Signal characteristic parameter estimation method based on wavelet decomposition

Abstract

The invention provides a signal characteristic parameter estimation method based on wavelet decomposition, and belongs to the field of signal time-frequency analysis. According to the method, a signal time domain high-frequency component is obtained through a wavelet decomposition method, then a high-frequency component noise average amplitude is calculated through a histogram statistical sorting method, starting and ending positions of a burst signal are determined according to a time domain amplitude threshold method, and finally parameters such as retention time, time interval and frequency domain bandwidth of the burst signal are calculated according to the starting and ending positions. And more basis is provided for signal classification. The method is suitable for signals with the center frequency not hopped or frequency hopping signals with the center frequency as the center and the jumping range not larger than one fourth of the monitoring bandwidth, and has the advantages of being small in resource consumption, easy to achieve and the like.

Description

Technical Field

[0001] The present invention belongs to the field of signal time-frequency analysis, and particularly relates to a method for estimating signal characteristic parameters based on wavelet decomposition. Background Art

[0002] With the development of electronic countermeasures and anti-jamming communication, the detection, recognition, blind detection and recognition of communication signals under non-cooperative conditions are becoming increasingly important. The traditional recognition methods relying solely on the time domain or the frequency domain can no longer meet the requirements for signal detection. Therefore, multi-domain joint recognition methods in the time domain, frequency domain and spatial domain have emerged.

[0003] Currently, most of the existing technologies adopt methods such as wavelet transform, energy detection, EMD and short-time Fourier transform (STFT). Among them, the energy detection method has weak detection ability for burst short-time signals; the EMD method is complex to implement, consumes a large amount of resources and cannot achieve real-time calculation; for the STFT method, when the time window is determined, its time resolution and frequency resolution are already determined, which cannot meet the requirements of multi-resolution analysis; in addition, the wavelet transform method has a large amount of calculation, consumes more resources, takes a long time to complete a signal detection, and is not suitable for real-time signal detection scenarios. Summary of the Invention

[0004] In view of the above deficiencies, the present invention provides a method for estimating signal characteristic parameters based on wavelet decomposition. This method is applicable to signals with non-hopping center frequencies or frequency-hopping signals with a hopping range not greater than one-fourth of the monitoring bandwidth centered on the center frequency, and has the characteristics of less resource consumption and simple implementation, and can provide more basis for signal classification.

[0005] The object of the present invention is achieved through the following solutions:

[0006] A method for estimating signal characteristic parameters based on wavelet decomposition, comprising the following steps:

[0007] Step 1, receiving and tracking signals frame by frame;

[0008] Step 2, processing each frame of signal, finding the positions of the first occurrence and the second occurrence of the signal, and obtaining the time length, signal bandwidth and time interval of the signal.

[0009] Further, the specific manner of Step 2 is:

[0010] Step 201, converting the tracking signal from an analog signal into a digital signal, and the sampling frequency of the digital signal is ;

[0011] Step 202, performing digital down-conversion on the digital signal to obtain a digital baseband time-domain signal y = , , … , where 0, 1, …, 2n are the number of signal points, there are N frequency points in total, and N = 2n + 1;

[0012] Step 203: Select the wavelet function db7, and obtain the decomposition high-pass filter Hi_D_filter according to the Mallat algorithm;

[0013] Step 204: Convolve the digital baseband time-domain signal with Hi_D_filter to obtain the high-frequency component cD1 = 、 … , cD1 contains Q elements in total, and Q = n + 1;

[0014] Step 205: Calculate the time-domain signal amplitude of each point i in the high-frequency component cD1 , i = 0, 2, …, 2n;

[0015] Step 206: Find the maximum value and the minimum value in the time-domain signal amplitude, and calculate the range ;

[0016] Step 207: Divide the Q time-domain signal amplitudes into 64 spaces, the space number it belongs to is , where is the division gradient;

[0017] Step 208: Count the number of amplitudes in the space and the average amplitude ;

[0018] Step 209: In the way of histogram statistics, obtain the maximum value that meets the following conditions:

[0019]

[0020] And calculate the average amplitude obtained by statistics under this condition:

[0021]

[0022] where m is the cumulative number of amplitudes that meet the conditions;

[0023] Step 210: Set the threshold , the value of the threshold factor is 2.5 to 3, and the Q time-domain signal amplitudes are compared with the threshold one by one in orderCompare them. Use the fact that the amplitude of the time-domain signal is greater than T as the start or end flag of the signal. Denote the indices corresponding to the amplitudes of the time-domain signal at the first start and end of the signal as and , and denote the index corresponding to the amplitude of the time-domain signal at the second start of the signal as ; The index refers to the unique serial number set for each frequency point in the digital baseband time-domain signal starting from the first frame of the signal;

