Bandwidth identification method for multipath signals

Abstract

The invention discloses a bandwidth identification method for multiple paths of signals, belongs to the technical field of wireless communication, and particularly relates to the bandwidth identification method for the multiple paths of signals. Comprising the following steps: performing time-frequency domain transformation on a received baseband signal to obtain a power spectrum of the signal; performing down-sampling to extract an upper envelope of the power spectrum; estimating the average power of the noise in the sampling bandwidth, and segmenting the signal and the noise; uniformly dividing the effective bandwidth into a plurality of parts, and counting the energy of the de-noised signal in each part of bandwidth to obtain the energy distribution of the signal; and estimating the bandwidth of each path of signal according to the energy distribution, and calculating the carrier frequency and the signal-to-noise ratio. The method is suitable for multi-path bandwidth identification of various types of signals such as WiFi signals, mobile communication signals and frequency modulation broadcast signals, and the center carrier frequency and the signal-to-noise ratio of each path of signal are estimated.

Description

Technical Field

[0001] The invention belongs to the technical field of wireless communications and is a bandwidth identification method for multi-channel signals. Background Art

[0002] In modern communication systems, signal bandwidth is one of the important indicators for evaluating signal transmission capabilities. Signal bandwidth is usually defined as the frequency range occupied by the part of the signal spectrum that contains the main energy. In communication systems, signal bandwidth determines the signal transmission rate, anti-interference ability and spectrum efficiency. For example, in wireless communications, signal bandwidth directly affects the capacity and coverage of the system; in wired communications, signal bandwidth determines the rate and quality of data transmission.

[0003] Accurately identifying signal bandwidth is of great significance for optimizing communication system performance, achieving interference suppression, and analyzing opportunity signals. By identifying the signal bandwidth, appropriate filters can be designed to reduce noise and interference, thereby improving the signal-to-noise ratio and communication quality. At the receiving end of the communication system, accurately identifying the signal bandwidth can help design a suitable demodulator to accurately restore the original information, which is crucial for correctly demodulating the signal. For fields such as wireless sensing that require the use of opportunity signals, identifying the signal bandwidth can help determine the location of the opportunity signal band and provide prior information for subsequent demodulation and perception. In the field of WiFi signal analysis, accurately analyzing how many signals there are in the space and their bandwidth size is the prerequisite for further analysis of other parameters.

[0004] The traditional signal bandwidth estimation method is the template analysis method based on the power spectrum. This method estimates the bandwidth based on the power spectrum obtained after the signal is transformed in the time-frequency domain. By roughly estimating the bandwidth of the signal and adjusting the length of the window function during the transformation process, a relatively smooth power spectrum can be obtained. On this basis, a zero-crossing search is performed to obtain the bandwidth estimation result of the signal. This method is simple and easy to implement, but it is difficult to identify signals with complex spectra and signals with complex noise backgrounds, and it cannot be identified when there are multiple signals in the spectrum.

[0005] Signal bandwidth identification based on deep learning is a new method that has emerged in recent years. This method uses deep neural networks to automatically learn and extract spectral features from a large amount of data to detect signal bandwidth. Compared with traditional methods, methods based on deep learning can adapt to signals of different types and bandwidths without manual parameter adjustment, and are more robust to noise and interference. However, this method requires a large amount of labeled data for training, and the cost of data acquisition and labeling is high. On the other hand, the internal working mechanism of the deep learning model is relatively complex, lacks interpretability, and has poor generalization ability for new data outside the training set. Summary of the invention

[0006] In an actual signal receiving scenario, the received signal spectrum contains several signals, and the bandwidth of each signal is uncertain. Therefore, the present invention proposes a bandwidth identification method for multiple signals, which can detect the bandwidth of each signal within the received signal spectrum in real time and estimate their respective carrier frequencies and signal-to-noise ratios.

[0007] The bandwidth identification method of the multi-channel signal of the present invention comprises the following specific steps:

[0008] Step 1: Perform a time-frequency domain transformation on the received baseband signal to obtain the power spectrum of the signal, wherein the time-frequency domain transformation adopts a fast Fourier transform.

[0009] Step 2: If the power spectrum can be downsampled, extract the upper envelope of the power spectrum by downsampling, and then execute step 3; if downsampling is not possible, directly execute step 3. Whether the power spectrum is downsampled depends on the comparison result between the original resolution of the power spectrum and the pre-set bandwidth estimation accuracy.

