signal based on Its Sparse Decomposition, can be used for the occasion such as test and quality inspection of detection, decoding and the track shift frequency transceiver of general iron under strong noise environment, the special and high speed railway vehicle mounted cab signal of visitor.

Background technology

Frequency shift modulated track circuit is the fundamental equipments of railway signal, it connects train operation and signal demonstration etc., be used for supervising the situation that takies of circuit, send frequency-shift signaling by transmitter along track, transmit front track seizure condition information and train operation control information to receiver and locomotive, ensure train operating safety and improve driving efficiency and transmit running information by frequency shift modulated track circuit to train.

At present, ZPW-2000 type track frequency-shifting signal is widely adopted on China railways main line, Line for Passenger Transportation and high-speed railway, and is established as the unified standard of railway frequency-shift automatic blocking system from now on.ZPW-2000 is the continuous frequency shift keying signal of a kind of phase place (frequency-shift keying, be called for short FSK), there is identical modulation system and modulation parameter with French UM-71 track frequency-shifting signal, it is the mode of proportion modulation, low-frequency information is transferred to compared with on high carrier frequency, to form, amplitude is constant, frequency is made periodically variable FM signal with the amplitude of low frequency signal, and mathematical notation is:

s(t)＝Acos(2πf _ct+h(t)) (1)

$h\left(t\right)=\left\{\begin{array}{c}2\mathrm{\&pi;\&Delta;ft},\frac{-1}{4f_d}<t<\frac{1}{4{\mathrm{<figure-callout\; id="2"\; label="f\; d"\; filenames="HDA0000467031060000021.png"\; state="\{\{state\}\}">f</figure-callout>}}_d}\\ 2\mathrm{\&pi;\&Delta;f}(\frac{1}{2f_d}-t),\frac{1}{4f_d}<t<\frac{3}{4f_d}\end{array}\right.---\left(2\right)$

Wherein, A, f _c, f _dbe respectively amplitude, carrier frequency and the low frequency of track frequency-shifting signal s (t), Δ f is frequency deviation.According to railway technology standard, frequency deviation Δ f=11Hz, eight nominal carrier frequency f _c∈ { 1698.7,1701.4,1998.7,2001.4,2298.7,2301.4,2598.7,2601.4}Hz, 18 nominal low frequency f _dstart to be uniformly-spaced incremented to 29Hz with 1.1Hz from 10.3Hz.

The effect major embodiment that ZPW-2000 track frequency-shifting signal detects is both ways: on the one hand, the vehicle-mounted cab signaling equipment of train in transit need to carry out online detection and decoding to the track frequency-shifting signal receiving, and extracts the carrier frequency f of frequency-shift signaling _cwith low frequency f _dparameter, takies situation and target range information to obtain front track; On the other hand, track frequency-shifting signal equipment upper track needs to carry out comprehensive quality inspection before using, and wherein crucial one is that frequency-shift signaling carrier frequency and low-frequency parameter to transmitter and receiver carries out high precision detection, so that the hidden danger of quality of discovering device in time.Therefore the quality of frequency-shift signaling detection algorithm, is directly connected to the traffic safety of train, and particularly, in high-speed line and low signal-to-noise ratio situation, frequency-shift signaling detects and seems particularly important accurately and reliably.

