Method for determining starting and ending time and starting and ending frequency of signal based on time-frequency analysis

Abstract

The invention relates to a method for determining starting and ending time and starting and ending frequency of a signal based on time-frequency analysis, which is characterized by comprising the following steps of: on the basis of a time-frequency analysis image and a related method, firstly, extracting frequency data in a signal frequency range in pairs, judging starting and ending time of an effective signal in a time domain after judging the effective signal in the frequency domain; and taking out the continuous time section and the continuous time length of the corresponding effective signal, finding out the maximum amplitude, and further determining the start-stop frequency and time of the effective signal. Judging whether an effective signal exists at the position or not according to the data characteristics in the frequency and time starting and ending positions; the method has the advantages in the application aspect of determining the frequency and time starting and ending positions of the signal, especially the burst signal.

Description

Technical Field

[0001] This invention relates to a method for determining the start and end times and start and end frequencies of a signal based on time-frequency analysis. Background Technology

[0002] A crucial task of information science is to acquire accurate information about the objective world. However, noise is inherent in any system, and extracting the true signal becomes difficult when the measured signal is weak and submerged in strong noise. Therefore, the problem of extracting useful signals from noise is attracting increasing attention. Furthermore, burst signals in a system, compared to continuous long signals, have uncertain start and end points, exhibiting strong concealment; moreover, they are shorter. These factors significantly increase the difficulty of detecting and identifying such signals.

[0003] In response to the increasingly complex use of the spatial spectrum and the application background requirement of accurately detecting the start and end times of typical airborne wireless signal targets, this study conducts research on key technologies such as wireless signal detection based on time-frequency analysis and edge detection. The main focus is on determining the occupancy of the spectrum in a certain frequency band in terms of time and spectrum: determining the location, bandwidth, start and end times, and spectral intensity of the spectrum occupied by signals with a certain strength, thereby achieving more sensitive signal detection and providing support for improving the intelligence level of spatial spectrum wireless signal monitoring. Summary of the Invention

[0004] Because the original frequency and time ranges are difficult to achieve high precision in time detection, the time errors of signal start and end are relatively large. Therefore, to address the requirement for high precision in signal detection start and end times, a signal start and end detection program was designed. This program aims to achieve accurate determination of signal start and end positions through signal detection within a relatively small frequency and time domain range, thus better serving wireless signal detection technology.

[0005] The technical solution of this invention is characterized by including the following steps:

[0006] Step 1: Calculate the Short-Time Fourier Transform (SFT) of the signal. The spectral position occupied by each signal is determined by its SFT plot. After inputting a signal and determining a suitable sampling rate, calculate the SFT over a period of time, resulting in SFT values ​​at discrete time and frequency points (time and frequency discrete grids). Plot the SFT of this signal, taking the dB values ​​for plotting. The program extracts frequency slices of the signal in pairs, with each pair of frequencies exceeding the noise floor threshold considered as a valid signal, and performs frequency domain detection on each valid signal.

[0007] Step 2: Calculate the frequency profile. Record the maximum value of all time points at a certain frequency point in the STRIF plot as the profile value of that frequency point. All the profile values ​​of the frequency points in the frequency domain form the frequency profile of the short-time Fourier transform image of the signal under test. Convert the short-time Fourier transform image of the frequency point into a one-dimensional frequency profile to improve the frequency accuracy in a single frequency domain.

[0008] Step 3: Using the frequency resolution calculated in Step 1, determine whether the recorded frequency profile exceeds the noise floor threshold. If it does, record it as a valid frequency point. Record the start and end frequencies of the exceeding portion as a frequency pair, which can be used as the start and end points of the signal frequency. If the start and end frequencies are found to be isolated points, they can be ignored.

[0009] Step 4: Analyze each SFT sub-image recorded on the short-time Fourier transform image of the signal under test. Record the maximum value of each frequency point at each time point to form a time profile.

[0010] Step 5: Analyze and statistically process the recorded time profile data. Detect signal presence using the range of start and end frequencies, time segments, and duration. Record the start and end times of segments exceeding the noise floor threshold. Calculate the time pairs and time segments, and record the maximum signal-to-noise ratio (SNR) within each time segment as the SNR of the signal. Isolate time points are not ignored. This time pair can be used as the start and end times of a valid signal.

[0011] Step Six: In signal presence detection, a signal presence detection rule is formed using the range of start and end frequencies, the number of time segments, and the duration of the time segment. This rule is based on the stringency requirements for signal presence judgment. The product of the range of start and end frequencies and the number of time segments must be greater than threshold 1, and the duration of the maximum time segment must be greater than threshold 2. Threshold 1 and threshold 2 are determined according to the stringency requirements for signal detection.