[0024] Step 211, calculate the time length when the signal appears:

[0025]

[0026] where U represents the difference between the frame number of the signal frame where the index is located and the frame number of the signal frame where the index is located;

[0027] Step 212, intercept the digital baseband time-domain signal data between the index and the index . If is not an integer power of 2, then fill 0 at the end of the intercepted data until the data length is , where k is an integer; perform FFT operation on the obtained data, and obtain the power spectrum of each point according to the FFT operation result , t = 0, 1, …, , count the number of frequency points Q where the power spectrum is greater than , and finally obtain the signal bandwidth ;

[0028] Step 213, calculate the time interval when the signal appears:

[0029]

[0030] where M is the total number of frames from the signal frame where the index is located to the signal frame where the index is located;

[0031] Thus far, the signal characteristic parameters , and are obtained.

[0032] The present invention has the following advantages compared with the background technology:

[0033] 1. The present invention adopts the histogram statistical sorting method, and the computational time complexity is , with high computational efficiency and good real-time performance.

[0034] 2. The present invention adopts a wavelet decomposition algorithm, which can obtain the holding time and time interval parameters of a burst signal. This method consumes less resources and is simple to implement.

[0035] 3. The present invention combines signal time positioning and uses the FFT method to obtain the frequency domain bandwidth parameter of a burst signal, which is simple and easy to implement.

[0036] 4. The present invention is applicable to signals with a non-hopping center frequency or frequency-hopping signals with a hopping range not greater than one-quarter of the monitoring bandwidth centered on the center frequency, and can provide more bases for signal classification. Specific Embodiment

[0037] The present invention will be further described in detail below.

[0038] A method for estimating signal characteristic parameters based on wavelet decomposition. This method obtains the high-frequency component in the time domain of the signal through wavelet decomposition, then calculates the average amplitude of the high-frequency component noise through the histogram statistical sorting method, determines the start and end positions of the burst signal according to the time domain amplitude threshold method, and finally calculates parameters such as the holding time, time interval, and frequency domain bandwidth of the burst signal based on the start and end positions. The specific steps are as follows:

[0039] Step 1, receive and track the signal frame by frame;

[0040] Step 2, process each frame of the signal to find the positions of the first occurrence and the second occurrence of the signal, and obtain the time length, signal bandwidth, and time interval of the signal. The specific method is as follows:

[0041] Step 201, convert the tracked signal from an analog signal to a digital signal, and the sampling frequency of the digital signal is ;

[0042] Step 202, perform digital down-conversion on the digital signal in the FPGA to obtain the digital baseband time domain signal y = 、 、 … , where 0, 1, …, 2n are the signal points, with a total of N frequency points, and N = 2n + 1;

[0043] Step 203, select the wavelet function "db7", and according to the Mallat algorithm, obtain the wavelet filter: the decomposition high-pass filter Hi_D_filter;

[0044] Step 204, convolve the digital baseband time domain signal with Hi_D_filter to obtain the high-frequency component cD1 = 、 … , cD1 contains a total of Q elements, and Q = n + 1;

[0045] Step 205, calculate the time-domain signal amplitude of each point i in the high-frequency component cD1 , where i = 0, 2, …, 2n;

[0046] Step 206, find the maximum value and the minimum value in the time-domain signal amplitude, and calculate the range ;

[0047] Step 207, divide the Q time-domain signal amplitudes into 64 spaces, and the space number they belong to is , where is the division gradient. When , is 1 dB. When , ;

[0048] Step 208, count the number of amplitudes in space and the average amplitude ;

[0049] Step 209, in the way of histogram statistics, obtain the maximum value that meets the following conditions:

[0050]

[0051] and calculate the average amplitude statistically obtained under this condition:

[0052]

[0053] where m is the cumulative number of amplitudes that meet the conditions;

[0054] can be updated once when processing each frame of signal, or other update frequencies can be set according to engineering requirements;

[0055] Step 210, set the threshold , and the value of the threshold factor is 2.5 to 3. Compare the Q time-domain signal amplitudes with the threshold one by one in order. Use the time-domain signal amplitude greater than T as the start or end flag of the signal. That is, when the time-domain signal amplitude is greater than T for the first time, it means the signal starts. When the time-domain signal amplitude is greater than T for the second time, it means the signal ends. When the time-domain signal amplitude is greater than T for the third time, it means the start of the next signal, and so on;

