A soft decision method for GFSK signal based on multi-sampling point adaptive weighting

Abstract

This invention relates to the field of wireless communication technology, specifically disclosing a soft-decision method for GFSK signals based on multi-sampling point adaptive weighting. The method includes: extracting phase change features from the I / Q signal and performing three-level anti-interference processing to obtain an optimized phase difference sequence; performing frequency offset correction and symbol synchronization on the optimized phase difference sequence based on a predefined local ZC preamble sequence and a local gold code sequence in the frame structure; selecting three sampling points (middle, left, and right) from eight times the oversampling points in each symbol period; performing interference compensation on the data from the three sampling points and then weighting and fusing them to obtain a weighted fused sequence; constructing a reference sequence based on the local gold code sequence, and calculating the log-likelihood ratio between the reference sequence and the weighted fused sequence bit by bit as the soft decision value; and performing subsequent decoding based on the soft decision value. This invention can stably reduce the bit error rate and maximize the system coding gain in low signal-to-noise ratio and multipath interference scenarios.

Description

Technical Field

[0001] This invention relates to the field of wireless communication technology, and more specifically to a soft decision method for GFSK signals based on multi-sampling point adaptive weighting. Background Technology

[0002] GFSK (Gaussian Frequency Shift Keying) is a core modulation scheme in communication systems such as Bluetooth Low Energy (BLE) and low-Earth orbit satellite IoT, and its demodulation performance directly determines the reliability of the communication link. Traditional GFSK demodulation methods generally employ single-point sampling decision, classical bit synchronization algorithms, and hard decision methods. These methods have the following inherent drawbacks:

[0003] Single-point sampling decision only selects a single sample value from the maximum open point of the eye diagram for decision, which wastes the redundant information brought by 8 times oversampling, is extremely sensitive to timing jitter - even a small offset of the sampling point will lead to decision error, and has limited noise resistance, with a significant increase in bit error rate in low signal-to-noise ratio (SNR≤-10dB) scenarios.

[0004] Classical bit synchronization algorithms have high computational complexity, rely on phase-locked loops (PLLs) for locking, have long convergence times, and are sensitive to inter-symbol interference (ISI). Synchronization accuracy drops significantly in multipath and Doppler frequency offset scenarios.

[0005] Hard decision directly outputs 0 / 1 bits, completely losing the reliability information of bit decisions. It cannot provide soft input for iterative decoding such as Turbo, resulting in limited system coding gain and difficulty in meeting the requirements of long-distance, high-reliability communication.

[0006] These traditional methods severely limit the system's anti-interference capability and coding gain. Therefore, how to stably reduce the bit error rate and maximize the system coding gain in low signal-to-noise ratio and multipath interference scenarios is a problem that urgently needs to be solved by those skilled in the art. Summary of the Invention

[0007] In view of the above problems, this invention proposes a soft decision method for GFSK signals based on multi-sampling point adaptive weighting, so as to overcome the above problems or at least partially solve them.

[0008] To achieve the above objectives, the present invention adopts the following technical solution: This invention provides a soft decision method for GFSK signals based on multi-sampling point adaptive weighting, comprising the following steps: S1. Perform phase difference detection on the frequency offset compensated I / Q signal and extract the signal phase change characteristics; S2. Perform three-level anti-interference processing on the signal phase change characteristics to obtain the optimized phase difference sequence; S3. Based on the predefined local ZC preamble sequence and local gold code sequence in the frame structure, perform frequency offset correction and symbol synchronization on the optimized phase difference sequence; S4. Select three sampling points—the middle, left, and right—from the 8 times oversampling points in each symbol period; S5. After interference compensation, the data from the three sampling points are weighted and fused to obtain a weighted fused sequence. S6. Based on the last n bits of the local gold code sequence, construct a reference sequence, calculate the log-likelihood ratio between the reference sequence and the weighted fusion sequence bit by bit, and use it as the soft decision value; perform subsequent decoding based on the soft decision value.

[0009] Furthermore, in S1, the phase difference detection method for the I / Q signals is as follows: first, the I / Q signals are sequentially unwrapped and normalized, and then the difference between adjacent phases is calculated on the normalized signals to obtain the signal phase change characteristics.