[0010] Step 3: Estimate the noise power within the sampling bandwidth by plotting the probability density function of the signal power spectrum and the cumulative distribution function to find the intersection point, and separate the signal and the noise. To enhance the stability of the recognition results, linear coordinates are required to calculate the cumulative distribution function.

[0011] Step 4: Divide the effective bandwidth evenly into N part For each bandwidth, the pure signal power spectrum within the bandwidth is summed to obtain the power value of each bandwidth, and a sliding average operation is performed to obtain a smooth signal energy distribution. Signals with different bandwidths use different sizes of N part value to obtain a more accurate bandwidth estimate.

[0012] Step 5: Estimate the bandwidth of each signal based on the energy distribution, and calculate the carrier frequency and signal-to-noise ratio. The edge search method is used to estimate the bandwidth.

[0013] The advantages of the present invention are:

[0014] 1. The bandwidth identification method of multi-channel signals of the present invention uses the probability density function and cumulative distribution function of the signal power spectrum to find the intersection point to determine the segmentation point between the signal and the background noise, which can obtain a more accurate noise power estimation and improve the bandwidth identification performance.

[0015] 2. The bandwidth identification method of multiple signals of the present invention can effectively reduce the interference of noise on detection by cutting the background noise and counting the signal energy of each unit interval within the effective bandwidth. It can also process more complex background noise, thereby improving the robustness of bandwidth identification.

[0016] 3. The bandwidth identification method of multiple signals of the present invention processes data by extracting the envelope on the power spectrum and performing downsampling, which greatly reduces the calculation amount of the background noise segmentation and bandwidth estimation steps and improves the real-time performance of bandwidth identification.

[0017] 4. Compared with traditional bandwidth identification, the bandwidth identification method of multiple signals of the present invention can simultaneously identify the bandwidths of multiple signals, and can also accurately identify the bandwidths of multiple signals when there is a large difference in the signal-to-noise ratio of each signal, and has a wider range of application scenarios. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Figure 1 This is an overall flow chart of the multi-channel signal bandwidth identification method of the present invention.

[0019] Figure 2 It is the original power spectrum and the upper envelope power spectrum.

[0020] Figure 3 The intersection of PDF and CDF is used to find the split point between signal and noise.

[0021] Figure 4 This is a comparison chart of the noise power estimate and the signal power spectrum.

[0022] Figure 5 This is the signal energy distribution and bandwidth detection diagram.

[0023] Figure 6 This is a diagram showing the recognition results of 5G WiFi signals using the bandwidth recognition method for multi-channel signals of the present invention.

[0024] Figure 7 The figure is a recognition result diagram of FM broadcasting signals using the bandwidth recognition method for multi-channel signals of the present invention.

[0025] Figure 8 The present invention is a recognition result diagram of a certain test signal using the bandwidth recognition method for multi-channel signals of the present invention. DETAILED DESCRIPTION

[0026] The present invention will be further described in detail below in conjunction with the accompanying drawings.

[0027] Traditional signal bandwidth identification methods can only identify the bandwidth of a single signal. One of the reasons is that the bandwidth identification process is easily affected by noise. In actual receiving scenarios, the signal-to-noise ratio may decrease as the channel environment changes, thereby affecting the algorithm's identification of the number of signals and their bandwidth. The bandwidth identification method for multi-channel signals proposed in the present invention can achieve signal-to-noise segmentation by estimating noise power, minimizing the impact of noise on bandwidth identification. On this basis, by statistically analyzing the signal energy of each interval within the effective bandwidth, a smooth signal energy distribution can be obtained, and the bandwidth of multi-channel signals can be identified by edge detection.

[0028] The bandwidth identification method of the multi-channel signal of the present invention is as follows: Figure 1 As shown, the specific steps are:

[0029] Step 1: Perform a time-frequency domain transform on the received baseband signal (original signal) to obtain the power spectrum of the signal.