[0056] Record the indices corresponding to the time-domain signal amplitudes at the first start and end of the signal as and ,record the index corresponding to the amplitude of the time-domain signal when the signal starts for the second time as ; The index refers to the unique serial number set for each frequency point in the digital baseband time-domain signal starting from the first frame of the signal;

[0057] Step 211, calculate the time length when the signal appears:

[0058]

[0059] where U represents the difference between the frame number of the signal frame where the index is located and the frame number of the signal frame where the index is located; for example, if the index appears in the 3rd frame of the signal and the index appears in the 5th frame of the signal, then U = 5 - 3 = 2.

[0060] Step 212, intercept the digital baseband time-domain signal data between the index and the index . If is not an integer power of 2, then adopt the principle of padding with 0 values nearby, and add 0 at the end of the intercepted data until the data length is , where k is an integer; perform FFT operation on the obtained data, and obtain the power spectrum of each point according to the FFT operation result , t = 0, 1, …, , count the number of frequency points Q whose power spectrum is greater than , and finally obtain the signal bandwidth ;

[0061] Step 213, calculate the time interval when the signal appears:

[0062]

[0063] where M is the total number of frames from the signal frame where the index is located to the signal frame where the index is located; for example, if the index appears in the 5th frame of the signal and the index appears in the 8th frame of the signal, then M = 8 - 5 + 1 = 4.

[0064] So far, the signal characteristic parameters , and are obtained.

[0065] The present invention has the characteristics of less resource consumption and simple implementation, and can provide more basis for signal classification.

Claims

1. A signal characteristic parameter estimation method based on wavelet decomposition, characterized in that: The following steps are involved: Step 1, receiving tracking signals frame by frame; Step 2, process each frame of the signal, find the positions of the first and second occurrences of the signal, and obtain the time length, signal bandwidth, and time interval of the signal.

2. The signal characteristic parameter estimation method based on wavelet decomposition according to claim 1 is characterized in that: The specific method of step 2 is: Step 201, convert the tracking signal from an analog signal to a digital signal, the sampling frequency of the digital signal is ; Step 202, digital down-convert the digital signal to obtain a digital baseband time domain signal y=[ , , … ], where 0, 1, …, 2n are the number of signal points, with a total of N frequency points, N = 2n + 1; Step 203, select the wavelet function db7, and obtain the decomposition high-pass filter Hi_D_filter according to the Mallat algorithm; Step 204, convolve the digital baseband time domain signal with Hi_D_filter to obtain a high frequency component cD1=[ , … ], cD1 contains a total of Q elements, Q=n+1; Step 205, calculate the time domain signal amplitude of each point i in the high frequency component cD1 , i=0,2,…,2n; Step 206, find the maximum value in the time domain signal amplitude and minimum value , and calculate the range ; Step 207, divide the Q time domain signal amplitudes into 64 spaces, The space number is ,in, To divide the gradient; Step 208: Statistical space The number of amplitudes in and the average amplitude ; Step 209, according to the histogram statistics method, obtain the maximum value that meets the following conditions value: And calculate the average amplitude obtained under this condition: Where m is the number of cumulative amplitudes that meet the conditions; Step 210, setting threshold , threshold factor The value of is 2.5~3, and the amplitudes of Q time domain signals are compared with the threshold one by one in sequence. The time domain signal amplitude greater than T is used as the sign of the start or end of the signal, and the index corresponding to the time domain signal amplitude when the signal first starts and ends is recorded as and , the index corresponding to the amplitude of the time domain signal when the signal starts for the second time is recorded as ; The index refers to a unique serial number set for each frequency point in the digital baseband time domain signal starting from the first frame signal; Step 211, calculate the duration of the signal appearance: Among them, U represents the index The frame number and index of the signal frame The difference in the frame number of the signal frame; Step 212, extract index To Index If the digital baseband time domain signal data between If it is not an integer power of 2, add 0 to the end of the intercepted data until the data length is , k is an integer; perform FFT operation on the obtained data, and obtain the power spectrum of each point according to the FFT operation result ,t=0,1,…, , the statistical power spectrum is greater than The frequency point number Q, finally get the signal bandwidth ; Step 213, calculate the time interval between signal occurrences: Among them, M is the index The signal frame to index The total number of frames in the signal frame; At this point, the signal characteristic parameters are obtained , and .

Images