A high-frequency sampling sensor quantization noise flattening method

Abstract

This invention relates to the field of digital signal processing technology for sensors, and discloses a method for quantization noise reduction in high-frequency sampling sensors. By acquiring the original sampling sequence, sampling period, and quantization step size, an original feasible quantization interval sequence is established. Then, peak block extraction, effective period boundary construction, normalized phase establishment, phase grid sampling, local slope generation, back-projection boundary convergence, full-time reconstruction, and pruning output are performed sequentially to form a quantization noise reduction process. This scheme processes data under the constraints of period-phase consistency and quantization interval reliability, reducing the impact of quantization jumps, improving period alignment stability, preserving local variation trends, and preventing the reconstruction results from deviating from the original quantization reliability range.

Description

Technical Field

[0001] This invention relates to the field of sensor digital signal processing technology, specifically to a method for quantization noise reduction in high-frequency sampling sensors. Background Technology

[0002] High-frequency sampling sensors are widely used in vibration monitoring, periodic displacement detection, current sampling, pressure pulsation detection, and acoustic signal acquisition. In these applications, the sensor output is typically converted from analog to digital to form a discrete digital sequence, which is then processed by a processing unit for state identification, waveform analysis, or feature extraction. When the sampling frequency is high but the quantization resolution is limited, significant quantization noise can easily be superimposed on the discrete output. This is especially true for signals with periodic variations, where quantization noise can cause waveform stepping, peak distortion, phase neighborhood jitter, and local slope distortion, thus affecting the accuracy of subsequent analysis results.

[0003] In existing technologies, quantization noise is typically addressed by increasing the resolution of the analog-to-digital converter (ADC), enhancing the precision of the front-end hardware, performing moving averages on the sampled sequence, low-pass filtering, fixed-window smoothing, or interpolation reconstruction based on adjacent samples. The basic idea behind these technologies is to reduce quantization errors through hardware improvements or to weaken high-frequency jitter by averaging the discrete sequence. Some technologies also utilize oversampling to increase the number of sampling points, combined with digital filtering to obtain a smoother output. However, existing technologies still have the following shortcomings: Firstly, increasing the ADC resolution or improving the precision of the analog front-end significantly increases system cost, power consumption, and circuit complexity, hindering widespread application under limited hardware conditions. Secondly, conventional smoothing, averaging, or low-pass processing primarily relies on directly reducing fluctuations in the time neighborhood, which can easily weaken the true details in periodic signals, causing peaks, inflection points, and locally rapidly changing parts to be over-smoothed, resulting in the loss of effective information. For high-frequency sampled signals with significant periodicity, existing technologies typically fail to fully utilize the phase correspondence between different periods, and cannot uniformly map cross-period repetitive information to the same phase reference for constrained fusion, thus making it difficult to further compress the feasible range corresponding to the quantization error. In addition, some existing reconstruction methods do not strictly constrain the quantization interval boundaries of individual sampling points, and the reconstruction results may deviate from the original quantization physical boundaries, affecting the reliability and consistency of the results.

[0004] Therefore, this paper aims to propose a method for reducing quantization noise in high-frequency sampling sensors. First, each discrete sample value is restored to the original quantization feasible interval with upper and lower boundaries. Then, taking advantage of the characteristic that periodic signals have similar phase structures in different effective periods, the sampling information at the same or similar phase positions at different times and in different periods is folded back to a unified phase grid for convergence processing. Summary of the Invention

[0005] This invention provides a method for reducing quantization noise in high-frequency sampling sensors, which helps to solve the problems mentioned in the background art.

[0006] This invention provides the following technical solution: a method for reducing quantization noise in a high-frequency sampling sensor, comprising:

[0007] Collect the original sampling sequence, sampling period and quantization step size, and establish and save the original quantization feasible interval sequence according to each discrete sample value;

[0008] Extract peak blocks in chronological order, obtain representative peak sampling sequence numbers, filter adjacent representative peak sampling sequence number pairs, and form effective period boundary sequence and effective period length;

[0009] For each effective period, a normalized phase is established for the sampling point corresponding to the sampling sequence number, the number of phase grid points in each period is obtained, and a standard phase sequence of phase grid points is formed.

[0010] Establish a ring phase difference processing rule to process the phase difference between the normalized phase and the standard phase of the phase grid. Select the grid number with the smallest ring phase difference processing result within each effective period to obtain the sampled value of the foldback grid and the ring phase deviation.

[0011] The initial center value is calculated based on the sampled values ​​of the foldback grid points, and the center values ​​of the previous adjacent grid point and the next adjacent grid point are obtained to form a local phase slope.

[0012] Based on the sampled values ​​of the back projection grid points, the quantization step size, the local phase slope, and the annular phase deviation, the lower boundary and upper boundary of the back projection are established, and converged to form the lower boundary of the interval convergence, the upper boundary of the interval convergence, and the boundary median of the phase grid points;

[0013] A processing domain is constructed to participate in phase folding and temporal reconstruction. The nearest phase grid point is selected for the sampling sequence number within the processing domain. The pre-clipping reconstruction value is obtained by combining the boundary median, local phase slope and ring phase difference of the nearest phase grid point. The corresponding discrete sampling value is read from the sampling sequence number outside the processing domain as the pre-clipping reconstruction value, thus forming the pre-clipping reconstruction sequence.

[0014] Based on the original quantization feasible interval sequence, each pre-clipping reconstructed value is pruned, and the quantization noise-smoothed sequence is output in the order of sampling number.

[0015] Optionally, the acquisition of the original sampling sequence, sampling period, and quantization step size, and the establishment and storage of the original quantization feasible interval sequence according to each discrete sample value, specifically includes:

[0016] The original sampling sequence output by the high-frequency sampling sensor is acquired. The original sampling sequence output by the high-frequency sampling sensor consists of multiple discrete sampling values ​​arranged in the order of sampling. The sampling period, quantization step size, total number of sampling points, and sampling sequence number corresponding to each discrete sampling value are read. The numerical unit of each discrete sampling value is the sensor output unit.