[0010] Furthermore, in S2, the three-level anti-interference processing includes: anti-interference limiting processing, 8 times sampling point accumulation processing, and phase normalization processing; The anti-interference limiting process is as follows: set a threshold value of t. If the absolute value of the difference between adjacent phases is greater than t, it is determined that there is sudden interference. At this time, the current phase difference value is clamped to t or -t by hard limiting. The 8x sampling point accumulation processing is as follows: Under 8x oversampling, each symbol period contains 8 sampling points. For each symbol period, starting from the first sampling point, the 8 sampling points are summed sequentially to obtain the effective phase energy of that symbol period. Phase normalization is performed by normalizing the signal phase change characteristics to... between.

[0011] Furthermore, S3 includes: The optimized phase difference sequence and the local ZC preamble sequence are subjected to sliding autocorrelation operation. The position corresponding to the peak value of the autocorrelation operation exceeding the preset threshold is taken as the starting position of ZC preamble timing, thus completing the preamble timing. The phase difference sequence segment corresponding to the starting position obtained from the ZC preamble timing is extracted, and the phase difference sequence segment is divided into multiple sub-segments according to the preset window length. The phase mean of each sub-segment is calculated, and the calculated phase mean is subtracted from each sampling point in the sub-segment to obtain the phase sequence after frequency offset correction. From the frequency offset corrected phase differential signal sequence, locate the end position of the ZC preamble, extract the subsequent sequence segment and perform sliding correlation operation with the local gold sequence, and take the position corresponding to the peak value of the correlation operation exceeding the preset threshold as the starting position of the synchronization word to complete the synchronization word frame synchronization.

[0012] Furthermore, in S4, the middle sampling point is the maximum open position of the eye diagram, and the 4th or 5th sampling point is taken; the left sampling point is the position of the middle sampling point shifted to the left by 2 sampling clock cycles; the right sampling point is the position of the middle sampling point shifted to the right by 2 sampling clock cycles.

[0013] Furthermore, S5 includes: Configure the basic weights for the middle sampling point, the left sampling point, and the right sampling point; For the three sampled data points within the first symbol period, only backward interference compensation is performed; for the three sampled data points within the last symbol period, only forward interference compensation is performed; for the three sampled data points within any intermediate symbol period other than the first and second symbol periods, both forward interference compensation and backward interference compensation are performed respectively. After performing forward interference compensation or backward interference compensation on the three sampling points within each symbol period, mean filtering is then applied. Within each symbol period, the data from the three sample points after mean filtering are weighted and summed according to the configured basic weights to obtain a weighted fusion sequence.

[0014] Furthermore, in S5, the mean filter is: After forward interference compensation or backward interference compensation, three independent sampling point sequences are obtained, namely the left sampling point sequence, the middle sampling point sequence, and the right sampling point sequence. For each sampling point sequence, the average of the sampling point data of the current symbol period and the previous several symbol periods is calculated to obtain the filtered sampling point data of the current symbol period.

[0015] Furthermore, in S5, forward interference compensation includes: Take the absolute value of the difference between the amplitudes of the three sampling points in the previous symbol period and the amplitudes of the three sampling points in the current symbol period at corresponding positions, and then sum them up to get the forward interference coefficient for the current symbol period. Forward interference compensation is performed point-by-point on the three sampling points of the current symbol period based on the forward interference coefficient.

[0016] Furthermore, in S5, backward interference compensation includes: Take the absolute value of the difference between the amplitudes of the three sampling points in the next symbol period and the amplitudes of the three sampling points in the current symbol period at corresponding positions, and then sum them up to get the backward interference coefficient for the current symbol period. Backward interference compensation is performed point-by-point on the three sampling points of the current symbol period based on the backward interference coefficient.

[0017] Furthermore, in S6, the calculation process for the soft decision threshold includes: Based on the last n bits of the local gold code sequence, construct two reference sequences ref0 and ref1, where ref0 represents the sequence obtained by adding a 0 to the end of the reference sequence when the current bit is 0, and ref1 represents the sequence obtained by adding a 1 to the end of the reference sequence when the current bit is 1. The correlation values ​​between the weighted fusion sequence and ref0 and ref1 are calculated bit by bit and denoted as sum0 and sum1 respectively. The log-likelihood ratio is generated based on the difference between sum0 and sum1, and used as the soft decision value. If the difference between sum0 and sum1 is greater than 0, the soft decision value is biased towards bit 0; if it is less than 0, the soft decision value is biased towards bit 1.

