Signal saliency analysis method based on local noise amplitude probability model

Abstract

The invention relates to a signal processing method, in particular to a signal saliency analysis method based on a local noise amplitude probability model, and aims to solve the problems that when signal detection is carried out by using the prior art, the background noise amplitude of a signal is relatively large, so that a detection condition is mistakenly triggered, and a detection result is influenced. In the operation process, firstly, an original time sequence waveform containing a signal to be detected is determined and subjected to primary processing, the linear trend of the original time sequence waveform is removed, and filtering is carried out; secondly, respectively selecting a noise amplitude statistical window and one or more signal windows comprising to-be-detected signals in the filtering waveform, and respectively calculating the root-mean-square amplitude of noise and the root-mean-square amplitude of the signals in the signal windows; then, calculating the probability that the root-mean-square amplitude of the noise reaches or exceeds the root-mean-square amplitude of the signal in the signal window; and finally, setting a judgment condition, and judging whether the to-be-detected signal is significant relative to background noise or not according to the obtained probability.

Description

Technical Field

[0001] The invention relates to a signal processing method, in particular to a signal significance analysis method based on a local noise amplitude probability model. Background Art

[0002] In the field of signal processing, accurately detecting the required target signal from complex background noise is a very important research topic, which has important applications in many fields such as wireless communication, medical imaging, geophysical exploration, satellite remote sensing imaging, speech signal processing, etc. However, background noise may contain a variety of interferences, such as the uncertainty of the signal source itself, the internal noise of electronic equipment, and the noise generated by external environmental factors, all of which may affect the detection and analysis of the signal.

[0003] Therefore, usually before signal detection, the signal-to-noise ratio must be improved first. Filtering and other methods can be used to suppress noise and amplify the target signal. If the signal-to-noise ratio of the signal is high, the energy ratio method, high-order statistics method, periodogram method and other methods can generally detect the signal well; but if the signal-to-noise ratio of the signal itself is low, then when the above methods are applied, it is impossible to accurately determine whether the detected signal is a weak signal or a large-amplitude noise or interference.

[0004] On the other hand, there are two different situations of background noise. One is that the background noise is relatively stable, with small fluctuations in amplitude, which has limited impact on signal detection; the other is that the background noise is not stable, with more interference with large amplitudes. In the latter case, the application of traditional signal detection methods, such as the energy ratio method, may affect the signal detection due to the large amplitude interference in the noise, triggering the detection conditions, thereby increasing the false detection rate. Summary of the invention

[0005] The main purpose of the present invention is to solve the problem that the background noise amplitude of the signal is large in the traditional signal detection method, thereby falsely triggering the detection condition and affecting the detection result, and to provide a signal significance analysis method based on the local noise amplitude probability model.

[0006] The idea of ​​the present invention is to firstly model the probability distribution of the local amplitude of the background noise, and then determine the probability that the amplitude of the noise itself reaches or exceeds the amplitude of the signal to be detected, so as to evaluate whether the signal to be detected is significant enough relative to the noise, and then complete the significance analysis of the signal.

[0007] In order to achieve the above invention objectives and complete the above invention concepts, the present invention provides the following technical solutions:

[0008] A signal significance analysis method based on a local noise amplitude probability model is characterized in that it comprises the following steps:

[0009] Step 1: Determine a section of the original timing waveform containing the signal to be detected, and record the arrival time of the signal to be detected as t s , and at the arrival time t of the signal to be detected s There are no other signals to be detected within the time range of length T before and after;

[0010] Step 2, removing the linear trend of the original time series waveform to obtain a waveform after the linear trend is removed;

[0011] Step 3, using m bandpass digital filters of preset frequency intervals to perform bandpass filtering on the waveform after removing the linear trend, to obtain m segments of filtered waveforms in the preset frequency intervals; wherein m is an integer greater than 0;

[0012] Step 4: In each segment of the filtered waveform of the preset frequency interval of m segments, k signal windows of different lengths including the signal to be detected are selected, and the root mean square amplitude A of the signal in each signal window is calculated. sk (m) ; Among them, the length of each signal window is T sk , k is an integer greater than 0;

[0013] Step 5: In each of the m segments of the filtering waveform in the preset frequency interval, a segment of length T is taken before the signal to be detected. nw The time window of is used as the noise amplitude statistical window, and T>T nw >T sk ; In the noise amplitude statistics window, according to the set sliding step size T step Calculate the root mean square amplitude A of the noise in the window with the same length as each signal window nk , get k noise amplitude vectors