[0017] For each discrete sample value corresponding to a sampling number, subtract half a quantization step size from the current discrete sample value to obtain the lower boundary of the original quantization feasible interval, and add half a quantization step size to the current discrete sample value to obtain the upper boundary of the original quantization feasible interval. The lower boundary and the upper boundary of the original quantization feasible interval form the original quantization feasible interval corresponding to the current discrete sample value.

[0018] The original sampling sequence and the original quantization feasible interval sequence composed of each original quantization feasible interval are saved in the order of sampling sequence number.

[0019] Optionally, the step of extracting peak blocks in chronological order, obtaining representative peak sampling numbers, filtering adjacent representative peak sampling number pairs, and forming an effective period boundary sequence and effective period length specifically includes:

[0020] Within the internal sampling sequence of the original sampling sequence, all consecutive integer blocks that cannot be extended forward or backward are extracted in ascending time order as peak blocks. The starting sampling number of each peak block is not less than the sampling number corresponding to the second discrete sampling value, and the ending sampling number of each peak block is not greater than the sampling number corresponding to the second-to-last discrete sampling value. The discrete sampling value at the starting sampling number is greater than the discrete sampling value at the sampling number preceding the starting sampling number, and the discrete sampling value at the ending sampling number is greater than the discrete sampling value at the sampling number following the ending sampling number.

[0021] For each peak block, check all integer sampling numbers from the start sampling number of the current peak block to the end sampling number of the current peak block, and retain the consecutive integer blocks whose discrete sampling values ​​at each sampling number are equal to the discrete sampling value at the start sampling number of the current peak block as peak blocks;

[0022] For each peak block, obtain the arithmetic mean of the starting sampling number and the ending sampling number of the current peak block, and use the arithmetic mean as the representative peak sampling number of the current peak block;

[0023] Traverse the representative peak sampling number pairs that are adjacent in time, calculate the interval between the next representative peak sampling number and the previous representative peak sampling number, and retain the adjacent representative peak sampling number pairs with an interval greater than or equal to four sampling numbers.

[0024] The adjacent representative peak sampling number pairs are renumbered in ascending order of time. Each pair of adjacent representative peak sampling number is used as the start and end sampling number of an effective period. The effective period boundary sequence is formed according to the renumbered start and end sampling numbers of each effective period.

[0025] For each valid period, the length of the current valid period is obtained by subtracting the starting sampling number of the current valid period from the ending sampling number of the current valid period.

[0026] Optionally, the step of establishing a normalized phase for each effective period corresponding to the sampling point, obtaining the number of phase grid points per period, and forming a standard phase sequence of phase grid points specifically includes:

[0027] Set the full cycle phase. For each valid cycle, starting from the beginning sampling number of the current valid cycle and ending at the sampling number before the end sampling number of the current valid cycle, obtain the number offset of each sampling number relative to the beginning sampling number of the current valid cycle. Calculate the ratio of the number offset to the length of the current valid cycle and multiply the ratio result by the full cycle phase to obtain the normalized phase of the sampling point corresponding to the corresponding sampling number.

[0028] Obtain the shortest effective period length among all effective period lengths. If one-third of the shortest effective period length is less than four, then take four as the number of phase grid points per period. If one-third of the shortest effective period length is not less than four, then take one-third of the shortest effective period length as the number of phase grid points per period.

[0029] The phase grid number is set according to the number of phase grid points in each cycle. Starting from the zeroth phase grid point and ending at the last phase grid point, for each phase grid point number, the phase of the complete cycle is multiplied by the current phase grid point number and then divided by the number of phase grid points in each cycle to obtain the standard phase of the current phase grid point. The phase grid points are then arranged according to their numbers to form a standard phase sequence.

[0030] Optionally, the step of establishing a ring phase difference processing rule, processing the phase difference between the normalized phase and the standard phase of the phase grid, selecting the grid point with the smallest ring phase difference processing result in each effective period, and obtaining the foldback grid point sampling value and the ring phase deviation, specifically includes:

[0031] A ring-shaped phase difference processing rule is established based on the complete cycle phase. Half of the complete cycle phase is the positive half-cycle phase. The negative half-cycle phase is obtained by taking the negative positive half-cycle phase, and the negative complete cycle phase is obtained by taking the negative complete cycle phase. When the input phase difference is greater than the positive half-cycle phase and less than the complete cycle phase, the input phase difference is subtracted from the complete cycle phase as the ring-shaped phase difference processing result. When the input phase difference is within the closed interval between the negative half-cycle phase and the positive half-cycle phase, the input phase difference is used as the ring-shaped phase difference processing result. When the input phase difference is not less than the negative complete cycle phase and less than the negative half-cycle phase, the input phase difference is added to the complete cycle phase as the ring-shaped phase difference processing result.

[0032] Within each effective period, for each phase grid point, the phase difference between the normalized phase of each sampling point corresponding to the sampling number in the current effective period and the standard phase of the current phase grid point is obtained. The ring phase difference processing result is obtained according to the ring phase difference processing rule. The ring phase difference processing results are compared according to their numerical values, and the sampling number with the smallest ring phase difference processing result is selected as the grid point sampling number at the current phase grid point in the current effective period.

[0033] Read the discrete sampled value of the grid point sampling sequence number at the current phase grid point in the current effective period in the original sampling sequence, and use the read discrete sampled value as the foldback grid point sampled value at the current phase grid point in the current effective period;

[0034] For the grid point sampling sequence number of the current effective period at the current phase grid point, the normalized phase of the sampling point corresponding to the grid point sampling sequence number of the current effective period at the current phase grid point is subtracted from the standard phase of the current phase grid point, and the corresponding ring phase difference processing result is obtained according to the ring phase difference processing rule. The corresponding ring phase difference processing result is used as the ring phase deviation of the current effective period at the current phase grid point.

[0035] Optionally, the step of calculating the initial center value based on the sampled values ​​of the folded-back grid points, obtaining the center values ​​of the previous adjacent grid point and the center values ​​of the next adjacent grid point, and forming a local phase slope specifically includes:

[0036] For each phase grid point, collect the sampled values ​​of the back-turn grid points at the current phase grid point for all valid periods, calculate the arithmetic mean of the collected back-turn grid point sampled values, and use the arithmetic mean as the initial center value of the current phase grid point;

[0037] For each phase grid point, obtain the center value of the next adjacent grid point. The center value of the next adjacent grid point of non-last phase grid points is taken as the initial center value of the next adjacent grid point of the current phase grid point, and the center value of the next adjacent grid point of the last phase grid point is taken as the initial center value of the zero phase grid point.