At present, detection method for track frequency-shifting signal mainly contains: document " the track circuit UM71 signal demodulation based on wavelet transformation realizes " (Zhang Lihong, Zhao Mingbo, measuring technology journal, 21 (6): 36-40,2007) first track frequency-shifting signal is carried out to wavelet transformation, then utilize wavelet ridge to obtain the carrier frequency of frequency-shift signaling, in recycling, the duration of lower side frequency calculates the low frequency of frequency-shift signaling.Document " applied research of EMD algorithm in Frequency-shift Signal Demodulation Using " (Liu Yanhong, Ma Ruijun, Wei Xueye, electronic surveying and instrument journal, 20 (5): 34-38,2006) after first frequency-shift signaling to be decoded being decomposed with EMD, then decomposed signal is carried out to Fourier's frequency domain spectrum peak search acquisition carrier frequency and low frequency.Document " application and the DSP thereof of ZFFT algorithm in railway frequency-shift signal is analyzed realizes " (Wu Zhongqi, Yang Shiwu, Source of Railway Communication and Signalling, 44 (07): 32-34,2008) first frequency-shift signaling is carried out to Fast Fourier Transform (FFT) (FFT), then in frequency domain, the maximum spectrum of search peak obtains carrier frequency, finally near frequency range carrier frequency is carried out to frequency spectrum amplification (Zoom-FFT), again calculates the low frequency of frequency-shift signaling by spectrum peak search.Document " design of railway frequency-shift signal detection system and realization " (Yang Fan, hoopoe China, Liu Ze, electronic surveying and instrument journal, 24 (05): 500-505,2010) track frequency-shifting signal is owed after sampling, utilized FFT conversion to obtain carrier frequency at the maximum spectrum of frequency domain search peak, and adopt energy barycenter method alignment technique to improve the accuracy of detection of carrier frequency, then utilize Zoom-FFT technique computes low frequency.Chinese invention patent application discloses a kind of ZPW-2000 track frequency-shifting signal interpretation method based on Duffing oscillator for No. 201110108599.7.The method is utilized the susceptibility of Duffing oscillator to homogenous frequency signal and the immunity feature to noise, low signal-to-noise ratio ZPW-2000 track frequency-shifting signal is successively input in 4 layered transducer elements and 18 layered transducer elements, changes and realize carrier frequency and low frequency decoding according to the phasor of oscillator.

The above-mentioned detection algorithm based on small echo or frequency domain spectrum peak search, can realize the carrier frequency of high s/n ratio track frequency-shifting signal and the detection of low frequency and decoding, but will lose efficacy under low signal-to-noise ratio.In addition, low frequency aberration that the detection algorithm based on small echo continues duration calculation according to upper lower side frequency is large, precision is low, and carrier frequency low frequency accuracy of detection based on frequency domain spectrum peak search is subject to the theory restriction of sampling rate and sampling duration.For low signal-to-noise ratio track frequency-shifting signal, though above-mentioned method based on Duffing oscillator can realize the decoding of low signal-to-noise ratio carrier frequency low frequency, its decoding principle has determined that the method does not possess the high precision detectability of carrier frequency low-frequency parameter.In addition, the parallel organization of Duffing layered transducer elements is convenient to hardware and is realized, but the decoding speed of software while realizing will be a greater impact.

Summary of the invention

The object of this invention is to provide a kind of low signal-to-noise ratio ZPW-2000 track frequency-shifting signal high-precision detecting method based on Its Sparse Decomposition, the method can be carried out high precision detection to the carrier frequency of low signal-to-noise ratio ZPW-2000 track frequency-shifting signal and low frequency, and the detection speed of software while realizing meets application request completely.

The present invention realizes its goal of the invention, and the technical scheme adopting is:

Low signal-to-noise ratio track frequency-shifting signal high-precision detecting method based on Its Sparse Decomposition, the steps include:

Carrier frequency (the f of A to track frequency-shifting signal _c), low frequency (f _d) carry out coarseness detection

(A.1) be first configured to the over-complete dictionary of atoms D of track frequency-shifting signal Its Sparse Decomposition ₁, D ₁in atom g _i,j(t) select the cycle of following characteristics to be

cosine signal:

$g_{i,j}\left(t\right)=\left\{\begin{array}{c}\cos \left(2\mathrm{\&pi;}(f_{c,i}+11)t\right),-\frac{1}{4f_{d,j}}<t\&le;\frac{1}{4f_{d,j}}\\ \cos (2\mathrm{\&pi;}(f_{c,i}-11)t+\frac{11\mathrm{\&pi;}}{f_{d,j}}),\frac{1}{4f_{d,j}}<t\&le;\frac{3}{4f_{d,j}}\end{array}\right.$