[0014] Step 6: Assuming that the noise amplitude follows a normal distribution, calculate the m×k noise amplitude vectors respectively. Reach or exceed the signal RMS amplitude A within the signal window sk (m) The probability P sk (m) , we get a probability set P = {P s1 (1) ,P s2 (1) ,......,P sk (1) ;P s1 (2) ,P s2 (2) ,......,P sk (2) ;......;Ps1 (m) ,P s2 (m) ,......,P sk (m)};

[0015] Step 7: Construct the function f(A,ε), which represents the number of elements in the set A whose values ​​are less than ε; set the judgment condition as: f(P,P th )>c, substitute the probability set obtained in step 6 into this condition for judgment. When the probability set P satisfies f(P,P th )>c, the signal to be detected is judged to be significant relative to the background noise; if this condition is not met, the signal to be detected is judged to be insignificant relative to the background noise; and then the signal significance analysis based on the local noise amplitude probability model is completed; wherein, P th is the preset probability threshold; c is an integer greater than or equal to 0, indicating that the elements in the preset set P should be less than the probability threshold P th The number of .

[0016] Furthermore, step 6 is specifically as follows:

[0017] Step 6.1: Calculate m×k noise amplitude vectors respectively The mean μ nk (m) and standard deviation σ nk (m) ;

[0018] Step 6.2: Calculate m×k noise amplitude vectors respectively Reach or exceed the root mean square amplitude A of each signal including the signal to be detected sk (m) Probability Get a probability set

[0019] P={P s1 (1) ,P s2 (1) ,......,P sk (1) ;P s1 (2) ,P s2 (2) ,......,P sk (2) ;......;P s1 (m) ,P s2 (m) ,......,P sk (m)}; where x is the value of the random variable.

[0020] Furthermore, in step 4, the k signal windows with different lengths including the signal to be detected are represented as [t s ,t s +T sk ].

[0021] Furthermore, the arrival time t of the signal to be detected s , the length of the signal window T sk , the length of the noise amplitude statistical window T nw , sliding step length T step The unit is seconds.

[0022] Compared with the prior art, the present invention has the following beneficial effects:

[0023] 1. The present invention proposes a signal significance analysis method based on a local noise amplitude probability model. First, a model is established based on the local noise amplitude of the original waveform. On this basis, the probability that the amplitude of the noise itself reaches or exceeds the amplitude of the signal containing the signal to be detected is calculated. This method can adapt to different noise environments, and can optimize the detection performance by adjusting the model parameters to improve the reliability of signal detection.

[0024] 2. The signal significance analysis method based on the local noise amplitude probability model proposed in the present invention can significantly reduce the false detection rate when the background noise is unstable and there is a lot of noise interference with large amplitude.

[0025] 3. The signal significance analysis method based on the local noise amplitude probability model proposed in the present invention has low computational complexity and can be used for real-time signal processing. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] Figure 1 A schematic flow chart of a signal significance analysis method based on a local noise amplitude probability model of the present invention;

[0027] Figure 2 is an original timing waveform diagram containing a signal to be detected used in the first embodiment of the present invention;

[0028] Figure 3 This is a timing waveform diagram after filtering using a bandpass filter with a frequency band of 1-2 Hz in Embodiment 1 of the present invention;

[0029] Figure 4 is an original timing waveform diagram containing a signal to be detected used in the second embodiment of the present invention;

[0030] Figure 5The timing waveform diagrams are after filtering using bandpass filters of different frequency bands in the second embodiment of the present invention, wherein (a) is the timing waveform diagram after filtering using a bandpass filter with a frequency band of 0.5-2 Hz; (b) is the timing waveform diagram after filtering using a bandpass filter with a frequency band of 2-4 Hz; (c) is the timing waveform diagram after filtering using a bandpass filter with a frequency band of 3-6 Hz. DETAILED DESCRIPTION

[0031] The present invention is further described below in conjunction with the accompanying drawings and specific embodiments. Obviously, the described embodiments are part of the embodiments of the present invention, rather than all of the embodiments. Based on the technical solutions in the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work belong to the scope of protection of the present invention.

[0032] The method flow charts of the following two embodiments are as follows Figure 1 shown.

[0033] Embodiment 1

[0034] The timing waveform of the original signal containing the signal to be detected used in this embodiment is as follows: Figure 2 As shown in the figure, the horizontal axis time is the arrival time of the signal, the unit is seconds; the vertical axis is the amplitude of the signal, which is a dimensionless quantity, the unit is 1.