[0038] For each phase grid point, obtain the center value of the previous adjacent grid point. The center value of the previous adjacent grid point of the zero phase grid point is taken as the initial center value of the last phase grid point, and the center value of the previous adjacent grid point of the non-zero phase grid point is taken as the initial center value of the previous adjacent grid point of the current phase grid point.

[0039] For each phase grid point, the center value difference is obtained by subtracting the center value of the previous adjacent grid point from the center value of the next adjacent grid point of the current phase grid point. The phase interval scale is obtained by dividing four times pi by the number of phase grid points per cycle. The local phase slope of the current phase grid point is obtained by dividing the center value difference by the phase interval scale.

[0040] Optionally, the step of establishing the lower and upper boundaries of back projection based on the sampled values ​​of the back projection grid points, the quantization step size, the local phase slope, and the annular phase deviation, and converging to form the lower boundary of the interval convergence, the upper boundary of the interval convergence, and the boundary median of the phase grid points, specifically includes:

[0041] For each effective period and each phase grid point, the sampled value of the foldback grid point at the current phase grid point in the current effective period is reduced by half a quantization step to obtain the first intermediate boundary value. The local phase slope of the current phase grid point is multiplied by the annular phase deviation of the current effective period at the current phase grid point to obtain the back projection correction value. The back projection correction value is subtracted from the first intermediate boundary value to obtain the back projection lower boundary of the current effective period at the current phase grid point.

[0042] For each effective period and each phase grid point, add half a quantization step size to the sampled value of the foldback grid point at the current phase grid point in the current effective period to obtain the second intermediate boundary value. Multiply the local phase slope of the current phase grid point by the annular phase deviation at the current phase grid point in the current effective period to obtain the back projection correction value. Subtract the back projection correction value from the second intermediate boundary value to obtain the back projection upper boundary of the current effective period at the current phase grid point.

[0043] For each phase grid point, collect the back projection lower boundaries of all valid periods at the current phase grid point, and take the back projection lower boundary with the highest value as the interval convergence lower boundary of the current phase grid point;

[0044] For each phase grid point, collect the back projection upper boundaries of all valid periods at the current phase grid point, and take the back projection upper boundary with the lowest value as the interval convergence upper boundary of the current phase grid point.

[0045] For each phase grid point, add the lower boundary of the interval convergence of the current phase grid point to the upper boundary of the interval convergence of the current phase grid point, and then divide by two to obtain the boundary median of the current phase grid point.

[0046] Optionally, the process domain for phase folding and temporal reconstruction is constructed by selecting the nearest phase grid point for the sampling sequence number within the processing domain, and obtaining the pre-clipping reconstructed value by combining the boundary median, local phase slope, and ring phase difference processing results of the nearest phase grid point. For the sampling sequence number outside the processing domain, the corresponding discrete sampling value is read as the pre-clipping reconstructed value, forming a pre-clipping reconstructed sequence. Specifically, this includes:

[0047] A processing domain for phase folding and timing reconstruction is constructed, and the sampling numbers from the start sampling number to the end sampling number that fall within any valid period are included in the processing domain for phase folding and timing reconstruction.

[0048] For each sampling number in the processing domain involved in phase folding and timing reconstruction, the phase difference between the normalized phase of the sampling point corresponding to the current sampling number and the standard phase of each phase grid point is obtained. The ring phase difference processing result corresponding to each phase grid point is obtained according to the ring phase difference processing rule. The ring phase difference processing results are compared according to their numerical values. The phase grid point with the smallest ring phase difference processing result is selected as the nearest phase grid point. When there are two identical minimum ring phase difference processing result values, the phase grid point with the earlier phase grid point number is selected as the nearest phase grid point.

[0049] For each sampling number in the processing domain involved in phase folding and temporal reconstruction, the boundary median and local phase slope of the nearest phase grid point are read. The phase difference between the normalized phase of the sampling point corresponding to the current sampling number and the standard phase of the nearest phase grid point is obtained. The ring phase difference processing result of the sampling point corresponding to the current sampling number relative to the nearest phase grid point is obtained by processing according to the ring phase difference processing rule. The ring phase difference processing result of the sampling point corresponding to the current sampling number relative to the nearest phase grid point is multiplied by the local phase slope of the nearest phase grid point and then added to the boundary median of the nearest phase grid point to obtain the pre-clipping reconstructed value of the sampling point corresponding to the current sampling number.

[0050] For each sampling number outside the processing domain involved in phase folding and timing reconstruction, the discrete sampling value corresponding to the current sampling number in the original sampling sequence is directly read, and the discrete sampling value corresponding to the current sampling number in the original sampling sequence is used as the pre-clipping reconstruction value of the sampling point corresponding to the current sampling number.

[0051] All pre-cropping reconstructed values ​​are arranged in chronological order of sampling number to form a pre-cropping reconstructed sequence.

[0052] Optionally, the step of pruning each reconstructed value before pruning based on the original quantization feasible interval sequence and outputting the quantization noise-smoothed sequence in the order of sampling number specifically includes:

[0053] Establish clipping rules: when the input value is below the lower clipping bound, output the lower clipping bound; when the input value is within the closed interval between the lower and upper clipping bounds, output the input value; when the input value is above the upper clipping bound, output the upper clipping bound.

[0054] For each sampling number, subtract half a quantization step from the discrete sample value corresponding to the current sampling number to obtain the lower boundary of the original quantization feasible interval corresponding to the current sampling number. Use the lower boundary of the original quantization feasible interval corresponding to the current sampling number as the lower bound of the clipping. Add half a quantization step to the discrete sample value corresponding to the current sampling number to obtain the upper boundary of the original quantization feasible interval corresponding to the current sampling number. Use the upper boundary of the original quantization feasible interval corresponding to the current sampling number as the upper bound of the clipping. Clip the reconstructed value before clipping corresponding to the current sampling number according to the clipping processing rules to obtain the output value after quantization noise reduction corresponding to the current sampling number.