Wherein f _c,i(i=0,1 ..., 7) and f _d,j(j=0,1 ..., 17) be called atom g _i,j(t) carrier frequency and low frequency, and 8 nominal carrier frequencies and 18 nominal low frequencies, the i.e. f of difference value ZPW-2000 track frequency-shifting signal _c,i∈ { 1698.7,1701.4,1998.7,2001.4,2298.7,2301.4,2598.7,2601.4}Hz, f _d,j∈ 10.3+1.1 × j, and j=0,1,2 ..., 17}Hz.

(A.2) establishing signals and associated noises to be detected is s _n(t)=s (t)+n (t), s (t) is ZPW-2000 track frequency-shifting signal, n (t) is noise.Utilize match tracing (MP) algorithm, at former word bank D ₁middle search and s _n(t) the atom g of inner product maximum _i,j(t), be designated as best atom

meet: $|<s\left(t\right),g_{\hat{i},\hat{j}}\left(t\right)>|=\underset{g_{i,j}\left(t\right)\&Element;{\mathrm{<figure-callout\; id="1"\; label="D"\; filenames="HDA0000467031060000021.png"\; state="\{\{state\}\}">D</figure-callout>}}_1}{\sup }|<s\left(t\right),g_{i,j}\left(t\right)>|,$ Wherein <, > is inner product operation, | .| is for taking absolute value, and sup is for getting maxima operation symbol.The coarseness carrier frequency detected value of track frequency-shifting signal s (t)

with low frequency detected value

be respectively best atom

carrier frequency and low frequency value,

Carrier frequency (the f of B to track frequency-shifting signal _c), low frequency (f _d) carry out fine granularity detection

(B.1) according to the coarseness carrier frequency of step (A.2) and low frequency

detected value, selects second the over-complete dictionary of atoms D of atomic structure structure shown in step (A.1) ₂, D ₂in each atom g _i,j(t) carrier frequency

low frequency

wherein d ₁, d ₂determined respectively Δ f by the nominal carrier frequency of track frequency-shifting signal and the frequency departure scope of nominal low frequency _c, Δ f _dbe respectively the carrier frequency of s (t) and the step-size in search that low frequency detects, set by required accuracy of detection.

(B.2) similar step (A.2), at former word bank D ₂in utilize MP algorithm to try to achieve and s _n(t) the best atom of inner product maximum

the carrier frequency of track frequency-shifting signal s (t) and the fine granularity detected value of low frequency are D ₂middle best atom

carrier frequency and low frequency value, so far the high precision that, has realized the frequency parameter of low signal-to-noise ratio lower railway frequency-shift signaling s (t) detects.

Compared with prior art, the invention has the beneficial effects as follows:

(1) the present invention adopts with the isostructural cosine signal of ZPW-2000 track frequency-shifting signal and builds over-complete dictionary of atoms as atom, then on these former word banks, carry out the nonopiate Its Sparse Decomposition of track frequency-shifting signal, anti-noise ability than tradition based on the Orthogonal Decomposition such as small echo and FFT method is strong, overcome the detection Problem of Failure of traditional detection method in low signal-to-noise ratio situation, also having overcome interpretation method based on Duffing oscillator can only the non-detectable defect of decoding;

(2) compared with prior art, the high precision that the present invention can realize ZPW-2000 track frequency-shifting signal carrier frequency and low frequency under low signal-to-noise ratio detects, and sensing range and accuracy of detection controlled, only need former word bank D be set simply according to actual testing requirement ₂parameter (Δ f _c, Δ f _d, d ₁, d ₂);