[0030] The received baseband signal is first transformed into the time-frequency domain through Fast Fourier Transformation (FFT); then the spectrum shift fftshift operation is performed to move the second half of the FFT result (corresponding to negative frequencies) to the front and the first half (corresponding to positive frequencies) to the back, so as to rearrange the frequency axis so that the frequency axis changes continuously from negative to positive, which is convenient for further analysis. On this basis, the logarithmic coordinates are converted, and the modulus value of the fftshift result is taken as the logarithm with the base 10 and multiplied by 10 to obtain the power spectrum of the baseband signal.

[0031] Step 2: Downsampling is performed to extract the upper envelope of the power spectrum.

[0032] Before downsampling the power spectrum to extract the upper envelope, the downsampling factor needs to be calculated. First, the original resolution R of the signal power spectrum is calculated. 1 for:

[0033]

[0034] Among them, f s is the original signal sampling rate, N 1 is the number of power spectrum signal points.

[0035] Then, the original resolution R 1 The bandwidth estimation accuracy R is set to 0 (set to 1kHZ) for comparison, if R 0 Greater than R 1 , the downsampling factor M can be calculated as:

[0036]

[0037] The symbol Indicates rounding down.

[0038] On this basis, the power spectrum can be downsampled to extract the upper envelope. In the original power spectrum, every M points are divided into a group in order, and the maximum value is extracted from each group to form a new power spectrum, which is an M-fold downsampled spectrum and the upper envelope of the original power spectrum. Since the power spectrum resolution changes after downsampling, the power spectrum resolution after downsampling needs to be calculated as:

[0039] R2 =M·R 1 (3)

[0040] The comparison between the envelope of the power spectrum extracted by downsampling and the original power spectrum is as follows: Figure 2 As shown (the signal analyzed in the embodiment of the present invention is an actual WiFi signal in the 5G frequency band, with a frequency of 5200 MHz and a sampling rate of 122.88 MHz).

[0041] If R 0 Less than R 1 , no downsampling is performed and the power spectrum resolution remains unchanged.

[0042] After performing sliding averaging on the power spectrum after downsampling or the power spectrum without downsampling (the sliding window length is 11), the noise power estimation of step 3 is performed.

[0043] Step 3: Estimate the noise power within the sampling bandwidth and separate the signal from the noise.

[0044] To facilitate subsequent processing, the minimum value of the power spectrum (downsampled or not downsampled) is adjusted first, specifically: the minimum value of the power spectrum is calculated, and then the minimum value is subtracted from the entire power spectrum to complete the adjustment so that the minimum value of the power spectrum is zero.

[0045] Before estimating the noise power, it is necessary to plot the probability density function (PDF) and cumulative distribution function (CDF) of the power spectrum after minimum zeroing. The probability density function uses logarithmic coordinates, while the cumulative distribution function uses linear coordinates.

[0046] When drawing the probability density function, the power spectrum after the minimum value is zeroed is divided into N according to the amplitude from low to high. bins Equal parts (N bins is 1000), the width of each interval is:

[0047]

[0048] Among them, P max With P min are the maximum and minimum values ​​of the power spectrum, P min = 0. Find the probability density of each interval. For the ith interval, the probability density function is calculated as follows:

[0049]

[0050] Among them, n iis the number of times the power spectrum appears in the i-th amplitude interval, N 2 is the number of power spectrum signal points after downsampling. The obtained result is windowed with a length of 1 / W bin The sliding average of is the final signal power spectrum probability density distribution.

[0051] On this basis, we further draw the cumulative distribution function. First, we transform N bins The central power spectrum amplitude of the interval is converted from logarithmic coordinates to linear coordinates. The conversion formula is:

[0052]

[0053] Among them, Y i is the central power spectrum amplitude of the logarithmic coordinates of the i-th interval, X i is the corresponding linear power spectrum amplitude. Then the power spectrum probability density distribution is weighted by the power spectrum of the linear coordinates, and the cumulative sum is calculated and normalized to obtain the final signal power spectrum cumulative distribution. The calculation formula is:

[0054]

[0055] Among them, CDF i is the cumulative distribution value of the signal power spectrum in the i-th interval, X i is the central power spectrum amplitude of the i-th interval linear coordinate, PDF i is the probability density value of the ith interval.

[0056] Plot the probability density function and cumulative distribution function of the signal power spectrum on the same graph, and find the intersection of the two functions. When there are more than two intersections, find the rightmost intersection as the split point. The horizontal coordinate corresponding to the split point is the estimated noise power (in dB), such as Figure 3 The position of the noise power corresponding to the split point in the power spectrum is shown as Figure 4 shown.