[0055] Arrange all the quantized noise-smoothed output values ​​in the order of sampling number, and output the quantized noise-smoothed sequence.

[0056] The present invention has the following beneficial effects:

[0057] 1. Each discrete sampled value is expanded into a reliable range defined by the quantization step size, ensuring that subsequent smoothing processing is always constrained by both the original sampled data and the quantization accuracy. This scheme first establishes a sequence of feasible original quantization intervals, defining clear upper and lower boundaries for each sampling number. Subsequent reconstruction and cropping are based on this interval, thus avoiding excessive deviation in the smoothing results. By transforming sensor quantization errors into operable boundary constraints, the smoothing process possesses both noise reduction capabilities and preserves the physical reliability of the original data. This approach is particularly suitable for scenarios with significant quantization steps at high-frequency sampling and reliable single-point sampled values ​​but unsmooth continuous trends.

[0058] 2. By extracting peak blocks and obtaining representative peak sampling numbers, peak regions that may contain plateau peaks, continuous peaks, or adjacent sampled values ​​are uniformly transformed into a stable periodic boundary basis. This scheme identifies peak regions through continuous integer blocks and uses the midpoint between the start and end sampling numbers as the representative peak sampling number, which can reduce the impact of quantization plateaus and local jumps on periodic boundaries. This improves the stability and repeatability of period division, ensuring that each subsequent effective period has clear start and end boundaries, providing a reliable foundation for phase normalization, grid point sampling, and cross-period information fusion. It also solves the problem of unstable period identification in existing methods when peaks are not sharp or sampled values ​​are stepped.

[0059] 3. This scheme maps effective periods of different lengths to the same period phase scale and determines the number of phase grid points based on the effective period length, enabling comparison and fusion of sampling points within different periods under a unified phase framework. By establishing a normalized phase for the sampling sequence number within each effective period, even if the number of sampling points differs across periods, they can be unified into the same phase reference system. Converting time sequence alignment to phase alignment makes subsequent phase grid sampling and cross-period statistics more accurate, preserving the structural characteristics of periodic signals while reducing misalignment errors caused by changes in period length. This solves the problem of insufficient adaptability to periodic fluctuations in existing fixed-window or fixed-sequence processing methods.

[0060] 4. The phase difference between the normalized phase of the sampling point corresponding to the sampling sequence number and the standard phase of the phase grid is processed through a ring-shaped phase difference processing rule. This ensures that sampling points at the beginning and end of the phase are correctly included in the adjacency judgment, and the closest grid sampling sequence number is selected for each phase grid point within each effective period. This scheme uses a ring-shaped foldback concept to convert the phase difference within a complete period into a processing result suitable for the continuity of the period boundary. The corresponding discrete sampling values ​​are then read as the foldback grid point sampling values, and the ring-shaped phase deviation is obtained. This improves the phase accuracy of grid point sampling, avoids the incorrect exclusion of sampling points near the period boundary, and makes the data at the same phase grid point in different effective periods more comparable. It solves the problems of discontinuity at the beginning and end of the phase, large sampling deviation, and unstable cross-period fusion in existing period alignment methods.

[0061] 5. An initial center value is formed based on the sampled values ​​of the foldback grid points at the same phase grid point for each effective period. This center value is then combined with the center values ​​of the preceding and following adjacent grid points to obtain the local phase change trend. This scheme constructs a local slope using the center values ​​of adjacent phase grid points, ensuring that subsequent backprojection and reconstruction consider not only the center amplitude at the grid point but also the influence of the phase offset of the sampling point relative to the grid point on the amplitude. This enhances the ability of the reconstruction process to express local waveform trends, avoids the result from being flattened into an overly blunted average curve, and solves the problem that existing filtering methods easily lose edge variations, peak and valley shapes, and local continuous trends.

[0062] 6. By combining the sampled values ​​from the back-projection grid, the quantization step size, the local phase slope, and the ring phase bias, the lower and upper boundaries of the back-projection are formed respectively. Then, the boundary information of all effective periods is aggregated to obtain the lower boundary, upper boundary, and median of the convergence interval for each phase grid. This scheme uses residual back-projection to back-project the sampled values ​​deviating from the grid along the local phase slope direction, enabling boundary aggregation of data from different effective periods at more consistent phase positions. By incorporating quantization boundaries, phase bias, and local trends into the calculation, the resulting median of the boundaries is closer to the reliable value of the true periodic waveform at the standard phase grid, solving the problems of existing cross-period averaging being susceptible to phase shifts and unable to simultaneously account for quantization constraints and local trends.

[0063] 7. Instead of processing only the data at the phase grid points, this approach constructs a processing domain for phase folding and temporal reconstruction. For each sampling number within the processing domain, the nearest phase grid point is selected. The processing results, including the boundary median, local phase slope, and the annular phase difference between the current sampling number and the nearest phase grid point, are combined to generate the pre-clipping reconstructed value for each sampling point. This scheme performs phase-related reconstruction on each sampling number within the processing domain while retaining the original discrete sampling values ​​as pre-clipping reconstructed values ​​for sampling numbers outside the processing domain. This ensures the reconstruction process covers the entire temporal sequence and has boundary safety. It achieves a continuous mapping from the phase grid point results to the complete sampling sequence, utilizing periodic commonality to suppress quantization noise and avoiding forced reconstruction of unsuitable regions. This solves the problems of incomplete reconstruction range, coarse processing of non-grid point data, and insufficient temporal continuity in existing methods.