(3) conventional Its Sparse Decomposition need to build huge former word bank, and the atom that search is mated with signal most one by one in this former word bank, has caused the operand that Its Sparse Decomposition is huge.The present invention detects thought by the thickness two-stage and is applied in Its Sparse Decomposition process, is guaranteeing, under the constant prerequisite of carrier frequency and low frequency accuracy of detection, significantly to have reduced the computation complexity of algorithm, requirement of real time;

(4) the present invention, in the time of the former word bank of structure, according to the technical manual of ZPW-2000 track frequency-shifting signal and frequency detection accuracy requirement, determines carrier frequency, low-frequency range and the step-size in search of atom, has greatly reduced D ₁and D ₂former word bank scale, does the calculated amount of the best atom match search that has both significantly reduced algorithm like this, has also significantly reduced the space requirement of atom library storage, is convenient to software and hardware and realizes.

Below in conjunction with drawings and Examples, the present invention is described in further detail

Accompanying drawing explanation

Fig. 1 is the inventive method process flow diagram.

Fig. 2 is the ZPW-2000 track frequency-shifting signal oscillogram in the embodiment of the present invention 1.Wherein: 2 (a) component is the oscillogram of the emulation frequency-shift signaling s (t) of formula (1) generation, its amplitude A=1, carrier frequency f _c=1701Hz, low frequency f _d=10.4Hz, sample frequency f _s=6000Hz; 2 (b) component is the signals and associated noises s after s (t) stack white Gaussian noise _n1(t) oscillogram, signal to noise ratio snr=-10dB; 2 (c) component is the signals and associated noises s after s (t) stack Gauss coloured noise _n2(t) oscillogram, SNR=-10dB.

Fig. 3 is the carrier frequency testing result contrast of the present invention and the track frequency-shifting signal of Zoom-FFT algorithm under different signal to noise ratio (S/N ratio)s.The wherein amplitude A=1 of track frequency-shifting signal, low frequency f _d=10.3Hz, carrier frequency f _c=1698.7Hz, sampling f _s=6000Hz, data length N=8192.

Fig. 4 is the low frequency testing result contrast of the present invention and the track frequency-shifting signal of Zoom-FFT algorithm under different signal to noise ratio (S/N ratio)s.The wherein amplitude A=1 of track frequency-shifting signal, sampling f _s=6000Hz, data length N=8192, carrier frequency f _c=2001.4Hz, low frequency f _d=18Hz.

Fig. 5 is the low frequency testing result contrast of the present invention and the track frequency-shifting signal of Zoom-FFT algorithm under different signal to noise ratio (S/N ratio)s.The wherein carrier frequency f of track frequency-shifting signal _c=2301.4Hz, low frequency f _d=22.4Hz, other parameter is identical with Fig. 4.

Embodiment

With reference to Fig. 1, the specific embodiment of the present invention is that a kind of low signal-to-noise ratio ZPW-2000 track frequency-shifting signal high-precision detecting method based on Its Sparse Decomposition, the steps include:

Carrier frequency (the f of A to track frequency-shifting signal _c), low frequency (f _d) carry out coarseness detection

(A.1) be first configured to the over-complete dictionary of atoms D of track frequency-shifting signal Its Sparse Decomposition ₁, D ₁in atom g _i,j(t) select the cycle of following characteristics to be

cosine signal:

$g_{i,j}\left(t\right)=\left\{\begin{array}{c}\cos \left(2\mathrm{\&pi;}(f_{c,i}+11)t\right),-\frac{1}{4f_{d,j}}<t\&le;\frac{1}{4f_{d,j}}\\ \cos (2\mathrm{\&pi;}(f_{c,i}-11)t+\frac{11\mathrm{\&pi;}}{f_{d,j}}),\frac{1}{4f_{d,j}}<t\&le;\frac{3}{4f_{d,j}}\end{array}\right.$

Wherein f _c,i(i=0,1 ..., 7) and f _d,j(j=0,1 ..., 17) be called atom g _i,j(t) carrier frequency and low frequency, and 8 nominal carrier frequencies and 18 nominal low frequencies, the i.e. f of difference value ZPW-2000 track frequency-shifting signal _c,i∈ { 1698.7,1701.4,1998.7,2001.4,2298.7,2301.4,2598.7,2601.4}Hz, f _d,j∈ 10.3+1.1 × j, and j=0,1,2 ..., 17}Hz.