[0018] As can be seen from the above technical solution, compared with the prior art, the present invention has the following beneficial effects: 1. This invention combines three-point optimal sampling with weighted fusion method to make full use of oversampling redundant information, improves tolerance to timing jitter, significantly enhances anti-interference and anti-jitter capabilities, and greatly reduces bit error rate compared with traditional single-point sampling in low signal-to-noise ratio and multipath interference scenarios.

[0019] 2. This invention relies on ZC preamble and Gold synchronization word to achieve fast timing, improve synchronization efficiency and accuracy, and reduce convergence time and synchronization error.

[0020] 3. This invention utilizes known gold code sequences for correlation soft decision-making and outputs LLR soft decision information, providing complete confidence support for Turbo and LDPC iterative decoding, significantly improving the system coding gain, and enhancing the coding gain compared to hard decision schemes.

[0021] 4. This invention integrates a six-level collaborative processing flow, including phase difference detection, three-level anti-interference optimization, timing synchronization, three-point optimal sampling, adaptive weighted fusion, and synchronization word correlation soft decision. The entire process involves no complex calculations, and the hardware implementation requires only a small amount of logic resources. It can be directly integrated into existing BLE and low-orbit satellite IoT receivers without significant modifications to the hardware architecture, demonstrating strong compatibility and practicality. Attached Figure Description

[0022] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0023] Figure 1 This is a flowchart of the soft decision method for GFSK signals based on multi-sampling point adaptive weighting provided in the embodiments of the present invention; Figure 2 This is a flowchart of the three-level anti-interference processing provided in the embodiments of the present invention; Figure 3 This is a flowchart of the adaptive fusion of three sampling points provided in an embodiment of the present invention; Figure 4 This diagram illustrates the performance comparison of different algorithms. Detailed Implementation

[0024] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0025] like Figure 1 As shown, this embodiment of the invention discloses a soft decision method for GFSK signals based on multi-sampling point adaptive weighting, including the following steps: S1. Perform phase difference detection on the frequency offset compensated I / Q signal and extract the signal phase change characteristics; S2. Perform three-level anti-interference processing on the signal phase change characteristics to obtain the optimized phase difference sequence; S3. Based on the predefined local ZC preamble sequence and local gold code sequence in the frame structure, perform frequency offset correction and symbol synchronization on the optimized phase difference sequence; S4. Select three sampling points—the middle, left, and right—from the 8 times oversampling points in each symbol period; S5. After interference compensation, the data from the three sampling points are weighted and fused to obtain a weighted fused sequence. S6. Based on the last n bits of the local gold code sequence, construct a reference sequence, calculate the log-likelihood ratio between the reference sequence and the weighted fusion sequence bit by bit, and use it as the soft decision value; perform subsequent decoding based on the soft decision value.

[0026] The specific implementation methods of the above steps will be further explained below.

[0027] S1, Phase Difference Detection: Phase difference detection is performed on the frequency offset compensated I / Q signals. Specifically, the I / Q signals are first unwrapped and normalized sequentially. Then, the difference between adjacent phases is calculated on the normalized signals to obtain the signal phase change characteristics. This step provides accurate phase data support for subsequent demodulation and decision-making by the Bluetooth receiver, without additional redundant calculations, meeting the design requirements of low power consumption and high precision.

[0028] S2, Level 3 Anti-interference and Optimization: Based on the phase change characteristics of the signal output by S1, sudden interference is suppressed, effective phase energy is extracted, and noise interference is smoothed in sequence, ultimately outputting a high-fidelity, low-noise phase differential signal. Specifically, such as... Figure 2 As shown, the three-level anti-interference processing includes: anti-interference limiting processing, 8 times sampling point accumulation processing, and phase normalization processing.