[0057] Further, the power spectrum obtained in step 2 is subtracted from the noise power estimation value determined above. At this time, the negative values ​​in the power spectrum represent noise, and the positive values ​​represent signals. All positive values ​​are extracted to form a new pure signal power spectrum, which can achieve the separation of signal and noise.

[0058] Step 4: Divide the effective bandwidth evenly into N part The energy of the denoised signal in each bandwidth is counted to obtain the energy distribution of the signal.

[0059] The effective bandwidth is defined as the bandwidth occupied by the pure signal power spectrum in step 3. For each bandwidth, the pure signal power spectrum within the bandwidth is summed to obtain the power value of each bandwidth, and a sliding average operation (sliding window length is 3) is performed to obtain a smoother signal energy distribution.

[0060] For signals with wider bandwidth, N part As in this embodiment, it can be set to 100. For narrowband signals such as FM broadcast signals, N can be appropriately increased. part value to obtain a more accurate bandwidth estimate.

[0061] Step 5: Estimate the bandwidth of each signal based on the energy distribution, and calculate the carrier frequency and signal-to-noise ratio.

[0062] Draw the signal energy distribution diagram obtained in step 4, and draw a detection threshold line at the bottom of the signal energy distribution diagram (the threshold value set in the embodiment of the present invention is 1 / 2N part ), the left and right cutting points of the power spectrum of each signal can be used to determine the bandwidth boundary of each signal, such as Figure 5 As shown. When determining the cutting point, the search method can be used to implement it. For the signal energy distribution, from left to right, first look for the left rising edge greater than the threshold, then look for the right falling edge less than the threshold, and search the entire effective bandwidth alternately from left to right until the right boundary of the effective bandwidth is reached.

[0063] Record the frequency of each pair of left rising edge and right falling edge, which is the power spectrum boundary of each signal. The bandwidth of each signal is the right falling edge frequency minus the left rising edge frequency. The midpoint of the right falling edge frequency and the left rising edge frequency is the carrier frequency of each signal. The average value of all frequency energy between the right falling edge and the left rising edge is the signal-to-noise ratio of each signal. The final bandwidth identification result is as follows: Figure 6 shown.

[0064] The bandwidth identification results for FM broadcast signals and a test signal are as follows: Figure 7 and Figure 8 As shown. Among them, the frequency of the FM broadcast signal is 88.7MHz, and the sampling rate is 10MHz; the frequency of a certain test signal is 500MHz, and the sampling rate is 10MHz. It can be seen from the figure that the bandwidth identification method of the multi-channel signal of the present invention can accurately identify the bandwidth of the multi-channel signal, and can also accurately identify the multi-channel signal with a more complex background noise.

Claims

1. A bandwidth identification method for multi-channel signals, characterized in that: The specific steps are: Step 1: Perform a time-frequency domain transformation on the received baseband signal to obtain the power spectrum of the signal, wherein the time-frequency domain transformation adopts a fast Fourier transform; Step 2: If the power spectrum can be downsampled, extract the upper envelope of the power spectrum by downsampling, and then execute step 3; if downsampling is not possible, directly execute step 3; whether the power spectrum is downsampled depends on the comparison result between the original resolution of the power spectrum and the pre-set bandwidth estimation accuracy; Step 3: Estimate the noise power within the sampling bandwidth by plotting the probability density function of the signal power spectrum and the cumulative distribution function to find the intersection point, and separate the signal and the noise; in order to enhance the stability of the recognition result, linear coordinates are required to calculate the cumulative distribution function; Step 4: Divide the effective bandwidth evenly into N part For each bandwidth, the pure signal power spectrum within the bandwidth is summed to obtain the power value of each bandwidth, and a sliding average operation is performed to obtain a smooth signal energy distribution; signals with different bandwidths use different sizes of N part value to obtain a more accurate bandwidth estimate. Step 5: Estimate the bandwidth of each signal based on the energy distribution, and calculate the carrier frequency and signal-to-noise ratio; the edge search method is used to estimate the bandwidth.