(A.2) establishing signals and associated noises to be detected is s _n(t)=s (t)+n (t), in the present embodiment, the carrier frequency f of track frequency-shifting signal s (t) _c=1701Hz, low frequency f _d=10.4Hz, n (t) is white Gaussian noise, signal to noise ratio snr=-10dB, s _n(t) oscillogram is as shown in Fig. 2 (b).Utilize match tracing (MP) algorithm, at former word bank D ₁middle search and s _n(t) the atom g of inner product maximum _i,j(t), be designated as best atom

worked as

time, for best atom.The coarseness carrier frequency detected value of track frequency-shifting signal s (t)

with low frequency detected value

be respectively best atom g _1,0(t) carrier frequency and low frequency value, ${\hat{f}}_c=f_{c,1}=1701.4\mathrm{Hz},$ ${\hat{f}}_d=f_{d,0}=10.3\mathrm{Hz}.$

Carrier frequency (the f of B to track frequency-shifting signal _c), low frequency (f _d) carry out fine granularity detection

(B.1) according to the coarseness carrier frequency detected value of step (A.2) with low frequency detected value

select second the over-complete dictionary of atoms D of atomic structure structure shown in step (A.1) ₂.D in the present embodiment ₁=1Hz, d ₂=0.5Hz, Δ f _c=0.1Hz, Δ f _d=0.01Hz, D ₂in each atom g _i,j(t) carrier frequency

low frequency $f_{d,j}\&Element;[{\hat{f}}_d-d_2:{\mathrm{\&Delta;f}}_d:{\hat{f}}_d+d_2]=[9.8:0.01:10.8]\mathrm{Hz}.$

(B.2) similar step (A.2), at former word bank D ₂in utilize MP algorithm to try to achieve and s _n(t) the best atom of inner product maximum

worked as

time, for best atom.The carrier frequency of track frequency-shifting signal s (t) and the final detected value of low frequency are D ₂middle best atom g _6,60(t) carrier frequency and low frequency value,

so far the high precision that, has realized the frequency parameter of SNR=-10dB lower railway frequency-shift signaling s (t) detects.Effect of the present invention can be verified by following emulation experiment

Embodiment 1

In the present embodiment, emulation produces ZPW-2000 track frequency-shifting signal s (t), the amplitude A=1 of s (t), carrier frequency f _c=1701Hz, low frequency f _d=10.4Hz, sample frequency f _s=6000, data length N=8192.The partial waveform of s (t) is as shown in Fig. 2 (a) component, 2 (b) component is the partial waveform figure after s (t) stack-10dB white Gaussian noise, and 2 (c) component is the partial waveform figure after s (t) stack-10dB Gauss coloured noise.The white noise that coloured noise in this experiment is is 1 by variance obtains by a fourth-order band-pass wave filter, and the transport function of this wave filter is:

$H\left(z\right)=\frac{0.0201*(1-{2z}^{-1}+z^4)}{1-{1.637z}^{-1}+{2.237z}^{-2}-{1.3072z}^{-3}}$

Wherein, the normalization bound cutoff frequency of this wave filter is respectively 0.2Hz, 0.15Hz.