[0029] The anti-interference limiting processing identifies sudden interference by detecting abrupt changes in the phase difference between adjacent phases. Specifically, a threshold t is set. If the absolute value of the difference between adjacent phases is greater than t, sudden interference is identified. In this case, hard limiting is used to clamp the current phase difference to t or -t. In this embodiment, the threshold can be set to 0.625. If the absolute value of the difference between two consecutive difference values ​​is greater than 0.625, it indicates that the current term introduces excessive noise. The difference between the phase difference between the current sampling point and the previous sampling point is calculated. If this difference is greater than 0.625 or less than -0.625, sudden interference is identified. Hard limiting is used to clamp the current phase difference to 0.625 or -0.625, avoiding the cascading effects of sudden interference on subsequent processing. This method also has low computational complexity and meets low power consumption requirements.

[0030] The 8x sampling point accumulation processing is as follows: Under 8x oversampling, each symbol period contains 8 sampling points. For each symbol period, starting from the first sampling point, the sum of the 8 sampling points is taken sequentially as the effective phase energy of that symbol period, thus obtaining the cumulative value of phase change within each symbol period. This cumulative value corresponds to the frequency offset characteristic of the GFSK signal. To improve the signal-to-noise ratio, zero-padding is performed to ensure no data loss during the accumulation and summation.

[0031] Phase normalization is performed by normalizing the signal phase change characteristics to... Between. Phase normalization can solve the phase accumulation overflow problem, ensure the consistency and dimensional uniformity of phase data, and provide a standard input for subsequent Gaussian filtering.

[0032] S3, preamble and synchronization word timing, relies on the ZC preamble and Gold synchronization word in the frame structure (generated locally based on the sequence length and root index parameters given in the frame structure) to complete preamble timing, initial carrier frequency offset estimation, and synchronization word frame synchronization, specifically including: S31, ZC preamble timing process: Perform sliding autocorrelation operation on the optimized phase difference sequence and the local ZC preamble sequence. The position corresponding to the peak value of the autocorrelation operation exceeding the preset threshold is taken as the starting position of ZC preamble timing, thus completing the preamble timing.

[0033] S32. Initial carrier frequency offset estimation process: Extract the phase difference sequence segment corresponding to the starting position obtained by ZC preamble timing, divide the phase difference sequence segment into multiple sub-segments according to the preset window length (width=64), calculate the phase mean of each sub-segment, and subtract the calculated phase mean from each sampling point in the sub-segment to obtain the phase sequence after frequency offset correction.

[0034] S33. Locate the end position of the ZC preamble from the frequency offset corrected phase differential signal sequence, extract the subsequent sequence segment and perform sliding correlation operation with the local gold sequence, and take the position corresponding to the peak value of the correlation operation exceeding the preset threshold as the starting position of the synchronization word to complete the synchronization word frame synchronization.

[0035] This step provides a precise time and frequency reference for subsequent data frame demodulation and decoding, ensuring consistent frame and symbol synchronization between the receiver and transmitter.

[0036] S4. Single-point optimal sampling selection: For the symbol period of 8 times oversampling, three independent sampling points covering the effective open area of ​​the eye diagram are selected from the left, middle and right to obtain the three most representative sampling points; then the DC component of each sampling data is eliminated, which not only preserves the core effective information of the signal, but also reduces the complexity of subsequent calculations, and at the same time has complementary fault tolerance capability for timing errors.

[0037] The middle sampling point is the position where the eye diagram is at its maximum opening, usually the 4th or 5th sampling point; the left sampling point is the position where the middle sampling point is shifted to the left by 2 sampling clock cycles; and the right sampling point is the position where the middle sampling point is shifted to the right by 2 sampling clock cycles.

[0038] The selection criteria for the above three sampling point intervals are: 1) an interval of 2 sampling points to ensure that each point is relatively independent; 2) to cover the effective open area of ​​the eye diagram; and 3) to have a complementary effect on timing errors.

[0039] S5. Weighted Fusion: Before bit decision, interference from adjacent symbols in signal transmission is eliminated, and multi-sampled data is weighted and fused to improve the accuracy of subsequent bit decisions. The specific process is as follows: Figure 3 As shown, it specifically includes: S51. Configure the basic weights of the middle sampling point, the left sampling point, and the right sampling point.

