CN106814601A

CN106814601A - A kind of Noise Filter method of signal specific

Info

Publication number: CN106814601A
Application number: CN201710182251.XA
Authority: CN
Inventors: 傅德锦; 刘胜; 方军涛; 李俊; 王毅
Original assignee: Xiamen Nine Magnificent Communication Equipment Factory
Current assignee: Xiamen Nine Magnificent Communication Equipment Factory
Priority date: 2017-03-24
Filing date: 2017-03-24
Publication date: 2017-06-09
Anticipated expiration: 2037-03-24
Also published as: CN106814601B

Abstract

A kind of Noise Filter method of signal specific, according to observation A_m, it is assumed that observation A_mIt is exactly actual value Z_m, the guess value T with certain precision can be released by mapping relationship f (m n)_m‑n=A_m+ f (m n), because observed quantity A_mRandom noise has been superimposed, guess value T has been often led to_m‑nLarger deviation is there is, or because speculates the increasing deviation of the accumulation that points increase.The general principle of filtering algorithm of the invention assumes observation A_mEqual to Z_m, then according to observation A_mThe a series of guess value for obtaining should be equal to random noise r with observed difference score value_m, its amplitude 80% falls between [R R], according to difference value whether beyond [R R] to T_mIncrement or decreasing amount adjustment are carried out, T is increased with adjustment number of times_mZ can be become closer to_m。

Description

A kind of Noise Filter method of signal specific

Technical field

The present invention relates to field of signal processing, the Noise Filter method of particularly a kind of signal specific.

Background technology

In time service field, either Big Dipper time service, GPS time services, GLONASS time services, PTP time services, long-wave time service or short Ripple time service, temporal frequency transmission is carried out using wireless or wire message way, because receiver clock is unstable, system clock correction, pass The time service information of the factors cause receivers such as the delay jitter of defeated channel reduction often has certain randomized jitter.Passive fixed Often there is larger ripple because signal strength changes in position field, the arrival time of the time pulse in time domain for receiving It is dynamic.

Traditional linear filter method can effectively filter random noise, but some gradual changes simultaneously useful component Together it is passivated, becomes insensitive, the gradual change of time service information, such as Doppler in passive location such as caused by the clock correction of time dissemination system The gradual change of pulse arrival time caused by effect.Traditional linear filter method such as Kalman filtering has ratio just for linear signal Preferable effect.

The content of the invention

It is a primary object of the present invention to overcome drawbacks described above of the prior art, propose one kind according to a series of observation Value obtains the Noise Filter method of the signal specific of more accurate guess value, can effectively weaken random noise influence.

The present invention is adopted the following technical scheme that：

A kind of Noise Filter method of signal specific, it is characterised in that：Preset T_mRepresent Discrete Periodic Signals M-th supposition amount, A_mIt is m-th observation with certain error, K is filtering depth, and △ is adjustment stepping, and R is discrete week The jitter amplitude of phase signal, filter step is as follows：

1) when initial, T is made_m=A_m, n=1；

2) according to current observation A_mWith the relation of mapping function f (m-n), guess value T is drawn_m-n, and by T_m-nWith M-n observation A_m-nMake comparisons, if T_m-n+ R ＜ A_m-n, then according to adjustment stepping △ to T_mCarry out increasing amount adjustment；If T_m-n+ R ＞ A_m-n, then to T_mCarry out decreasing amount adjustment；N=n+1；

3) n ＜ K are judged, if so, then returning to 2)；If it is not, then exporting T_m。

Preferably, Z is preset_mIt is m-th actual value without error, it obeys mapping relations：Z_m-n=Z_m+f(m- n)；In step 2) in, it is assumed that current observation is equal to actual value, then the current observation and mapping function f (m-n) Relation is A_m-n=A_m+ f (m-n), the guess value for drawing is T_m-n=T_m+f(m-n)。

Preferably, it is described according to adjustment stepping △ to T_mCarry out increasing amount adjustment, including the two add, is subtracted, multiplied or Division operation.

Preferably, it is described according to adjustment stepping △ to T_mCarry out decreasing amount adjustment, including the two add, is subtracted, multiplied or Division operation.

Preferably, the R is related to the statistical properties of random noise.

Preferably, the R takes the 75%-85% of the random noise of the jitter amplitude of Discrete Periodic Signals.

From the above-mentioned description of this invention, compared with prior art, the present invention has the advantages that：

Filtering method of the invention is according to a series of observation A_mThe a series of guess value T for obtaining_m, with reference to discrete periodic With observation be compared guess value by the jitter amplitude R of signal, according to comparative result to T_mIncrement or decreasing amount adjustment are carried out, with Increasing for adjustment number of times, T_mZ can be become closer to_m, can effectively weaken random noise r_mInfluence.In addition, filtering side of the invention Method has relatively good filter effect to linear signal and nonlinear properties (such as periodic signal).

Brief description of the drawings

Fig. 1 is the flow chart of filtering method of the invention；

Fig. 2 is linear signal filter effect figure of the invention；

Fig. 3 is cyclical signal filter effect figure of the invention.

Specific embodiment

Below by way of specific embodiment, the invention will be further described.