[0012] Step 7: At this point, the start and end positions of the signal in the frequency and time domains within a small range can be obtained, which are the start and end positions of the frequency and time of the signal under test with high accuracy. Attached Figure Description

[0013] Figure 1 This is a flowchart of the algorithm of the present invention.

[0014] Figure 2 This is the short-time Fourier transform (STFT) image of the signal in this invention.

[0015] Figure 3 This is a frequency cross-sectional view of the frequency band for frequency domain detection in this invention.

[0016] Figure 4 This is an example diagram of the frequency domain detection frequency chip SFT of the present invention. Detailed Implementation

[0017] Because the original frequency and time ranges are difficult to achieve high time detection accuracy, the time errors of signal start and end are relatively large. Therefore, to address the requirement for high accuracy in signal detection start and end times, a signal start and end detection program was designed. This program aims to achieve high accuracy in signal start and end positions through signal detection within a relatively small frequency and time domain range, thus better serving wireless signal detection technology.

[0018] Step 1: Calculate the short-time Fourier transform of the signal, such as... Figure 2 Example. The spectral and temporal positions occupied by each signal are determined by its SFT plot. Set an appropriate fixed segment duration, read the sampling rate of the signal under test, and determine the frequency resolution as the ratio of the sampling rate to the number of sampling points, i.e., F = fs / N; determine the temporal resolution as one-quarter of the window length, i.e., hop = wlen / 4. Use the Short-Time Fourier Transform (SFT) to plot the SFT plot of this signal.

[0019] The program extracts the frequency of the signal in pairs with a step size of one frequency grid. Each pair of frequencies is temporarily considered as a valid signal, and frequency domain detection is performed on each valid signal.

[0020] Name the file according to the occupancy of different frequency bands of the signal, and obtain the basic information of the signal under test, including the sampling rate, signal frequency range, etc.

[0021] Step 2: Calculate the frequency profile, such as... Figure 3 As shown, the maximum value at all times of a certain frequency point in the STRIF plot is recorded as the profile value of that frequency point. All profile values ​​of the frequency points in the frequency domain form the frequency profile of the short-time Fourier transform image of the signal under test, thus converting the short-time Fourier transform image of that frequency point into a one-dimensional frequency profile.

[0022] Step 3: Using the frequency resolution calculated in Step 1, check whether the recorded frequency profile exceeds the noise floor threshold. If it exceeds the noise floor threshold, record it as a valid frequency point. Record the start and end frequencies of the exceeding portion as a frequency pair, which can be used as the start and end points of the signal frequency. If the start and end frequencies are found to be isolated points, they can be ignored.

[0023] Step 4: Analyze the various SFT sub-graphs in the recorded frequency slices of the short-time Fourier transform image of the signal under test. For a sub-graph corresponding to a certain frequency slice, record the maximum value at each frequency point at each time point to form a time profile.

[0024] Step 5: For each time profile, check whether it exceeds the noise floor threshold. Record the start and end times of each time pair exceeding the noise floor threshold as a time segment. Count the number of start and end time pairs, i.e., the number of time segments. Record the maximum signal-to-noise ratio (SNR) within each time segment as the SNR of the signal. Isolation points are not ignored. An example of a time segment is shown below. Figure 4 .

[0025] Step Six: At this point, the start and end positions of the candidate signal in the frequency and time domains can be obtained, i.e., the start and end positions of the frequency and time of the signal under test with high accuracy. After comparison with the STRIF diagram of the signal, it is found that it can reflect the start and end positions of the effective signal with high accuracy.

[0026] However, useful signals still need further evaluation.

[0027] The presence of a signal is detected by using the start and end frequencies, time profile, and time slack as criteria. If the start and end frequency range Δf and the time slack N are related... t The product is greater than the threshold T1, i.e., Δf*N t >T1, maximum time duration t m Greater than the threshold T2, i.e., t m If the value is greater than T2, then the signal can be considered to exist.

[0028] Case Study:

[0029] This invention uses real-time signals from the Beijing area as experimental data for detection. The signal numbered JCF120 was selected; its filename contains information such as the sampling rate and frequency range. The basic signal data was read and imported using MATLAB software. The duration of each segment was set to 0.02 seconds, and the sampling rate was set to 153.6 eΩ. 6 Hz, total file size is 7.5e 9 The number of bytes selected for processing, starting from the beginning of the file, is the product of the signal duration and the sampling rate. The receive buffer size is 2 bytes. * sampling rate fs * Segment duration length.