In certain deviation range of nominal carrier frequency and nominal low frequency, carrier frequency and the low frequency of randomly changing s (t), obtain the noisy frequency-shift signaling s of a series of non-nominal frequency parameters _n(t), then adopt embodiment of the present invention method to carry out frequency parameter detection to these signals, testing result is as shown in table 1, table 2.Wherein table 1 is that non-nominal carrier frequency detected value and carrier frequency detect error, and table 2 is that non-nominal low frequency detected value and low frequency detect error, and signal to noise ratio (S/N ratio) is the white Gaussian noise environment of SNR=-10dB, Monte Carlo simulation 500 times.

Non-nominal carrier frequency testing result (white Gaussian noise SNR=-10dB) under the different low frequencies of table 1

Non-nominal low frequency testing result (white Gaussian noise SNR=-10dB) under the different carrier frequency of table 2

Table 1 illustrates, under the white Gaussian noise environment of SNR=-10dB, non-nominal carrier frequency all can correctly detect and carrier frequency to detect error be 0, be much better than that carrier frequency detects 10 ^-1the skill rule requirement of magnitude.Table 2 illustrates, under the white Gaussian noise environment of SNR=-10dB, non-nominal low frequency all can correctly detect and low frequency detects error all in 0.003Hz, be much better than that low frequency detects 10 ^-2the skill rule requirement of magnitude.Consolidated statement 1 and table 2 can find out, carrier frequency of the present invention and low frequency accuracy of detection totally exceed an order of magnitude than the frequency detection accuracy index of railway skill rule regulation.Change noise into Gauss coloured noise, other simulated conditions is constant, also has same detection effect.Experiment 1 has shown that the present invention can carry out high precision detection to the ZPW-2000 track frequency-shifting signal of frequency parameter nominal value generation deviation.

Embodiment 2

ZPW-2000 track frequency-shifting signal frequency parameter is got respectively to 8 kinds of nominal carrier frequencies and 18 kinds of nominal low frequencies, the raw 8*18=144 kind of common property track frequency-shifting signal, other experiment condition, method of operating are identical with experiment 1.Under SNR=-10dB Gauss coloured noise, carrier frequency and low frequency to these 144 kinds of track frequency-shifting signals detect, and have all obtained satisfied result.Table 3 has provided the partial detection that nominal carrier frequency detects, and table 4 has provided the partial detection that nominal low frequency detects.

Nominal carrier frequency testing result (Gauss's coloured noise SNR=-10dB) under table 3 nominal low frequency

Nominal low frequency testing result (Gauss's coloured noise SNR=-10dB) under table 4 nominal carrier frequency

Table 3 and table 4 illustrate, under Gauss's coloured noise environment of SNR=-10dB, the detection effect of nominal carrier frequency of the present invention and nominal low frequency is consistent with experiment 1.In addition, change the Gauss's coloured noise in experiment 2 into white Gaussian noise, when other simulated conditions is constant, also obtain similar testing result.Result shows, nominal carrier frequency and the low frequency of the present invention to low signal-to-noise ratio ZPW-2000 track frequency-shifting signal all can carry out high precision detection.

Performance simulation result is comprehensively analyzed

For further illustrating the detection performance of the present invention to ZPW-2000 track frequency-shifting signal, lowest detection signal to noise ratio (S/N ratio) of the present invention and implementation efficiency are discussed respectively below.

1. lowest detection signal to noise ratio (S/N ratio)

Lowest detection signal to noise ratio (S/N ratio) shows that detection algorithm meets the lowest detection ability that can reach when accuracy of detection requires under low signal-to-noise ratio environment, has reflected detectivity and the noiseproof feature of detection algorithm to weak signal.The Zoom-FFT(of the inventive method and widespread use being called for short to ZFFT below) high-precision detecting method contrasts, accuracy of detection and the anti-noise advantage of outstanding the inventive method.

Emulation experiment parameter: the amplitude A=1 of track frequency-shifting signal, sampling f _s=6000Hz, data length N=8192; It is 8192 that the Fourier transform of Zoom-FFT algorithm is counted, and enlargement factor is 20.