For signal specific, it is assumed that it is superimposed with the random noise that 75%-85% (preferably 80%) amplitude is [- R, R].Should The Mathematical Modeling of signal specific is A_m=Z_m+r_m,r_mFor 80% between interval [- R ,+R] random signal m-th value, A_mIt is The m observation with certain error, Z_mIt is m-th of the actual value without error, Z_mObey known mapping relations Z_m-n= Z_m+f(m-n).For example, f (m-n)=C (m-n) when representing linear signal, C is constant；F (m-n)=0 when representing periodic signal, As m-n=NT, N is integer, T is the cycle.

Reference picture 1, a kind of Noise Filter method of signal specific of the invention, it is filtering depth, △ to preset K It is adjustment stepping, R is the jitter amplitude of Discrete Periodic Signals.Can basis signal sample fluctuation characteristic, choose suitable K, △, R, filter step is as follows：

1) when initial, T is made_m=A_m, n=1.

2) assume that current observation is equal to actual value, then the relation of the current observation and mapping function f (m-n) is A_m-n=A_m+ f (m-n), the guess value for drawing is T_m-n=T_m+ f (m-n) (during periodic signal n=NT, N is integer, and T is the cycle). By T_m-nWith the m-n observation A_m-nMake comparisons, if T_m-n+ R ＜ A_m-n, then according to adjustment stepping △ to T_mCarry out increasing amount adjustment, That is T_m=T_m+△；If T_m-n+ R ＞ A_m-n, then to T_mCarry out decreasing amount adjustment, i.e. T_m=T_m-△；N=n+1.

3) n ＜ K are judged, if so, then returning to 2)；If it is not, then exporting T_m.With increasing for adjustment number of times, the T of output_mWill be more Come closer to Z_m, therefore, it is possible to replace observation A_mExport as a result.

The method of the present invention, above-mentioned increasing amount adjustment and decreasing amount adjustment also can use other computing mode realities such as multiplication, division It is existing.

Applicating example

Such as close with the signal specific model of the application 1PPS shake models for having navigation time dissemination system.Its A_m=Z_m+ r_m, Z_m-n=Z_m+ (m-n) * 1*f, f are the precise frequencies of local clock, r_mIt is random between interval [- 100ns ,+100ns] Noise.A_mIt is m-th observation of arrival time of 1PPS, Z_mIt is m-th actual value of arrival time of 1PPS, r_mIt is navigation time service The random clock correction of 1PPS of system.Can obtain a series of being substantially better than 1PPS arrival times and observing by the filtering algorithm of the application Amount A_mSupposition arrival time T_m, and then restore 1PPS of the precision better than ± 100ns.

Reference picture 2 is linear signal filter effect figure, and the sample of signal is:F (m-n)=0, r_mIt is fluctuation in [- 10 10] Random noise be superimposed the infrequent noise of [- 50 50].Abscissa is signal sequence number, ordinate signal value.

Reference picture 3 is cyclical signal filter effect figure, and sample of signal is:Cycle is 50 cosine waves of signaling point, i.e. f (m-n) n=NT=0, is worked as, N is integer, during T=50, r_mFor the random noise of fluctuation in [- 10 10] is superimposed one [- 50 50] Infrequent noise.Abscissa is signal sequence number, ordinate signal value.

Specific embodiment of the invention is above are only, but design concept of the invention is not limited thereto, it is all to utilize this Design carries out the change of unsubstantiality to the present invention, all should belong to the behavior for invading the scope of the present invention.

Claims

1. a kind of Noise Filter method of signal specific, it is characterised in that：Preset T_mRepresent Discrete Periodic Signals m Individual supposition amount, A_mIt is m-th observation with certain error, K is filtering depth, and △ is adjustment stepping, and R is discrete periodic letter Number jitter amplitude, filter step is as follows：

1) when initial, T is made_m=A_m, n=1；

2. the Noise Filter method of a kind of signal specific as claimed in claim 1, it is characterised in that：Preset Z_mFor M-th actual value without error, it obeys mapping relations：Z_m-n=Z_m+f(m-n)；In step 2) in, it is assumed that current observation Value is equal to actual value, then the relation of the current observation and mapping function f (m-n) is A_m-n=A_m+ f (m-n), draws Guess value is T_m-n=T_m+f(m-n)。

3. the Noise Filter method of a kind of signal specific as claimed in claim 1, it is characterised in that：In step 2) in, It is described according to adjustment stepping △ to T_mCarry out increasing amount adjustment, including the two add, subtract, multiplied or division operation.

4. the Noise Filter method of a kind of signal specific as claimed in claim 1, it is characterised in that：In step 2) in, It is described according to adjustment stepping △ to T_mCarry out decreasing amount adjustment, including the two add, subtract, multiplied or division operation.

5. the Noise Filter method of a kind of signal specific as claimed in claim 1, it is characterised in that：The R with it is random The statistical properties of noise are related.

6. the Noise Filter method of a kind of signal specific as claimed in claim 1, it is characterised in that：The R takes discrete The 75%-85% of the random noise of the jitter amplitude of periodic signal.