[0064] 8. After obtaining the pre-clipping reconstructed sequence, the reconstructed values ​​are not directly used as the final output. Instead, boundary pruning is performed point-by-point on each pre-clipping reconstructed value based on the original quantization feasible interval sequence, ensuring that the final output value always falls within the quantization confidence interval of the corresponding discrete sampled value. This scheme uses pruning rules to pull reconstructed values ​​below the lower boundary back to the lower boundary, and reconstructed values ​​above the upper boundary back to the upper boundary, directly retaining reconstructed values ​​within the interval. This provides a final safety constraint for the entire smoothing result, ensuring that the output sequence has a smooth and continuous trend without deviating from the original quantization observation range. This solves the problems of overshoot, undershoot, physically unbelievable output, and mismatch with the original sampling accuracy that may occur in existing smoothing methods. Attached Figure Description

[0065] Figure 1 This is a schematic diagram of the process of the present invention.

[0066] Figure 2 This is a schematic diagram of the original quantization feasible interval sequence construction process of the present invention.

[0067] Figure 3 This is a schematic diagram illustrating the construction process of the effective period boundary sequence and effective period length of the present invention.

[0068] Figure 4 This is a schematic diagram of the annular phase difference processing and grid sampling process of the present invention.

[0069] Figure 5 This is a schematic diagram of the process for constructing the processing domain and generating the reconstructed sequence before pruning according to the present invention. Detailed Implementation

[0070] 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.

[0071] Example, refer to Figure 1 A method for reducing quantization noise in a high-frequency sampling sensor, comprising:

[0072] Collect the original sampling sequence, sampling period and quantization step size, and establish and save the original quantization feasible interval sequence according to each discrete sample value;

[0073] Extract peak blocks in chronological order, obtain representative peak sampling sequence numbers, filter adjacent representative peak sampling sequence number pairs, and form effective period boundary sequence and effective period length;

[0074] For each effective period, a normalized phase is established for the sampling point corresponding to the sampling sequence number, the number of phase grid points in each period is obtained, and a standard phase sequence of phase grid points is formed.

[0075] Establish a ring phase difference processing rule to process the phase difference between the normalized phase and the standard phase of the phase grid. Select the grid number with the smallest ring phase difference processing result within each effective period to obtain the sampled value of the foldback grid and the ring phase deviation.

[0076] The initial center value is calculated based on the sampled values ​​of the foldback grid points, and the center values ​​of the previous adjacent grid point and the next adjacent grid point are obtained to form a local phase slope.

[0077] Based on the sampled values ​​of the back projection grid points, the quantization step size, the local phase slope, and the annular phase deviation, the lower boundary and upper boundary of the back projection are established, and converged to form the lower boundary of the interval convergence, the upper boundary of the interval convergence, and the boundary median of the phase grid points;

[0078] A processing domain is constructed to participate in phase folding and temporal reconstruction. The nearest phase grid point is selected for the sampling sequence number within the processing domain. The pre-clipping reconstruction value is obtained by combining the boundary median, local phase slope and ring phase difference of the nearest phase grid point. The corresponding discrete sampling value is read from the sampling sequence number outside the processing domain as the pre-clipping reconstruction value, thus forming the pre-clipping reconstruction sequence.

[0079] Based on the original quantization feasible interval sequence, each pre-clipping reconstructed value is pruned, and the quantization noise-smoothed sequence is output in the order of sampling number.

[0080] First, by acquiring the original sampling sequence, sampling period, and quantization step size, an original quantization feasible interval sequence is established. Then, through peak block extraction, effective period boundary construction, normalized phase establishment, phase grid sampling, local slope generation, back-projection boundary convergence, full-time reconstruction, and final pruning output, a complete processing chain is formed from the original discrete sampling data to the smoothed sequence. By establishing the original quantization feasible interval sequence, the problem that existing direct filtering methods easily ignore the sensor quantization confidence range is solved. By constructing the effective period boundary sequence and normalized phase, the problem that periodic high-frequency data is difficult to stably align between different periods is solved. By phase grid sampling, local phase slope, and back-projection boundary convergence, the problem that simple mean smoothing easily weakens the local trend of the waveform is solved. By pre-pruning reconstruction and original quantization feasible interval pruning, the problem that the reconstruction result may deviate from the original quantization constraints is solved. Compared with existing direct low-pass filtering, moving average, or fixed window smoothing, this scheme does not simply weaken fluctuations, but smooths them under the joint constraints of periodic phase consistency and quantization interval reliability. Therefore, it can reduce the impact of quantization jumps while better preserving the phase structure, local variation trend, and reliable boundary of the sampled data of the periodic signal. It is suitable for scenarios in high-frequency sampling sensors where quantization noise is obvious but the original signal still has periodic regularity.

[0081] Reference Figure 2 The acquisition of the original sampling sequence, sampling period, and quantization step size, and the establishment and storage of the original quantization feasible interval sequence according to each discrete sample value, specifically includes:

[0082] The original sampling sequence output by the high-frequency sampling sensor is acquired. The original sampling sequence output by the high-frequency sampling sensor consists of multiple discrete sampling values ​​arranged in the order of sampling. The sampling period, quantization step size, total number of sampling points, and sampling sequence number corresponding to each discrete sampling value are read. The numerical unit of each discrete sampling value is the sensor output unit.

[0083] For each discrete sample value corresponding to a sampling number, subtract half a quantization step size from the current discrete sample value to obtain the lower boundary of the original quantization feasible interval, and add half a quantization step size to the current discrete sample value to obtain the upper boundary of the original quantization feasible interval. The lower boundary and the upper boundary of the original quantization feasible interval form the original quantization feasible interval corresponding to the current discrete sample value.

[0084] The original sampling sequence and the original quantization feasible interval sequence composed of each original quantization feasible interval are saved in the order of sampling sequence number.

[0085] Acquire the output sequence of the high-frequency sampling sensor: and read the sampling period With quantization step size ;in, This is the original sampling sequence, where each element is a discrete sample value arranged in the order of sampling. For the first One original sampled value, in sensor output units; For discrete sampling sequence numbers; This represents the total number of sampling points;

[0086] For each sampling number The quantization interval function is established as follows: ;in, For the first The quantization feasible interval corresponding to each sample value;

[0087] The original sampling sequence and the quantization interval sequence are preserved.