(1) carrier frequency f _clowest detection signal to noise ratio (S/N ratio)

Fig. 3 is the carrier frequency testing result of the present invention and the track frequency-shifting signal of ZFFT algorithm under different signal to noise ratio (S/N ratio)s, wherein the carrier frequency f of track frequency-shifting signal _c=1698.7Hz, low frequency f _d=10.3Hz.Fig. 3 illustrates: in the time of SNR >=-25dB, the present invention can accurately detect carrier frequency and detect error is 0, and ZFFT only has and in the time of SNR >=-5dB, just can reach accurate detection.If detect according to carrier frequency the technical indicator that error is not more than 0.2Hz, as the scope that dotted line indicated in Fig. 3, the present invention meets the demands in the time of SNR >=-25dB, and ZFFT could meet testing requirement at SNR >-15dB.In the time that the carrier frequency of track frequency-shifting signal and low frequency are got other value, also there is similar results.Therefore, the carrier frequency lowest detection signal to noise ratio (S/N ratio) of the inventive method is at least than the low 10dB of ZFFT algorithm, and under equal accuracy of detection condition, the anti-noise ability that the inventive method detects track frequency-shifting signal carrier frequency is than the good 20dB of ZFFT left and right.

(2) low frequency f _dlowest detection signal to noise ratio (S/N ratio)

Fig. 4 and Fig. 5 are the inventive method and the low frequency testing result of ZFFT algorithm to the track frequency-shifting signal under different signal to noise ratio (S/N ratio)s.Wherein, the carrier frequency f of Fig. 4 _c=2001.4Hz, low frequency f _d=18Hz; The carrier frequency f of Fig. 5 _c=2301.4Hz, low frequency f _d=22.4Hz.

Fig. 4 illustrates: for ZPW-2000 track frequency-shifting signal, in the time of signal to noise ratio snr >=-15dB, the present invention can accurately detect low frequency and detect error is 0, and ZFFT just can reach accurate detection in the time of SNR >=-5dB.If detect according to low frequency the technical indicator that error is not more than 0.02Hz, scope as the dotted line in Figure 4 indicated, the present invention meets testing requirement in the time of SNR >=-22dB, and ZFFT needs SNR >=-13dB could meet testing requirement.

Get other value (as shown in Figure 5) for carrier frequency and low frequency, experimental result and Fig. 4 are similar.Therefore, low frequency lowest detection signal to noise ratio (S/N ratio) of the present invention and anti-noise ability will good 10dB left and right than ZFFT.

Synthesizing map 3～Fig. 5 can obtain: allow and detect in error on railway skill rule, carrier frequency lowest detection signal to noise ratio (S/N ratio) of the present invention is SNR >=-25dB, and low frequency lowest detection signal to noise ratio (S/N ratio) is SNR >=-22dB.Under the same conditions, detection performance of the present invention is far superior to the detection performance of ZFFT algorithm, has further embodied superiority and the validity of the inventive method.

2, implementation efficiency

The present invention is a kind of low signal-to-noise ratio ZPW-2000 track frequency-shifting signal detection method based on Its Sparse Decomposition, although the operand of conventional Its Sparse Decomposition is large, but the present invention adopts thickness two-stage atoms searching scheme and sets atomic parameter according to ZPW-2000 actual frequency parameter area, the operand that has greatly reduced former word bank scale and atom match tracing, makes algorithm reach requirement of real-time.In the Matlab of above-mentioned experiment emulation, the high precision detection total time of track frequency-shifting signal carrier frequency and low frequency only needs 0.33 second, also will reduce if detect elapsed time while adopting C language to realize, meets the requirement of real-time that track frequency-shifting signal detects completely.