[0040] S52. For the three sampled data points within the first symbol period, only backward interference compensation is performed; for the three sampled data points within the last symbol period, only forward interference compensation is performed; for the three sampled data points within any intermediate symbol period other than the first and second symbol periods, both forward interference compensation and backward interference compensation are performed respectively.

[0041] Forward interference compensation includes: Take the absolute value of the difference between the amplitudes of the three sampling points in the previous symbol period and the amplitudes of the three sampling points in the current symbol period at corresponding positions, and then sum them up to get the forward interference coefficient K_force for the current symbol period. The forward interference coefficient can quantify the interference intensity of the previous symbol to the current symbol and is used to cancel point-by-point forward interference. Forward interference compensation is performed point-by-point on the three sampling points of the current symbol period based on the forward interference coefficient. The calculation process for forward interference compensation on the sampling points of the current symbol period can be expressed as follows: Where i = 1, 2, 3, corresponding to the left sampling point, the middle sampling point, and the right sampling point, respectively. This represents the amplitude after forward interference compensation for the i-th sampling point in the current symbol period n. The amplitude of the i-th sampling point in the current symbol period n before forward interference compensation; As a compensation factor, it can be preset or adaptively adjusted according to channel characteristics; Indicates the forward interference coefficient; It represents the amplitude of the i-th sampling point in the previous symbol period n-1.

[0042] Backward interference compensation includes: Take the absolute value of the difference between the amplitudes of the three sampling points in the next symbol period and the amplitudes of the three sampling points in the current symbol period at corresponding positions, and then sum them up to get the backward interference coefficient K_back for the current symbol period. Backward interference compensation is performed point-by-point on the three sampling points of the current symbol period based on the backward interference coefficient. The calculation method of backward interference compensation is similar to that of forward interference compensation.

[0043] S53. After performing forward or backward interference compensation on the three sampled data points within each symbol period, mean filtering is performed. Mean filtering eliminates the DC offset of the compensated data, allowing the data to be distributed around the zero point, facilitating subsequent decision-making. The main function of this step is to quantify and cancel the interference of the previous symbol on the current symbol, while correcting the data baseline to avoid the DC component affecting the decision-making. Specifically, the mean filtering process is as follows: After forward interference compensation or backward interference compensation, three independent sampling point sequences are obtained, namely the left sampling point sequence, the middle sampling point sequence, and the right sampling point sequence. For each sampling point sequence, the average of the sampling point data of the current symbol period and the previous several symbol periods is calculated to obtain the filtered sampling point data of the current symbol period.

[0044] S54. Within each symbol period, the three sample points after mean filtering are weighted and summed according to the configured basic weights. The weighted sample point data of each symbol period are combined into a weighted fusion sequence.

[0045] For the intermediate symbol period, after forward interference compensation of the sampling points, a mean filter is performed. After backward interference compensation, a mean filter is performed again to eliminate the DC offset after bidirectional compensation. This step adds backward interference compensation on top of forward interference compensation to achieve bidirectional interference cancellation and further purify the signal. After interference compensation in both directions, the data is significantly purified. The three sampling points are fused according to dynamic weights, and the final weighted fusion sequence used for bit decision is output. The value of this sequence directly reflects the probability of the corresponding position being "0" or "1", for subsequent bit decision.

[0046] In this step, bidirectional elimination of inter-symbol interference is achieved through forward and backward interference coefficients; mean filtering can eliminate DC offset of data multiple times to ensure that data is distributed around 0 point and improve decision accuracy; combined with fixed base weights and dynamic interference coefficients, multi-sample data is fused into a single weight sequence to provide a reliable basis for bit decision.

[0047] S6. Based on the correlation soft decision of the synchronization word, using the local gold code sequence in the known frame structure from S3, the last 7 bits of the synchronization word are taken, and the log-likelihood ratio (LLR) of the data bits is determined bit by bit through sliding correlation, instead of directly outputting 0 / 1 decision values, which is more suitable for subsequent decoding. The specific process is as follows: Based on the last n bits (7 in this embodiment) of the local gold code sequence, two reference sequences ref0 and ref1 are constructed. ref0 represents the sequence obtained by adding a 0 to the end of the reference sequence when the current bit is 0, and ref1 represents the sequence obtained by adding a 1 to the end of the reference sequence when the current bit is 1. ref0 and ref1 are respectively represented as:

[0048] The correlation values ​​between the weighted fusion sequence and ref0 and ref1 are calculated bit by bit and denoted as sum0 and sum1 respectively. The log-likelihood ratio is generated based on the difference between sum0 and sum1, and is used as the soft decision value LLR. If the difference between sum0 and sum1 is greater than 0, the soft decision value LLR is biased towards bit 0; if it is less than 0, the soft decision value is biased towards bit 1.