[0088] Reference Figure 3 The step of extracting peak blocks in chronological order, obtaining representative peak sampling numbers, filtering adjacent representative peak sampling number pairs, and forming an effective period boundary sequence and effective period length specifically includes:

[0089] Within the internal sampling sequence of the original sampling sequence, all consecutive integer blocks that cannot be extended forward or backward are extracted in ascending time order as peak blocks. The starting sampling number of each peak block is not less than the sampling number corresponding to the second discrete sampling value, and the ending sampling number of each peak block is not greater than the sampling number corresponding to the second-to-last discrete sampling value. The discrete sampling value at the starting sampling number is greater than the discrete sampling value at the sampling number preceding the starting sampling number, and the discrete sampling value at the ending sampling number is greater than the discrete sampling value at the sampling number following the ending sampling number.

[0090] For each peak block, check all integer sampling numbers from the start sampling number of the current peak block to the end sampling number of the current peak block, and retain the consecutive integer blocks whose discrete sampling values ​​at each sampling number are equal to the discrete sampling value at the start sampling number of the current peak block as peak blocks;

[0091] For each peak block, obtain the arithmetic mean of the starting sampling number and the ending sampling number of the current peak block, and use the arithmetic mean as the representative peak sampling number of the current peak block;

[0092] Traverse the representative peak sampling number pairs that are adjacent in time, calculate the interval between the next representative peak sampling number and the previous representative peak sampling number, and retain the adjacent representative peak sampling number pairs with an interval greater than or equal to four sampling numbers.

[0093] The adjacent representative peak sampling number pairs are renumbered in ascending order of time. Each pair of adjacent representative peak sampling number is used as the start and end sampling number of an effective period. The effective period boundary sequence is formed according to the renumbered start and end sampling numbers of each effective period.

[0094] For each valid period, the length of the current valid period is obtained by subtracting the starting sampling number of the current valid period from the ending sampling number of the current valid period.

[0095] Extract all maximal contiguous integer blocks that satisfy the following conditions in ascending time order within the sampling sequence: Each peak block satisfies:

[0096] , , ;in, For the first The consecutive integer sequence range corresponding to each peak block; The peak block number; For the first The starting sampling sequence number of each peak block; For the first The termination sampling sequence number of each peak block; The total number of peak blocks;

[0097] And for any satisfying integers All of them have: ;

[0098] For each peak block, calculate the representative peak sampling number as follows:

[0099] , ;in, For the first The representative peak sampling number of each peak block;

[0100] All will be satisfied And satisfy Adjacent representative peak pairs Renumbered in ascending order of time: ;in, The candidate period boundary pair is formed by two adjacent sample numbers representing the peak values; For the first Each peak block corresponds to a peak sampling sequence number; For the first The start and end sampling numbers of each valid period; For the first The starting sampling sequence number of each valid cycle; For the first The effective cycle termination sampling sequence number; The valid period number; Total number of effective periods;

[0101] For each valid period number The effective period length is calculated as follows: ;in, For the first Effective cycle length.

[0102] The process of establishing a normalized phase for each effective period corresponding to the sampling point, obtaining the number of phase grid points per period, and forming a standard phase sequence of phase grid points specifically includes:

[0103] Set the full cycle phase. For each valid cycle, starting from the beginning sampling number of the current valid cycle and ending at the sampling number before the end sampling number of the current valid cycle, obtain the number offset of each sampling number relative to the beginning sampling number of the current valid cycle. Calculate the ratio of the number offset to the length of the current valid cycle and multiply the ratio result by the full cycle phase to obtain the normalized phase of the sampling point corresponding to the corresponding sampling number.

[0104] Obtain the shortest effective period length among all effective period lengths. If one-third of the shortest effective period length is less than four, then take four as the number of phase grid points per period. If one-third of the shortest effective period length is not less than four, then take one-third of the shortest effective period length as the number of phase grid points per period.

[0105] The phase grid number is set according to the number of phase grid points in each cycle. Starting from the zeroth phase grid point and ending at the last phase grid point, for each phase grid point number, the phase of the complete cycle is multiplied by the current phase grid point number and then divided by the number of phase grid points in each cycle to obtain the standard phase of the current phase grid point. The phase grid points are then arranged according to their numbers to form a standard phase sequence.

[0106] For each valid period number Establish the phase mapping function as follows:

[0107] , ;in, To process the first in the domain Normalized phase of each sampling point; Number the discrete sampling points;

[0108] The phase grid number is calculated as follows: ;in, This represents the number of phase grid points per cycle.

[0109] Establish a standard phase grid sequence, specifically as follows: , ;in, For the first The standard phase of each phase grid point; This represents the phase grid point number.

[0110] Reference Figure 4 The establishment of the ring phase difference processing rule, which processes the phase difference between the normalized phase and the standard phase of the phase grid, selects the grid point with the smallest ring phase difference processing result within each effective period, and obtains the foldback grid point sampling value and the ring phase deviation, specifically includes:

[0111] A ring-shaped phase difference processing rule is established based on the complete cycle phase. Half of the complete cycle phase is the positive half-cycle phase. The negative half-cycle phase is obtained by taking the negative positive half-cycle phase, and the negative complete cycle phase is obtained by taking the negative complete cycle phase. When the input phase difference is greater than the positive half-cycle phase and less than the complete cycle phase, the input phase difference is subtracted from the complete cycle phase as the ring-shaped phase difference processing result. When the input phase difference is within the closed interval between the negative half-cycle phase and the positive half-cycle phase, the input phase difference is used as the ring-shaped phase difference processing result. When the input phase difference is not less than the negative complete cycle phase and less than the negative half-cycle phase, the input phase difference is added to the complete cycle phase as the ring-shaped phase difference processing result.

[0112] Within each effective period, for each phase grid point, the phase difference between the normalized phase of each sampling point corresponding to the sampling number in the current effective period and the standard phase of the current phase grid point is obtained. The ring phase difference processing result is obtained according to the ring phase difference processing rule. The ring phase difference processing results are compared according to their numerical values, and the sampling number with the smallest ring phase difference processing result is selected as the grid point sampling number at the current phase grid point in the current effective period.