2. A bandwidth identification method for multi-channel signals as claimed in claim 1, characterized in that: In step 1, firstly, the time-frequency domain is transformed by fast Fourier transform; then the spectrum shift fftshift operation is performed to make the frequency axis change continuously from negative to positive; further, the modulus value of the fftshift result is taken as the base 10 logarithm and multiplied by 10 to obtain the power spectrum of the baseband signal.

3. A bandwidth identification method for multi-channel signals as claimed in claim 1, characterized in that: In step 2, the original resolution R1 is compared with the preset bandwidth estimation accuracy R0: If R0 is greater than R1, the downsampling factor is calculated The power spectrum is downsampled to extract the upper envelope. After downsampling, the power spectrum resolution is R2 = M·R1; If R0 is smaller than R1, no downsampling is performed and the power spectrum resolution remains unchanged.

4. A bandwidth identification method for multi-channel signals as claimed in claim 3, characterized in that: Native resolution f s is the original signal sampling rate, and N1 is the number of power spectrum signal points.

5. A bandwidth identification method for multi-channel signals as claimed in claim 1, characterized in that: In step 3, when drawing the probability density function, the power spectrum is divided into N parts from low to high according to the amplitude. bins Equal parts, the width of each interval is: Among them, P max With P min are the maximum and minimum values ​​of the power spectrum respectively; find the probability density of each interval, and for the i-th interval, the probability density function calculation formula is: Among them, n i is the number of occurrences of the power spectrum in the i-th amplitude interval, and N2 is the number of power spectrum signal points after downsampling; The result is windowed with a length of 1 / W bin The sliding average of is the final signal power spectrum probability density distribution; When plotting the cumulative distribution function, first N bins The central power spectrum amplitude of the interval is converted from logarithmic coordinates to linear coordinates. The conversion formula is: Among them, Y i is the central power spectrum amplitude of the logarithmic coordinates of the i-th interval, X i is the corresponding linear power spectrum amplitude. Then the power spectrum probability density distribution is weighted by the power spectrum of the linear coordinates, and the cumulative sum is calculated and normalized to obtain the final signal power spectrum cumulative distribution. The calculation formula is: Among them, CDF i is the cumulative distribution value of the signal power spectrum in the i-th interval, X i is the central power spectrum amplitude of the i-th interval linear coordinate, PDF i is the probability density value of the ith interval.

6. The bandwidth identification method of a multi-channel signal according to claim 1, characterized in that: In step 3, the signal and noise segmentation method is: Plot the probability density function and cumulative distribution function of the signal power spectrum on the same graph, find the intersection of the two functions, and when there are more than two intersections, find the rightmost intersection as the split point. The horizontal coordinate corresponding to the split point is the estimated noise power. Further, subtract the noise power estimate determined above from the power spectrum obtained in step 2. At this time, the negative values ​​in the power spectrum represent noise, and the positive values ​​represent signals. Extract all positive values ​​to form a new pure signal power spectrum to achieve the segmentation of signal and noise.

7. A bandwidth identification method for multi-channel signals as claimed in claim 1, characterized in that: In step 3, before plotting the probability density function and cumulative distribution function of the signal power spectrum, the minimum value of the power spectrum is adjusted. Specifically, the minimum value of the power spectrum is first calculated, and then the minimum value is subtracted from the entire power spectrum to complete the adjustment so that the minimum value of the power spectrum is zero.

8. A bandwidth identification method for multi-channel signals as claimed in claim 1, characterized in that: In step 4, for signals with wider bandwidth, N part As in this embodiment, it can be set to 100. For narrowband signals such as FM broadcast signals, N can be appropriately increased. part value to obtain a more accurate bandwidth estimate.

9. A bandwidth identification method for multi-channel signals as claimed in claim 1, characterized in that: The method for detecting the bandwidth of each signal is to search the entire effective bandwidth alternately from left to right for the signal energy distribution, first looking for the left rising edge greater than the threshold, then looking for the right falling edge less than the threshold, until reaching the right boundary of the effective bandwidth; record the frequency of each pair of left rising edge and right falling edge, which is the power spectrum boundary of each signal. The bandwidth of each signal is the right falling edge frequency minus the left rising edge frequency, the midpoint of the right falling edge frequency and the left rising edge frequency is the carrier frequency of each signal, and the average value of the energy of all frequency points between the right falling edge and the left rising edge is the signal-to-noise ratio of each signal.

Images