[0035] A signal significance analysis method based on a local noise amplitude probability model comprises the following steps:

[0036] Step 1: Determine a section of the original timing waveform containing the signal to be detected. The waveform is 1000s long and the signal to be detected is at 500s, that is, t s =500s, there is no other signal to be detected in this timing waveform;

[0037] Step 2, removing the linear trend of the original time series waveform to obtain a waveform after the linear trend is removed;

[0038] Step 3: Use a bandpass digital filter with a frequency band of 1-2 Hz to filter the waveform after removing the linear trend. The resulting waveform is as follows: Figure 3 As shown, the red arrow points out the signal to be detected;

[0039] Step 4: In the 1-2 Hz filtering waveform, three signal windows [500s, 510s], [500s, 502.5s], and [500s, 501.5s] including the signal to be detected are selected. The lengths of the three signal windows are T s1 =10s, T s2 =2.5s, T s3 =1.5s; calculate the root mean square amplitude A of the signal in the three windows respectivelys1 (1) , A s2 (1) and A s3 (1) ;

[0040] Step 5: In the 1-2 Hz filter waveform, take a length T before the signal to be detected nw =300s time window is used as the noise amplitude statistical window, and the right boundary of the time window is 5s away from the arrival time of the signal to be detected, that is, Figure 3 The interval [195s, 495s] is the noise amplitude statistical window. Within the noise amplitude statistical window, the sliding step size T step = 1s respectively calculated with the signal window T s1 , T s2 and T s3 The root mean square amplitude of the noise in the window of the same length is obtained, and three sets of vectors are obtained. and

[0041] Step 6.1: In this embodiment, the noise amplitude follows a normal distribution, and the vectors are calculated respectively. and The mean μ n1 (1) , μ n2 (1) , μ n3 (1) and standard deviation σ n1 (1) , σ n2 (1) , σ n3 (1) ;

[0042] Step 6.2: Calculate the probability that the noise amplitude reaches or exceeds the root mean square amplitude of the signal to be detected. The result is as follows:

[0043]

[0044] Thus we get the probability set P = {P s1 (1) ,P s2 (1) ,P s3 (1)};

[0045] Step 7: Set the judgment condition as: f(P,P th )>0, the signal to be detected is significant relative to the background noise, where the probability threshold P th =10 -5 ; Substitute the probability set P obtained in step 6.2 into this formula to obtain f({P s1(1) ,P s2 (1) ,P s3 (1)},10 -5 )=2>0, the condition is met, so the signal is significant.

[0046] Embodiment 2

[0047] The timing waveform of the original signal containing the signal to be detected used in this embodiment is as follows: Figure 4 As shown in the figure, the horizontal axis time is the arrival time of the signal, the unit is seconds; the vertical axis is the amplitude of the signal, which is a dimensionless quantity, the unit is 1.

[0048] A signal significance analysis method based on a local noise amplitude probability model comprises the following steps:

[0049] Step 1: Determine a section of the original waveform containing the signal to be detected. The waveform is 1000s long and the signal to be detected is at 500s, that is, t s =500s, there is no other signal to be detected in this timing waveform;

[0050] Step 2, removing the linear trend of the original time series waveform to obtain a waveform after the linear trend is removed;

[0051] Step 3: Use bandpass digital filters with frequency bands of 0.5-2 Hz, 2-4 Hz, and 3-6 Hz to filter the waveform after removing the linear trend. The obtained waveforms are as follows: Figure 5 (a) Figure 5 (b) Figure 5 (c) shows that all red arrows indicate signals to be detected;

[0052] Step 4: Take a signal window [500s, 505s] including the signal to be detected from the filtered waveforms of 0.5-2Hz, 2-4Hz, and 3-6Hz respectively. The length of the signal window is T s1 =5s, calculate the RMS amplitude A of the signal in the signal window s1 (1) ;

[0053] Step 5: In the filtered waveforms of 0.5-2 Hz, 2-4 Hz, and 3-6 Hz, take a length T before the signal to be detected. nw =300s time window is used as the noise amplitude statistical window, and the right boundary of the time window is 5s away from the arrival time of the signal to be detected, that is, Figure 5 (a) Figure 5 (b) Figure 5 In (c), the interval [195s, 495s] is the noise amplitude statistical window. In the noise amplitude statistical window, according to the sliding step size Tstep = 1s calculation and signal window T s1 The root mean square amplitude of the noise in the window of the same length, resulting in three sets of vectors

[0054] Step 6.1: In this embodiment, the noise amplitude follows a normal distribution, and the vector is calculated. The mean μ n1 (1) , μ n1 (2) , μ n1 (3) and standard deviation σ n1 (1) , σ n1 (2) , σ n1 (3) ;