[0113] Read the discrete sampled value of the grid point sampling sequence number at the current phase grid point in the current effective period in the original sampling sequence, and use the read discrete sampled value as the foldback grid point sampled value at the current phase grid point in the current effective period;

[0114] For the grid point sampling sequence number of the current effective period at the current phase grid point, the normalized phase of the sampling point corresponding to the grid point sampling sequence number of the current effective period at the current phase grid point is subtracted from the standard phase of the current phase grid point, and the corresponding ring phase difference processing result is obtained according to the ring phase difference processing rule. The corresponding ring phase difference processing result is used as the ring phase deviation of the current effective period at the current phase grid point.

[0115] The ring phase difference function is established as follows: ;in, It is a ring phase difference function. For input variables;

[0116] In each effective period Within, for each phase grid point The grid sampling function is constructed as follows:

[0117] ;in, For the first The closest to the first in the effective cycle The sampling sequence number of each standard phase grid point;

[0118] The extracted grid sample values ​​after the return are: ;in, For the first The effective period in the first The sampled values ​​selected at each phase grid point; For sampling sequence number The original sampled values;

[0119] The grid point ring phase deviation is calculated as follows: ;in, Sampling sequence number Corresponding sampling point relative to the first Ring phase deviation at standard phase grid points.

[0120] The process of calculating the initial center value based on the sampled values ​​of the foldback grid points, obtaining the center values ​​of the previous adjacent grid point and the center values ​​of the next adjacent grid point, and forming a local phase slope specifically includes:

[0121] For each phase grid point, collect the sampled values ​​of the back-turn grid points at the current phase grid point for all valid periods, calculate the arithmetic mean of the collected back-turn grid point sampled values, and use the arithmetic mean as the initial center value of the current phase grid point;

[0122] For each phase grid point, obtain the center value of the next adjacent grid point. The center value of the next adjacent grid point of non-last phase grid points is taken as the initial center value of the next adjacent grid point of the current phase grid point, and the center value of the next adjacent grid point of the last phase grid point is taken as the initial center value of the zero phase grid point.

[0123] For each phase grid point, obtain the center value of the previous adjacent grid point. The center value of the previous adjacent grid point of the zero phase grid point is taken as the initial center value of the last phase grid point, and the center value of the previous adjacent grid point of the non-zero phase grid point is taken as the initial center value of the previous adjacent grid point of the current phase grid point.

[0124] For each phase grid point, the center value difference is obtained by subtracting the center value of the previous adjacent grid point from the center value of the next adjacent grid point of the current phase grid point. The phase interval scale is obtained by dividing four times pi by the number of phase grid points per cycle. The local phase slope of the current phase grid point is obtained by dividing the center value difference by the phase interval scale.

[0125] For each phase grid point The initial central value function is constructed as follows: ;in, For the first Initial center values ​​for each phase grid point; For the first The effective period in the first The foldback sampled values ​​at each phase grid point;

[0126] The center values ​​of adjacent grid points in a cycle are:

[0127] , ;in, For the first The center value of the next adjacent grid point of each phase grid point; For the first The center value of the previous adjacent grid point of each phase grid point;

[0128] The local phase slope is calculated as follows: ;in, For the first Local phase slope of each phase grid point.

[0129] The process of establishing the lower and upper boundaries of back projection based on the sampled values ​​of the return grid points, the quantization step size, the local phase slope, and the annular phase deviation, and converging to form the lower boundary of the interval convergence, the upper boundary of the interval convergence, and the boundary median of the phase grid points, specifically includes:

[0130] For each effective period and each phase grid point, the sampled value of the foldback grid point at the current phase grid point in the current effective period is reduced by half a quantization step to obtain the first intermediate boundary value. The local phase slope of the current phase grid point is multiplied by the annular phase deviation of the current effective period at the current phase grid point to obtain the back projection correction value. The back projection correction value is subtracted from the first intermediate boundary value to obtain the back projection lower boundary of the current effective period at the current phase grid point.

[0131] For each effective period and each phase grid point, add half a quantization step size to the sampled value of the foldback grid point at the current phase grid point in the current effective period to obtain the second intermediate boundary value. Multiply the local phase slope of the current phase grid point by the annular phase deviation at the current phase grid point in the current effective period to obtain the back projection correction value. Subtract the back projection correction value from the second intermediate boundary value to obtain the back projection upper boundary of the current effective period at the current phase grid point.

[0132] For each phase grid point, collect the back projection lower boundaries of all valid periods at the current phase grid point, and take the back projection lower boundary with the highest value as the interval convergence lower boundary of the current phase grid point;

[0133] For each phase grid point, collect the back projection upper boundaries of all valid periods at the current phase grid point, and take the back projection upper boundary with the lowest value as the interval convergence upper boundary of the current phase grid point.

[0134] For each phase grid point, add the lower boundary of the interval convergence of the current phase grid point to the upper boundary of the interval convergence of the current phase grid point, and then divide by two to obtain the boundary median of the current phase grid point.

[0135] For each effective period and each phase grid point The lower boundary function of the back projection is established as follows:

[0136] ;in, For the first The effective period in the first The back-projected lower boundary at each phase grid point;

[0137] For each effective period and each phase grid point The upper boundary function of the back projection is established as follows:

[0138] ;in, For the first The effective period in the first The upper boundary of the back projection at each phase grid point;

[0139] For each phase grid point Calculate the convergence boundary of the interval, specifically as follows:

[0140] , ;in, For the first The lower boundary of the interval convergence of each phase grid point; For the first The interval converges at the upper boundary of each phase grid point;

[0141] The phase grid boundary median function is constructed as follows: ;in, For the first The boundary median of each phase grid point.

[0142] Reference Figure 5 The process domain involved in phase folding and temporal reconstruction is constructed. Within the processing domain, the nearest phase grid point is selected based on the sampling sequence number. The pre-clipping reconstruction value is obtained by combining the boundary median, local phase slope, and ring phase difference processing results of the nearest phase grid point. For sampling sequences outside the processing domain, the corresponding discrete sampling values ​​are read as pre-clipping reconstruction values, forming a pre-clipping reconstruction sequence. Specifically, this includes:

[0143] A processing domain for phase folding and timing reconstruction is constructed, and the sampling numbers from the start sampling number to the end sampling number that fall within any valid period are included in the processing domain for phase folding and timing reconstruction.