Claims (3)

1. the low signal-to-noise ratio track frequency-shifting signal high-precision detecting method based on Its Sparse Decomposition, the steps include:

Carrier frequency (the f of A to track frequency-shifting signal _c), low frequency (f _d) carry out coarseness detection

(A.1) be first configured to the over-complete dictionary of atoms D of track frequency-shifting signal Its Sparse Decomposition ₁, D ₁in atom g _i,j(t) select the cycle of following characteristics to be

cosine signal:

$g_{i,j}\left(t\right)=\left\{\begin{array}{c}\cos \left(2\mathrm{\&pi;}(f_{c,i}+11)t\right),-\frac{1}{4f_{d,j}}<t\&le;\frac{1}{4f_{d,j}}\\ \cos (2\mathrm{\&pi;}(f_{c,i}-11)t+\frac{11\mathrm{\&pi;}}{f_{d,j}}),\frac{1}{4f_{d,j}}<t\&le;\frac{3}{4f_{d,j}}\end{array}\right.$

Wherein f _c,i(i=0,1 ..., 7) and f _d,j(j=0,1 ..., 17) be called atom g _i,j(t) carrier frequency and low frequency, and 8 nominal carrier frequencies and 18 nominal low frequencies, the i.e. f of difference value ZPW-2000 track frequency-shifting signal _c,i∈ { 1698.7,1701.4,1998.7,2001.4,2298.7,2301.4,2598.7,2601.4}Hz, f _d,j∈ 10.3+1.1 × j, and j=0,1,2 ..., 17}Hz;

(A.2) establishing signals and associated noises to be detected is s _n(t)=s (t)+n (t), s (t) is ZPW-2000 track frequency-shifting signal, n (t) is noise; Utilize match tracing MP algorithm, at former word bank D ₁middle search and s _n(t) the atom g of inner product maximum _i,j(t), be designated as best atom

meet: $|<s\left(t\right),g_{\hat{i},\hat{j}}\left(t\right)>|=\underset{g_{i,j}\left(t\right)\&Element;D_1}{\sup }|<s\left(t\right),g_{i,j}\left(t\right)>|,$ Wherein <, > is inner product operation, | .| is for taking absolute value, and sup is for getting maxima operation symbol; The coarseness carrier frequency detected value of track frequency-shifting signal s (t)

with low frequency detected value

be respectively best atom

carrier frequency and low frequency value,

Carrier frequency (the f of B to track frequency-shifting signal _c), low frequency (f _d) carry out fine granularity detection

(B.1) according to the coarseness carrier frequency of step (A.2)

and low frequency

detected value, selects second the over-complete dictionary of atoms D of atomic structure structure shown in step (A.1) ₂, D ₂in each atom g _i,j(t) carrier frequency

low frequency

wherein d ₁, d ₂determined respectively Δ f by the nominal carrier frequency of track frequency-shifting signal and the frequency departure scope of nominal low frequency _c, Δ f _dbe respectively the carrier frequency of s (t) and the step-size in search that low frequency detects, set by required accuracy of detection;

(B.2) similar step (A.2), at former word bank D ₂in utilize MP algorithm to try to achieve and s _n(t) the best atom of inner product maximum

the carrier frequency of track frequency-shifting signal s (t) and the fine granularity detected value of low frequency are best atom in D2

carrier frequency and low frequency value, ${\hat{f}}_c=f_{c,\hat{i}},$ ${\hat{f}}_d=f_{d,\hat{j}};$

So far the high precision that, realizes the frequency parameter of low signal-to-noise ratio lower railway frequency-shift signaling s (t) detects.

2. the low signal-to-noise ratio track frequency-shifting signal high-precision detecting method based on Its Sparse Decomposition according to claim 1, is characterized in that: the carrier search step delta f in described step (B.1) _c=0.1Hz, low frequency step-size in search Δ f _d=0.01Hz.

3. the low signal-to-noise ratio track frequency-shifting signal high-precision detecting method based on Its Sparse Decomposition according to claim 1, is characterized in that: the carrier deviation d in described step (B.1) ₁=1Hz, low frequency deviation d ₂=0.5Hz.

Images