Blind signal detection method based on finite-state machine
Through the blind signal detection method based on finite state machine, the difficulty of signal detection under complex electromagnetic environment and noise interference is solved, and efficient and accurate signal detection is achieved, which is suitable for a variety of signal and noise conditions.
Patent Information
- Application Number
- CN202510172700.7
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-02-17
- Publication Date
- 2025-05-27
AI Technical Summary
Under complex electromagnetic environments and noise interference factors, traditional signal detection methods are difficult to achieve efficient and accurate blind signal detection, especially when unknown signals and prior information are difficult to obtain.
A blind signal detection method based on a finite state machine is adopted, and signals are collected and segmented, spectrum analysis and power spectral density calculation are performed, and the signal and noise are distinguished and judged based on the judgment threshold value and the finite state machine model.
This method can effectively suppress noise without prior information, improve the accuracy and reliability of signal detection, and is suitable for various signal and noise conditions.
Smart Images
Figure CN120050139A_ABST
Abstract
Description
Technical Field
[0001] The invention belongs to the technical field of wireless communications, and in particular relates to a blind signal detection method based on a finite state machine in this field. Background Art
[0002] In the current field of communication technology, with the increasing complexity of the electromagnetic environment and the diversification of noise interference factors, signal detection faces unprecedented challenges. Especially in applications such as electronic reconnaissance, wireless communication and radar systems, the received signals are often transmitted in complex electromagnetic environments. These environments may include electromagnetic interference of various frequencies, multipath effects, non-stationary noise and interference from non-cooperative signals, which cause signal distortion, attenuation and distortion during transmission. This not only increases the difficulty of signal detection, but also puts higher requirements on the accuracy and reliability of signal detection.
[0003] Although traditional signal detection methods have good detection performance under certain specific conditions, they usually require prior knowledge of the signal as support. However, in practical applications, especially in scenarios such as electronic reconnaissance and non-cooperative communication, the signals to be detected are often unknown, and the prior information of these signals is often impossible to obtain or difficult to accurately estimate, which greatly limits the application of traditional signal detection methods. In order to meet this challenge, researchers began to seek blind signal detection methods that do not require prior information. Existing blind signal detection methods, such as energy detection method, delay correlation method and conjugate symmetric sequence multiplication method, can achieve signal detection without prior information to a certain extent, but they each have obvious shortcomings. The energy detection method performs well under the background of Gaussian white noise, but its detection performance will drop significantly when the background noise is much larger than the signal or the background noise is non-stationary; the delay correlation method has poor detection performance in multipath and low signal-to-noise ratio environments; and the conjugate symmetric sequence multiplication method, although it makes up for the shortcomings of the delay correlation method to a certain extent, its calculation amount is very large and not practical. In summary, how to overcome noise and achieve blind signal detection in a complex electromagnetic environment and under noise interference factors has become an important issue that needs to be urgently solved in the current field of communication technology. Summary of the invention
[0004] The technical problem to be solved by the present invention is to provide a blind signal detection method based on a finite state machine. The method can effectively cope with complex electromagnetic environments and noise interference factors without prior information, and realize efficient and accurate signal detection.
[0005] The present invention adopts the following technical solution:
[0006] A blind signal detection method based on a finite state machine, the improvement of which is that it comprises the following steps:
[0007] Step 1: Collect the signal x(t) received within a period of time. First, convert the signal x(t) into a digital signal x(n), and then equally divide the digital signal x(n) into multiple frames of data with a fixed length. Continuously extract Y data points from each frame of data, so x(n) is divided into U frames of data with a length of Y, and each frame of data is expressed as x u (n), where n = 0, 1, 2, …, Y - 1;
[0008] Step 2: Conduct spectral analysis on the data. Use the fast Fourier transform to calculate the spectrum of the signal, and further obtain the power spectral density of the signal:
[0009] First, perform Fourier transform on the data. The specific formula is as follows:
[0010]
[0011] In the above formula, x u (n) is the signal sequence in the time domain before Fourier transform, X u (k) is the signal sequence in the frequency domain, k is the frequency number, and Y is the sequence length of the signal;
[0012] For the spectrum obtained after Fourier transform, only keep half of it;
[0013] Based on the spectrum result after Fourier transform, obtain the power spectral density of each frame of data:
[0014] F u (k) = |X u (k) 2
[0015] Step 3: Use a Gaussian kernel to smooth the power spectral density:
[0016] Use a Gaussian weighted moving average filter to smooth the noisy data. The formula of the Gaussian weight function is as follows:
[0017]
[0018] In the above formula, w(i) is the weight of the i-th sample point, c is the current sample point, and σ is the standard deviation of the Gaussian function;
[0019] Step 4: Calculate the decision threshold value according to the value of the power spectral density after smoothing:
[0020] Calculate the maximum and minimum values in the smoothed data of the current frame. The formula for calculating the decision threshold value is as follows:
[0021]
[0022] In the above formula, S u(k) is the smoothed data obtained after smoothing the power spectral density, min(·) is the minimum value operation, and max(·) is the maximum value operation;
[0023] Step 5, according to the decision threshold and the finite state machine model, judge the data within the starting point and ending point intervals, effectively distinguish signals and noise, and determine whether a signal exists:
[0024] The finite state machine model is set as: M = (Q, Reason, δ, q 0 , F), where Q = {q 0 , q 1 , q 2 , q 3} represents 4 finite state sets, Reason = {r 1 , r 2 , r 3 , r 4 , r 5} represents 5 state transition condition sets, δ represents the state transition function, q 0 ∈ Q represents the initial state, and F ∈ Q represents the final state;
[0025] q 0 is the initial state, starting to judge whether there is a signal. When the condition r 1 is satisfied, corresponding to the transition rule δ(q 0 , r 1 ) = q 1 , a state transition q 0 → q 1 occurs. The condition r 1 is that the value of the current point is less than the threshold value. The state transition indicates starting to search for the starting point of the signal; if r 1 is not satisfied, then continue to judge the next point;
[0026] q 1 is the state of searching for the starting point of the signal, starting to search for the starting point of the signal segment. When the condition r 2 is satisfied, corresponding to the transition rule δ(q 1 , r 2 ) = q 2 , a state transition q 1 → q 2 occurs. The condition r 2 is that the value of the current point is greater than or equal to the threshold value. The state transition indicates finding the starting point of the signal segment; if r 2 is not satisfied, then continue to judge the next point;
[0027] q 2 is the state of finding the starting point of the signal, judging whether the bandwidth meets the conditions. At the same time, when the conditions r 2 and r3 Perform state transition q 2 →q 3 Condition r 3 is that the length of the current signal segment is greater than the minimum effective bandwidth of the signal; if r is not satisfied 2 then perform the judgment of r 4 and when both condition r 1 and r 4 are satisfied, perform state transition q 2 →q 0 Condition r 4 is that the number of current singular points is greater than the noise bandwidth, switch to the initial state q 0 and redetect the signal; in other cases, continue to judge the next point;
[0028] q 3 is to find an effective signal segment, prepare to calculate its frequency and bandwidth status, and when both condition r 1 and r 4 are satisfied, perform state transition q 3 →q 0 Condition r 1 and r 4 are also satisfied on the basis of r 5 and r 3 →F, condition r 5 is that the length of the current signal segment is less than the maximum effective bandwidth of the signal, and in other cases, continue to judge the next point;
[0029] F is the final state, and use the found effective signal segment to calculate its frequency and bandwidth.
[0030] Furthermore, in step 2, obtain the power spectral density of each frame of data through the periodogram method.
[0031] Furthermore, in step 5, to calculate the frequency and bandwidth, first calculate the frequency resolution and frequency offset, and the formula is as follows:
[0032] Δf = fs / N
[0033]
[0034] In the above formula, fs represents the sampling frequency, N represents the number of sampling points, Δf represents the frequency resolution, F start represents the starting point of the effective signal segment, length represents the length of the signal segment, and ε represents the frequency offset;
[0035] According to the frequency resolution and frequency offset obtained above, obtain the corrected bandwidth and frequency of the effective signal segment:
[0036] B = length × Δf
[0037] f = f o + ε
[0038] In the above formula, f o represents the center frequency, B represents the bandwidth of the signal segment, and f represents the frequency of the signal segment.
[0039] The beneficial effects of the present invention are as follows:
[0040] The method disclosed by the present invention is universal for signals and can be applied to the detection of various signals. By smoothing the power spectral density, noise and interference can be effectively suppressed. At the same time, by adopting a finite state machine, each state in the signal detection process and the transition conditions between states are clearly defined, enabling comprehensive coverage of the signal detection states, accurate judgment of the signal states, and corresponding processing, thus avoiding the problem of state omission that may exist in traditional methods. This method can flexibly adapt to different signal and noise conditions, thereby improving the accuracy and reliability of signal detection. Description of the Drawings
[0041] Figure 1 is a schematic flow chart of the method of the present invention;
[0042] Figure 2 is a state transition diagram for determining the presence of a signal using a finite state machine;
[0043] Figure 3 is a process diagram for determining the presence of a signal using a finite state machine. Detailed Embodiments
[0044] In order to make the objectives, technical solutions, and advantages of the present invention clearer and more understandable, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.