[0049] Overall, firstly, this invention overcomes the limitations of single-point sampling by selecting three independent sampling points (left, center, and right) to cover the effective area of ​​the eye diagram. This utilizes redundant information to improve anti-interference capabilities and simplifies the bit synchronization process. Secondly, this invention highlights the value of core sampling points based on differentiated basic weights, dynamically adjusting the adaptation signal stable region and transition edges to balance information effectiveness and anti-interference capabilities. Thirdly, this invention uses the known end of the synchronization word as a reference and directly outputs LLR soft decision through sliding correlation, eliminating the need for additional training sequences and providing high-confidence soft information for iterative decoding. Finally, this invention employs basic mathematical operations (summation, dot product, and sign judgment) in each stage, without complex iterations or matrix operations, reducing computational load and meeting the resource constraints of low-power receivers.

[0050] Next, simulation experiments were conducted to further verify the performance of the invention.

[0051] The simulation experiment, based on a GFSK modulated signal (carrier frequency 2MHz, 8 times oversampling), constructs a multipath fading channel and a Doppler frequency offset scenario (frequency offset range ±60kHz, -720Hz / s Doppler rate of change). A soft decision is performed using a multi-sampling point adaptive weighting algorithm, such as... Figure 4 As shown, its performance is significantly better than the traditional single-point sampling hard decision scheme.

[0052] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on the differences from other embodiments. The same or similar parts between the various embodiments can be referred to each other.

[0053] The above description of the disclosed embodiments enables those skilled in the art to make or use the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A soft-decision method for GFSK signals based on multi-sampling-point adaptive weighting, characterized in that, Includes the following steps: S1. Perform phase difference detection on the frequency offset compensated I / Q signal and extract the signal phase change characteristics; S2. Perform three-level anti-interference processing on the signal phase change characteristics to obtain the optimized phase difference sequence; S3. Based on the predefined local ZC preamble sequence and local gold code sequence in the frame structure, perform frequency offset correction and symbol synchronization on the optimized phase difference sequence; S4. Select three sampling points—the middle, left, and right—from the 8 times oversampling points in each symbol period; S5. After interference compensation, the data from the three sampling points are weighted and fused to obtain a weighted fused sequence. S6. Based on the last n bits of the local gold code sequence, construct a reference sequence, and calculate the log-likelihood ratio between the reference sequence and the weighted fusion sequence bit by bit as the soft decision value. Subsequent decoding is based on soft decision values.

2. The soft decision method for GFSK signals based on multi-sampling point adaptive weighting as described in claim 1, characterized in that, In S1, the phase difference detection method for the I / Q signals is as follows: first, the I / Q signals are sequentially unwrapped and normalized, and then the difference between adjacent phases is calculated on the normalized signals to obtain the signal phase change characteristics.

3. The soft decision method for GFSK signals based on multi-sampling point adaptive weighting as described in claim 1, characterized in that, In S2, the three-level anti-interference processing includes: anti-interference limiting processing, 8 times sampling point accumulation processing, and phase normalization processing; The anti-interference limiting process is as follows: set a threshold value of t. If the absolute value of the difference between adjacent phases is greater than t, it is determined that there is sudden interference. At this time, the current phase difference value is clamped to t or -t by hard limiting. The 8x sampling point accumulation processing is as follows: Under 8x oversampling, each symbol period contains 8 sampling points. For each symbol period, starting from the first sampling point, the 8 sampling points are summed sequentially to obtain the effective phase energy of that symbol period. Phase normalization is performed by normalizing the signal phase change characteristics to... between.