[0055] Step 6.2: Obtain the probability that the noise amplitude reaches or exceeds the RMS amplitude of the signal containing the signal to be detected. The result is as follows:

[0056]

[0057] Thus we get the probability set P = {P s1 (1) ,P s1 (2) ,P s1 (3)};

[0058] Step 7: Set the judgment condition as: f(P,P th )>0, the signal to be detected is significant relative to the background noise, where the probability threshold P th =10 -5 ; Substitute the probability set P obtained in step 6.2 into this formula to obtain f({P s1 (1) ,P s1 (2) ,P s1 (3)},10 -5 )=0, the condition is not met, so the signal is not significant.

[0059] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit them. For ordinary professional and technical personnel in the field, the specific technical solutions recorded in the above embodiments can be modified, or some of the technical features therein can be replaced by equivalents, and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the scope of the technical solutions protected by the present invention.

Claims

1. A signal significance analysis method based on a local noise amplitude probability model, characterized in that: The following steps are involved: Step 1: Determine a section of the original timing waveform containing the signal to be detected, and record the arrival time of the signal to be detected as t s , and at the arrival time t of the signal to be detected s There are no other signals to be detected within the time range of length T before and after; Step 2, removing the linear trend of the original time series waveform to obtain a waveform after the linear trend is removed; Step 3, using m bandpass digital filters of preset frequency intervals to perform bandpass filtering on the waveform after removing the linear trend, to obtain m segments of filtered waveforms in the preset frequency intervals; wherein m is an integer greater than or equal to 0; Step 4: In each segment of the filtered waveform of the preset frequency interval of m segments, k signal windows of different lengths including the signal to be detected are selected, and the root mean square amplitude A of the signal in each signal window is calculated. sk (m) ; Among them, the length of each signal window is T sk , k is an integer greater than 0; Step 5: In each of the m segments of the filtering waveform in the preset frequency interval, a segment of length T is taken before the signal to be detected. nw The time window of is used as the noise amplitude statistical window, and T>T nw >T sk ; In the noise amplitude statistics window, according to the set sliding step size T step Calculate the root mean square amplitude A of the noise in the window with the same length as each signal window nk , get k noise amplitude vectors Step 6: Assuming that the noise amplitude follows a normal distribution, calculate the m×k noise amplitude vectors respectively. Reach or exceed the signal RMS amplitude A within the signal window sk (m) The probability P sk (m) , we get a probability set P = {P s1 (1) ,P s2 (1) ,......,P sk (1) ;P s1 (2) ,P s2 (2) ,......,P sk (2) ;......;P s1 (m) ,P s2 (m) ,......,P sk (m) }; Step 7: Construct the function f(A,ε), which represents the number of elements in the set A whose values ​​are less than ε; set the judgment condition as: f(P,P th )>c, substitute the probability set obtained in step 6 into this condition for judgment. When the probability set P satisfies f(P,P th )>c, the signal to be detected is judged to be significant relative to the background noise; if this condition is not met, the signal to be detected is judged to be insignificant relative to the background noise; and then the signal significance analysis based on the local noise amplitude probability model is completed; wherein, P th is the preset probability threshold; c is an integer greater than or equal to 0, indicating that the elements in the preset set P should be less than the probability threshold P th The number of .

2. The signal significance analysis method based on the local noise amplitude probability model according to claim 1 is characterized in that: Step 6 is as follows: Step 6.1: Calculate m×k noise amplitude vectors respectively The mean μ nk (m) and standard deviation σ nk (m) ; Step 6.2: Calculate m×k noise amplitude vectors respectively Reach or exceed the root mean square amplitude A of each signal including the signal to be detected sk (m) Probability Get a probability set P={P s1 (1) ,P s2 (1) ,......,P sk (1) ;P s1 (2) ,P s2 (2) ,......,P sk (2) ;......;P s1 (m) ,P s2 (m) ,......,P sk (m) }; where x is the value of the random variable.

3. The signal significance analysis method based on the local noise amplitude probability model according to claim 2 is characterized in that: In step 4, the k signal windows with different lengths including the signal to be detected are represented by [t s ,t s +T sk ].

4. A signal significance analysis method based on a local noise amplitude probability model according to any one of claims 1 to 3, characterized in that: The arrival time t of the signal to be detected s , the length of the signal window T sk , the length of the noise amplitude statistical window T nw , sliding step length T step The unit is seconds.

Images