[0045] First, a blind signal detection method based on a finite state machine disclosed by the present invention will be described in detail again. This method calculates a decision threshold based on the difference in power spectra between a signal and noise, and effectively distinguishes between the signal and noise according to the threshold and the finite state machine, and solves for the frequency and bandwidth of the signal. The received signal is detected based on the finite state machine to obtain a signal detection result. It can realize the detection of blind signals in real time, efficiently, and accurately. As Figure 1 shown, it includes the following steps:
[0046] Step 1, collect the signal x(t) received within a period of time. First, convert the signal x(t) into a digital signal x(n), and then equally divide the digital signal x(n) into multiple frames of data (binary data streams) with a fixed length. Y data points are continuously extracted from each frame of data. Then x(n) is divided into U frames of data with a length of Y, and each frame of data is expressed as x u (n), where n = 0, 1, 2, …, Y - 1; the following signal detection method is for each frame of data. Each frame of the divided data is detected segment by segment. The following process will only introduce the process of signal detection for one segment of data.
[0047] Step 2, perform spectral analysis on the data. Use the fast Fourier transform (FFT) to calculate the spectrum of the signal, and further obtain the power spectral density of the signal:
[0048] First, perform Fourier transform on the data. The specific formula is as follows:
[0049]
[0050] In the above formula, x u (n) is the signal sequence in the time domain before Fourier transform, X u (k) is the signal sequence in the frequency domain, k is the frequency number, and Y is the sequence length of the signal;
[0051] For the spectrum obtained after Fourier transform, only keep half of it, that is, the single-sided spectrum;
[0052] Based on the spectrum result after Fourier transform, obtain the power spectral density of each frame of data. Obtain the power spectral density of each frame of data through the periodogram method:
[0053] F u (k) = |X u (k) 2
[0054] Step 3, in order to eliminate the noise and mutations in the power spectrum, use a Gaussian kernel to smooth the power spectral density, and then retain the main frequency characteristics of the signal:
[0055] Use a Gaussian weighted moving average filter to smooth the noisy data. Different from the moving average filter, the weighted moving average filter assigns different weights to different samples when calculating the average value of the samples within the sliding window, and uses this weight vector to perform weighted average on the samples within the window. This weight vector defines the importance of each sample in the average calculation and can better adapt to the changes of the signal. The points closer to the current sample point are given higher weights, and the points farther from the current sample point are given lower weights. In this way, the recent data points are more emphasized, and the influence of the farther points on the average value is weakened, so as to achieve the smoothing effect. The formula of the Gaussian weight function is as follows:
[0056]
[0057] In the above formula, w(i) is the weight of the i-th sample point, c is the current sample point, and σ is the standard deviation of the Gaussian function;
[0058] Step 4: Calculate the decision threshold value according to the value of the power spectral density after smoothing. The selection of the threshold value is based on the difference in the power spectra between the signal and the noise.
[0059] After smoothing the power spectrum, the smoothed data in the current frame is obtained. Calculate the maximum and minimum values in the smoothed data in the current frame. Depending on the type of signal, the calculated maximum and minimum values are different. Determine the threshold value according to different signals. The formula for calculating the decision threshold value is as follows:
[0060]
[0061] In the above formula, S u (k) is the smoothed data obtained after smoothing the power spectral density, min(·) is the operation of taking the minimum value, max(·) is the operation of taking the maximum value, and to calculate the decision threshold value, the maximum and minimum values of the smoothed data need to be obtained first.
[0062] Step 5: According to the decision threshold value and the finite state machine model, judge the data within the start point and end point intervals, effectively distinguish the signal and the noise, and determine whether the signal exists:
[0063] In this step, comparing the power spectrum within the received signal interval with the decision threshold is not enough just relying on a single comparison process. In the actual environment, the presence of noise often interferes with signal detection. To effectively distinguish noise and signal, the method of the present invention adopts a finite state machine model. By setting multiple states to reflect various possible situations in the signal detection process, including the start, duration, termination of the signal, and noise, etc., the transition between states is based on the result of the comparison process. When the received signal characteristics match a certain state, the finite state machine will enter that state and make the next judgment according to the preset transition conditions. In this way, the finite state machine can summarize all the states and all the conditions in the comparison process to ensure that all situations are comprehensively covered.