4. The soft-decision method for GFSK signals based on multi-sampling point adaptive weighting as described in claim 1, characterized in that, S3 includes: The optimized phase difference sequence and the local ZC preamble sequence are subjected to sliding autocorrelation operation. The position corresponding to the peak value of the autocorrelation operation exceeding the preset threshold is taken as the starting position of ZC preamble timing, thus completing the preamble timing. The phase difference sequence segment corresponding to the starting position obtained from the ZC preamble timing is extracted, and the phase difference sequence segment is divided into multiple sub-segments according to the preset window length. The phase mean of each sub-segment is calculated, and the calculated phase mean is subtracted from each sampling point in the sub-segment to obtain the phase sequence after frequency offset correction. From the frequency offset corrected phase differential signal sequence, locate the end position of the ZC preamble, extract the subsequent sequence segment and perform sliding correlation operation with the local gold sequence, and take the position corresponding to the peak value of the correlation operation exceeding the preset threshold as the starting position of the synchronization word to complete the synchronization word frame synchronization.

5. The soft decision method for GFSK signals based on multi-sampling point adaptive weighting as described in claim 1, characterized in that, In S4, the middle sampling point is the maximum open position of the eye diagram, and the 4th or 5th sampling point is taken; the left sampling point is the position of the middle sampling point shifted to the left by 2 sampling clock cycles; the right sampling point is the position of the middle sampling point shifted to the right by 2 sampling clock cycles.

6. The soft decision method for GFSK signals based on multi-sampling point adaptive weighting as described in claim 1, characterized in that, S5 include: Configure the basic weights for the middle sampling point, the left sampling point, and the right sampling point; For the three sampled data points within the first symbol period, only backward interference compensation is performed; for the three sampled data points within the last symbol period, only forward interference compensation is performed; for the three sampled data points within any intermediate symbol period other than the first and second symbol periods, both forward interference compensation and backward interference compensation are performed respectively. After performing forward interference compensation or backward interference compensation on the three sampling points within each symbol period, mean filtering is then applied. Within each symbol period, the data from the three sample points after mean filtering are weighted and summed according to the configured basic weights to obtain a weighted fusion sequence.

7. The soft-decision method for GFSK signals based on multi-sampling point adaptive weighting as described in claim 6, characterized in that, In S5, the mean filter is: After forward interference compensation or backward interference compensation, three independent sampling point sequences are obtained, namely the left sampling point sequence, the middle sampling point sequence, and the right sampling point sequence. For each sampling point sequence, the average of the sampling point data of the current symbol period and the previous several symbol periods is calculated to obtain the filtered sampling point data of the current symbol period.

8. The soft decision method for GFSK signals based on multi-sampling point adaptive weighting as described in claim 6, characterized in that, In S5, forward interference compensation includes: Take the absolute value of the difference between the amplitudes of the three sampling points in the previous symbol period and the amplitudes of the three sampling points in the current symbol period at corresponding positions, and then sum them up to get the forward interference coefficient for the current symbol period. Forward interference compensation is performed point-by-point on the three sampling points of the current symbol period based on the forward interference coefficient.

9. The soft-decision method for GFSK signals based on multi-sampling point adaptive weighting as described in claim 6, characterized in that, In S5, backward interference compensation includes: Take the absolute value of the difference between the amplitudes of the three sampling points in the next symbol period and the amplitudes of the three sampling points in the current symbol period at corresponding positions, and then sum them up to get the backward interference coefficient for the current symbol period. Backward interference compensation is performed point-by-point on the three sampling points of the current symbol period based on the backward interference coefficient.

10. The soft decision method for GFSK signals based on multi-sampling point adaptive weighting as described in claim 1, characterized in that, In S6, the calculation process for the soft decision threshold includes: Based on the last n bits of the local gold code sequence, construct two reference sequences ref0 and ref1, where ref0 represents the sequence obtained by adding a 0 to the end of the reference sequence when the current bit is 0, and ref1 represents the sequence obtained by adding a 1 to the end of the reference sequence when the current bit is 1. The correlation values ​​between the weighted fusion sequence and ref0 and ref1 are calculated bit by bit and denoted as sum0 and sum1, respectively. The log-likelihood ratio is generated based on the difference between sum0 and sum1, and used as the soft decision value. If the difference between sum0 and sum1 is greater than 0, the soft decision value is biased towards bit 0; if it is less than 0, the soft decision value is biased towards bit 1.

Images