[0030] Step 1: Based on the above data, plot the short-time Fourier transform (SFT) of the JCF120 signal. The spectral position occupied by each signal is determined by its SFT plot. Since different signal intensities result in different colors and intensities, set an appropriate fixed segment duration. Read the sampling rate of the signal under test. Determine the frequency resolution based on the sampling rate as the ratio of the sampling rate to the number of sampling points, i.e., fres = fs / N; determine the time resolution as one-quarter of the window length, i.e., hop = wlen / 4. The program extracts the signal frequencies in pairs with a step size of one frequency grid. Each pair of frequencies is temporarily considered a valid signal, and frequency domain detection is performed on each valid signal. Name the file according to the different frequency band occupancy of the signal to obtain basic information about the signal under test, including the sampling rate and signal frequency range.

[0031] Step 2: Record the maximum value of all time points in the single frequency domain as the maximum value of that frequency, and convert the short-time Fourier transform image of the signal under test into a frequency profile.

[0032] Using the frequency resolution calculated in step one, the portion exceeding the noise floor threshold is taken as a frequency profile. All profile values ​​of the frequency points in the frequency domain form the frequency profile of the short-time Fourier transform image of the signal under test, and the short-time Fourier transform image of the frequency points is converted into a one-dimensional frequency profile.

[0033] Step 3: Using the frequency resolution calculated in Step 1, check whether the recorded frequency profile exceeds the noise floor threshold. If it exceeds the noise floor threshold, record it as a valid frequency point. Record the start and end frequencies of the exceeding portion as a frequency pair, which can be used as the start and end points of the signal frequency. If the start and end frequencies are found to be isolated points, they can be ignored.

[0034] In this embodiment, the noise floor threshold is estimated to be 82.111.

[0035] Step 4: Analyze the various SFT sub-graphs in the recorded frequency slices of the short-time Fourier transform image of the signal under test. For a sub-graph corresponding to a certain frequency slice, record the maximum value at each frequency point at each time point to form a time profile.

[0036] Step 5: For each time profile, check whether it exceeds the noise floor threshold. The time pairs exceeding the noise floor threshold are considered as one time segment, and their start and end times are recorded. The number of start and end time pairs is counted, i.e., the number of time segments. The maximum signal-to-noise ratio (SNR) within each time segment is recorded as the SNR of the signal. Isolated time points are not ignored. Example time segments are shown in Table 1. The example shows three segments detected, with an SNR of 3dB.

[0037] Table 1. Examples of time domain time truncation, time truncation number, frequency start and end, and maximum duration.

[0038] SIGN Frequency Grids Maximum duration Time cutoff Detection frequency s Detection frequency e Detection frequency s Detection frequency e Maximum strength Window noise 3 36 36 3 16302 16338 35493752 35524432 89.92269 82.11101

[0039] Step Six: At this point, the start and end positions of the candidate signal in the frequency and time domains can be obtained, i.e., the start and end positions of the frequency and time of the signal under test with high accuracy. After comparison with the STRIF diagram of the signal, it is found that it can reflect the start and end positions of the effective signal with high accuracy.

[0040] However, useful signals still need further evaluation. If the number of frequency grids occupied by the current frequency point is not less than 16 and the maximum number of frequency grids is not less than 20, and the signal-to-noise ratio is greater than 3dB, then the signal at that time is considered valid and marked as 3.

[0041] Table 2 shows the correspondence between the number of discrete-time grids and physical time (in seconds). The time of the first sampling point is considered to be 0 seconds. (This is written after the concept of discrete time).

[0042] Table 2. Correspondence between discrete-time grid numbers and physical time (unit: seconds)

[0043]

[0044] Table 3 shows the signal detection results obtained using the method of this patent. A total of 158 signals were detected in this file; the information for the first 19 signals is shown in the table below.

[0045] The starting frequency position of the valid signal is the starting number of the detection frequency interval (s) and the detection frequency interval (e). From this, the number of frequency grids occupied by the signal can be calculated, which is the ending number of the starting frequency position of the valid signal minus the starting number of the starting frequency position of the valid signal.

[0046] 'Maximum duration' is calculated by subtracting the start and end positions of each time segment for each valid signal; the recorded value here is the difference in the number of discrete time segments.

[0047] 'Maximum strength' is the maximum value of the effective signal strength.

[0048] 'Window noise' is the noise threshold value, which is estimated. Since the noise threshold value of the JCF120 signal changes relatively flat, the noise threshold value does not fluctuate much.

[0049] The 'detection time interval' consists of several time segments exceeding the noise threshold. In this example, if the detected time segment is [1, 65], it can be considered that each detected signal exists continuously from beginning to end without interruption. The maximum number of time segments in this embodiment is 65. Through the correspondence shown in Table 2, the time segment positions can be converted into corresponding physical times.

[0050] In summary, this patent can achieve the effect of determining the start and end frequencies and start and end times of a signal.