[0064] The finite state machine model is set as: M = (Q, Reason, δ, q 0 , F), where Q = {q 0 , q 1 , q 2 , q 3} represents a set of 4 finite states, Reason = {r 1,r 2 ,r 3 ,r 4 ,r 5} represents five sets of state transition conditions, δ represents the state transition function, q 0 ∈Q represents the initial state, and F∈Q represents the final state;
[0065] The specific state transitions are shown in the following table, Figure 2 For the corresponding state transition diagram. The following values need to be defined: the minimum signal bandwidth min_band (the minimum bandwidth length of the effective signal), the maximum signal bandwidth max_band (the maximum bandwidth length of the effective signal), and the noise bandwidth noise_band (the bandwidth length recognized as noise)
[0066] Table 1 State Transition Table
[0067]
[0068] In the above table, length represents the length of the effective signal bandwidth, that is, the length of the signal segment that satisfies the condition r 2 ; width represents the length of the singularity point, that is, the length of the signal segment that satisfies the condition r 1 .
[0069] q 0 is the initial state, starting to judge whether there is a signal. If the condition r 1 is satisfied, corresponding to the transition rule δ(q 0 ,r 1 ) = q 1 , a state transition q 0 →q 1 occurs. The condition r 1 is that the value of the current point is less than the threshold. The state transition indicates starting to find the starting point of the signal; if r 1 is not satisfied, then continue to judge the next point;
[0070] q 1 is the state of finding the starting point of the signal, starting to find the starting point of the signal segment. If the condition r 2 is satisfied, corresponding to the transition rule δ(q 1 ,r 2 ) = q 2 , a state transition q 1 →q 2 occurs. The condition r 2 is that the value of the current point is greater than or equal to the threshold. The state transition indicates finding the starting point of the signal segment; if r 2 is not satisfied, then continue to judge the next point;
[0071] q 2It is to find the starting point of the signal, judge whether the bandwidth meets the conditional state, and simultaneously meet the condition r 2 and r 3 , perform a state transition q 2 →q 3 , the condition r 3 is that the length of the current signal segment is greater than the minimum effective bandwidth of the signal. The conversion of this step state indicates that the signal bandwidth meets the minimum bandwidth and can be considered as an effective signal segment; if it does not meet r 2 , then perform the judgment of r 4 , and simultaneously meet the conditions r 1 and r 4 , perform a state transition q 2 →q 0 , the condition r 4 is that the number of current singular points is greater than the noise bandwidth. The conversion of this step state indicates that the noise is misjudged as a signal, and it switches to the initial state q 0 and re - perform the signal detection; in other cases, continue to judge the next point;
[0072] q 3 is to find an effective signal segment, prepare to calculate its frequency and bandwidth state, and simultaneously meet the conditions r 1 and r 4 , perform a state transition q 3 →q 0 , the conversion of this step state indicates that the number of singular points before the current point is greater than the noise bandwidth, and the currently detected is noise, and it is necessary to re - enter the initial state for the detection of effective signals; on the basis of meeting the conditions r 1 and r 4 also meet r 5 , perform a state transition q 3 →F, the condition r 5 is that the length of the current signal segment is less than the maximum effective bandwidth of the signal. The conversion of this step state indicates that the signal bandwidth meets the maximum bandwidth and the frequency and bandwidth can be estimated; in other cases, continue to judge the next point;
[0073] F is the final state, and using the found effective signal segment, calculate its frequency and bandwidth.
[0074] Embodiment 1. This embodiment discloses a blind signal detection method based on a finite - state machine, which detects the received signal and includes the following steps:
[0075] S1: Collect the signals received within a period of time, convert the signals into digital signals, and then divide the signals into binary data streams with a fixed length. In this embodiment, the length Y of each frame of data is set to 65536. The following signal detection method is for each frame of data. The segmented data of each frame is detected segment by segment. The following process will only introduce the process of signal detection for one segment of data.
[0076] S2: Conduct spectral analysis on the data, calculate the spectrum of the signal using the Fast Fourier Transform (FFT), and further obtain the power spectrum of the signal.
[0077] S3: To eliminate the noise and mutations in the power spectrum, use a Gaussian kernel to smooth the power spectrum, thereby retaining the main frequency characteristics of the signal. Use a Gaussian weighted moving average filter to smooth the noisy data. The sliding window size is set to 100. Different from the moving average filter, the weighted moving average filter assigns different weights to different samples when calculating the average value of the samples within the sliding window, and uses this weight vector to perform a weighted average on the samples within the window.
[0078] S4: Calculate the decision threshold based on the value of the smoothed power spectrum. The selection of the threshold is based on the difference in the power spectra between the signal and the noise, which can effectively distinguish between the signal and the noise. After smoothing the power spectrum, the smoothed data under the current frame is obtained. Calculate the maximum and minimum values in the smoothed data under the current frame, and determine the decision threshold value from the maximum and minimum values.
[0079] S5: According to the decision threshold value and the finite state machine model, judge the data within the start point and end point intervals, effectively distinguish between the signal and the noise, and determine whether the signal exists.
[0080] The following describes Figure 3 Judging the received signals. Starting from the initial point, compare the power spectrum and the threshold value for each point, and the state corresponding to each point is also different.
[0081] Point A is the q 0 Initial state. Starting from this point, judge whether there is a signal in the data segment. If S u (i) < P is satisfied, perform a state transition q 0 →q 1 .
[0082] The next point B is the q 1 State of searching for the start point of the signal. If S u (i) ≥ P is not satisfied, continue to judge the next point. Judge point by point until point C satisfies the condition S u (i) ≥ P, and it is possible to find the start point of the signal segment. Perform a state transition q 1→q 2 。
[0083] Point C is q 2 The starting point of the signal has been found, and the state of judging whether the bandwidth meets the condition starts. Judgments are made point by point from point C backwards until point D meets length≥min_band. length refers to the bandwidth from the currently judged point to point C. When it meets the minimum signal bandwidth min_band, a state transition to q occurs 2 →q 3 。
[0084] Judgment continues from point D. Starting from point E, the condition r is not met 2 . width is used to count the number of consecutive points that do not meet the condition until the number of data points that do not meet the condition at point F is greater than the noise bandwidth, that is, ++width>noise_band. At this time, the judgment of points is no longer continued downward, and a state transition to q occurs 3 →q 0 , indicating that the signal detection of the current data segment has ended, and it jumps to the initial state q 0 , and the signal detection of the next data segment is carried out. On this basis, when the bandwidth from point D to point E meets the condition of being less than the maximum signal bandwidth max_band, a state transition to q occurs 3 →F, calculate the frequency and bandwidth of the effective signal segment from point D to point E
[0085] To calculate the frequency and bandwidth, the frequency resolution and frequency offset need to be calculated first. The formula is as follows
[0086] Δf=fs / N
[0087]
[0088] In the above formula, fs represents the sampling frequency, N represents the number of sampling points, Δf represents the frequency resolution, F start represents the starting point of the effective signal segment, length represents the length of the signal segment, and ε represents the frequency offset
[0089] Based on the frequency resolution and frequency offset obtained above, the corrected bandwidth and frequency of the effective signal segment are obtained
[0090] B=length×Δf
[0091] f=f o +ε
[0092] In the above formula, f o represents the center frequency, B represents the bandwidth of the signal segment, and f represents the frequency of the signal segment
[0093] The signal detection of the next data segment starts from point F. Point F is q0 Initial state, starting from this point to judge the presence or absence of signals in the data segment, satisfying S u (i) < P, perform state transition q 0 →q 1 .
[0094] The next point, G point, is q 1 State of searching for the signal starting point, not satisfying S u (i) ≥ P, continue to judge the next point. Judge point by point until point H satisfies condition S u (i) ≥ P, may have found the starting point of the signal segment, perform state transition q 1 →q 2 .
[0095] Point H is q 2 Have found the signal starting point, start the state of judging whether the bandwidth meets the conditions. Calculate the bandwidth from the currently judged point to point H point by point starting from point H. When the minimum signal bandwidth min_band has not been met, the values of the points after point I are less than the threshold value, and width is used for counting until the number of data points that do not satisfy the conditions after point J is greater than the noise bandwidth, that is, ++width > noise_band. At this time, no longer continue to judge the points downward, perform state transition q 2 →q 0 , indicating that the signal detection of the current data segment has ended, jump to the initial state q 0 , and perform signal detection on the next data segment.
[0096] It should be understood that various forms of the flow shown above can be used, reordering, adding or deleting steps. For example, the steps described in this application can be executed in parallel, sequentially, or in different orders, as long as the desired results of the technical solutions disclosed in this application can be achieved, they are all within the protection scope of the present invention.