[0051] Table 3. Examples of relevant parameters for effective signals

[0052]

Claims

1. A method for determining the start and end times and start and end frequencies of a signal based on time-frequency analysis, characterized in that... Includes the following steps: Step 1: In the method for determining the start and end times and frequencies of signals based on time-frequency analysis, the spectral position occupied by each signal is determined by the short-time Fourier transform (STFT) image of that signal. After plotting the STFT image of this signal, each valid signal is found, and frequency domain detection is performed on each valid signal. Step 2: Calculate the frequency profile to form the frequency profile of the short-time Fourier transform image of the signal under test, and convert the short-time Fourier transform image of the frequency point into a one-dimensional frequency profile. Step 3: Detect whether the recorded frequency profile exceeds the noise floor threshold to determine the start and end points of the usable signal frequency. Step 4: Detect the portions of the extracted frequency and time pairs that exceed the noise floor threshold, and record the start and end frequencies and time positions of these excess portions as the start and end positions of the signal. This can significantly improve the accuracy of signal detection. Step 5: Detect and statistically analyze the recorded time profile data, and use the range of start and end frequencies, time segments, and duration to detect the presence of the signal. Step Six: By establishing signal presence detection rules based on the range of start and end frequencies, time truncation, and duration, and adhering to strict requirements for signal presence judgment, the start and end positions of the signal in the time and frequency domains within a small range can be obtained, thus achieving high-precision determination of the signal start and end times and frequencies. Step 7: At this point, the start and end positions of the signal in the frequency and time domains within a small range can be obtained, which are the start and end positions of the frequency and time of the signal under test with high accuracy.

2. The method for determining the start and end times and start and end frequencies of a signal based on time-frequency analysis according to claim 1, characterized in that: In step one, the effective signal is determined by the short-time Fourier transform of the signal. The method is as follows: after inputting a signal and determining an appropriate sampling rate, the short-time Fourier transform is used to draw the STRET diagram of the signal; the signal is extracted in pairs by the program with frequency slices exceeding the noise floor threshold as a group, and each pair of frequencies is temporarily regarded as an effective signal, and frequency domain detection is performed on each effective signal.

3. The method for determining the start and end times and start and end frequencies of a signal based on time-frequency analysis according to claim 1, characterized in that: The method for calculating the frequency profile in step two is as follows: Using a signal at a certain resolution, the portion exceeding the noise floor threshold is considered as a frequency slice. The maximum value over all time intervals at a given frequency point is recorded as the profile value for that frequency point. All profile values ​​of the frequency points within the frequency domain form the frequency profile of the short-time Fourier transform image of the signal under test. Converting the short-time Fourier transform image of this frequency point into a one-dimensional frequency profile improves the frequency accuracy in a single frequency domain.

4. The method for determining the start and end times and start and end frequencies of a signal based on time-frequency analysis according to claim 1, characterized in that: In step three, the start and end points of the usable signal frequency are determined by: using the frequency resolution calculated in step one, the recorded frequency profile is checked to see if it exceeds the noise floor threshold. If it exceeds the threshold, it is recorded as a valid frequency point. The start and end frequencies of the exceeding portion are recorded as a frequency pair, which can be used as the start and end points of the signal frequency. If the start and end frequencies are found to be isolated points, they can be ignored.

5. The method for determining the start and end times and start and end frequencies of a signal based on time-frequency analysis according to claim 1, characterized in that: In step four, the method for determining the maximum time occupied by the signal on the time axis is as follows: Analyze each SFT sub-graph of the recorded frequency slices on the short-time Fourier transform image of the signal under test. For a sub-graph corresponding to a certain frequency slice, record the maximum value at each frequency point at each time point to form a time profile.

6. The method for determining the start and end times and start and end frequencies of a signal based on time-frequency analysis according to claim 1, characterized in that: The method for detecting the existence of the signal in step five is as follows: For each time profile, detect whether the noise floor threshold is exceeded. Record the time pairs exceeding the noise floor threshold as a time segment, and count the start and end time pairs and the number of time segments. Record the maximum signal-to-noise ratio (SNR) within each time segment as the SNR of the signal. Isolation points are not ignored. This time pair can be used as the start and end times of a valid signal.

7. The method for determining the start and end times and start and end frequencies of a signal based on time-frequency analysis according to claim 1, characterized in that: The signal existence detection in step six is ​​performed by using the start and end frequencies, time profile, and time truncation as criteria to detect the signal's existence. If the start and end frequency range Δf and the time truncation N... t The product is greater than the threshold T1, i.e., Δf×N t >T1, maximum duration t m Greater than the threshold T2, i.e., t m If the value is >T2, then the signal can be considered to exist.

Images