Claims
1. A blind signal detection method based on a finite state machine, characterized in that: The steps include: Step 1: Collect the signal x(t) received within a period of time, first convert the signal x(t) into a digital signal x(n), and then divide the digital signal x(n) into multiple frames of fixed length. Each frame of data continuously extracts Y data points, then divide x(n) into U frames of length Y, and each frame of data is represented by x u (n), n=0, 1, 2,..., Y-1; Step 2: Perform spectrum analysis on the data, use fast Fourier transform to calculate the spectrum of the signal, and further obtain the power spectrum density of the signal: First, perform Fourier transform on the data. The specific formula is as follows: In the above formula, x u (n) is the signal sequence in the time domain before Fourier transform, X u (k) is the signal sequence in the frequency domain, k is the frequency number, and Y is the sequence length of the signal; Only half of the spectrum obtained after Fourier transform is retained; The power spectrum density of each frame of data is obtained based on the spectrum result after Fourier transform: F u (k)=|X u (k) 2 Step 3: Use Gaussian kernel to smooth the power spectrum density: Use Gaussian weighted moving average filter to smooth the noisy data. The formula of Gaussian weight function is as follows: In the above formula, w(i) is the weight of the i-th sample point, c is the current sample point, and σ is the standard deviation of the Gaussian function; Step 4: Calculate the decision threshold value according to the value of the power spectrum density after smoothing: Calculate the maximum and minimum values in the smoothed data of the current frame, and the formula for calculating the decision threshold is as follows: In the above formula, S u (k) is the smoothed data obtained after smoothing the power spectrum density, min(·) is the minimum value operation, and max(·) is the maximum value operation; Step 5: Based on the decision threshold and the finite state machine model, the data within the start point and end point intervals are judged to effectively distinguish between signals and noise and determine whether a signal exists: The finite state machine model is set as: M = (Q, Reason, δ, q0, F), where Q = {q0, q1, q2, q3} represents four finite state sets, Reason = {r1, r2, r3, r4, r5} represents five state transition condition sets, δ represents the state transition function, q0∈Q represents the initial state, and F∈Q represents the final state; q0 is the initial state, and the judgment of whether there is a signal begins. If the condition r1 is met, the corresponding transfer rule δ(q0,r1)=q1 is used to transfer the state q0→q1. The condition r1 is that the value of the current point is less than the threshold value. The state transition indicates the start of the search for the starting point of the signal. If r1 is not met, the judgment of the next point will continue. q1 is the state of searching for the starting point of the signal. It starts to search for the starting point of the signal segment. If the condition r2 is met, the corresponding transfer rule δ(q1,r2)=q2 is used to transfer the state q1→q2. The condition r2 is that the value of the current point is greater than or equal to the threshold value. The state transition indicates that the starting point of the signal segment has been found. If r2 is not met, the next point will be judged. q2 is to find the starting point of the signal and determine whether the bandwidth meets the conditional state. If conditions r2 and r3 are met at the same time, the state transition q2→q3 is performed. Condition r3 is that the length of the current signal segment is greater than the minimum effective bandwidth of the signal. If r2 is not met, r4 is judged. If conditions r1 and r4 are met at the same time, the state transition q2→q0 is performed. Condition r4 is that the number of current singular points is greater than the noise bandwidth. Switch to the initial state q0 to re-detect the signal. In other cases, continue to judge the next point. q3 is to find the valid signal segment and prepare to calculate its frequency and bandwidth state. If conditions r1 and r4 are met at the same time, the state transition q3→q0 is performed. If conditions r1 and r4 are met, r5 is also met, and the state transition q3→F is performed. Condition r5 is that the length of the current signal segment is less than the maximum effective bandwidth of the signal. In other cases, continue to judge the next point. F is the final state, using the found valid signal segment, its frequency and bandwidth are calculated.
2. The blind signal detection method based on finite state machine according to claim 1, characterized in that: In step 2, the power spectrum density of each frame of data is obtained by the periodogram method.
3. The blind signal detection method based on finite state machine according to claim 1, characterized in that: In step 5, the frequency and bandwidth must first be calculated by calculating the frequency resolution and frequency deviation, the formula is as follows: Δf=fs / N In the above formula, fs represents the sampling frequency, N represents the number of sampling points, Δf represents the frequency resolution, and F start Indicates the starting point of the valid signal segment, length indicates the length of the signal segment, and ε indicates the frequency deviation; According to the frequency resolution and frequency deviation obtained above, the correction bandwidth and frequency of the effective signal segment are obtained: B=length×Δf f=f o +e In the above formula, f o represents the center frequency, B represents the bandwidth of the signal segment, and f represents the frequency of the signal segment.