[0144] For each sampling number in the processing domain involved in phase folding and timing reconstruction, the phase difference between the normalized phase of the sampling point corresponding to the current sampling number and the standard phase of each phase grid point is obtained. The ring phase difference processing result corresponding to each phase grid point is obtained according to the ring phase difference processing rule. The ring phase difference processing results are compared according to their numerical values. The phase grid point with the smallest ring phase difference processing result is selected as the nearest phase grid point. When there are two identical minimum ring phase difference processing result values, the phase grid point with the earlier phase grid point number is selected as the nearest phase grid point.

[0145] For each sampling number in the processing domain involved in phase folding and temporal reconstruction, the boundary median and local phase slope of the nearest phase grid point are read. The phase difference between the normalized phase of the sampling point corresponding to the current sampling number and the standard phase of the nearest phase grid point is obtained. The ring phase difference processing result of the sampling point corresponding to the current sampling number relative to the nearest phase grid point is obtained by processing according to the ring phase difference processing rule. The ring phase difference processing result of the sampling point corresponding to the current sampling number relative to the nearest phase grid point is multiplied by the local phase slope of the nearest phase grid point and then added to the boundary median of the nearest phase grid point to obtain the pre-clipping reconstructed value of the sampling point corresponding to the current sampling number.

[0146] For each sampling number outside the processing domain involved in phase folding and timing reconstruction, the discrete sampling value corresponding to the current sampling number in the original sampling sequence is directly read, and the discrete sampling value corresponding to the current sampling number in the original sampling sequence is used as the pre-clipping reconstruction value of the sampling point corresponding to the current sampling number.

[0147] All pre-cropping reconstructed values ​​are arranged in chronological order of sampling number to form a pre-cropping reconstructed sequence.

[0148] The processing domain is constructed as follows: ;in, This is the processing domain involved in phase foldback and timing reconstruction; It is an existential quantifier;

[0149] For each satisfied The sampling sequence number is used to calculate the nearest phase grid point number:

[0150] When two minimum values ​​exist, the smaller phase grid point index is taken; where, To process the first in the domain The nearest phase grid point number corresponding to each sampling point;

[0151] For each sampling number The pre-cropping reconstruction function is constructed as follows:

[0152] ;in, For the first Reconstructed values ​​of each sampling point before cropping; For the first The boundary median corresponding to the nearest phase grid point of each sampling point; For the first The local phase slope corresponding to the nearest phase grid point of each sampling point; For the first The standard phase of each phase grid point;

[0153] The reconstruction sequence is as follows: ;in, This is the reconstructed sequence before pruning; This represents all reconstructed values ​​before cropping, arranged in order of sampling sequence number.

[0154] The process of pruning the reconstructed values ​​before pruning based on the original quantization feasible interval sequence and outputting the quantization noise-smoothed sequence in the order of sampling number specifically includes:

[0155] Establish clipping rules: when the input value is below the lower clipping bound, output the lower clipping bound; when the input value is within the closed interval between the lower and upper clipping bounds, output the input value; when the input value is above the upper clipping bound, output the upper clipping bound.

[0156] For each sampling number, subtract half a quantization step from the discrete sample value corresponding to the current sampling number to obtain the lower boundary of the original quantization feasible interval corresponding to the current sampling number. Use the lower boundary of the original quantization feasible interval corresponding to the current sampling number as the lower bound of the clipping. Add half a quantization step to the discrete sample value corresponding to the current sampling number to obtain the upper boundary of the original quantization feasible interval corresponding to the current sampling number. Use the upper boundary of the original quantization feasible interval corresponding to the current sampling number as the upper bound of the clipping. Clip the reconstructed value before clipping corresponding to the current sampling number according to the clipping processing rules to obtain the output value after quantization noise reduction corresponding to the current sampling number.

[0157] Arrange all the quantized noise-smoothed output values ​​in the order of sampling number, and output the quantized noise-smoothed sequence.

[0158] Construct the clipping function as follows ;in, Input values ​​for the trimming function. This is the lower bound of the clipping function. This is the upper bound of the clipping function;

[0159] For each sampling number The final output is calculated as follows:

[0160] ;in, For the first Output value after quantization noise reduction at each sampling point;

[0161] The output sequence after quantization noise reduction is as follows: ;in, This is the output sequence after quantization noise reduction; This represents all the final output values ​​arranged in order of sampling sequence number.

[0162] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0163] The above description is only a preferred embodiment of the present invention. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the technical principles of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A method for reducing quantization noise in a high-frequency sampling sensor, characterized in that, include: Collect the original sampling sequence, sampling period and quantization step size, and establish and save the original quantization feasible interval sequence according to each discrete sample value; Extract peak blocks in chronological order, obtain representative peak sampling sequence numbers, filter adjacent representative peak sampling sequence number pairs, and form effective period boundary sequence and effective period length; For each effective period, a normalized phase is established for the sampling point corresponding to the sampling sequence number, the number of phase grid points in each period is obtained, and a standard phase sequence of phase grid points is formed. Establish a ring phase difference processing rule to process the phase difference between the normalized phase and the standard phase of the phase grid. Select the grid number with the smallest ring phase difference processing result within each effective period to obtain the sampled value of the foldback grid and the ring phase deviation. The initial center value is calculated based on the sampled values ​​of the foldback grid points, and the center values ​​of the previous adjacent grid point and the next adjacent grid point are obtained to form a local phase slope. Based on the sampled values ​​of the back projection grid points, the quantization step size, the local phase slope, and the annular phase deviation, the lower boundary and upper boundary of the back projection are established, and converged to form the lower boundary of the interval convergence, the upper boundary of the interval convergence, and the boundary median of the phase grid points; A processing domain is constructed to participate in phase folding and temporal reconstruction. The nearest phase grid point is selected for the sampling sequence number within the processing domain. The pre-clipping reconstruction value is obtained by combining the boundary median, local phase slope and ring phase difference of the nearest phase grid point. The corresponding discrete sampling value is read from the sampling sequence number outside the processing domain as the pre-clipping reconstruction value, thus forming the pre-clipping reconstruction sequence. Based on the original quantization feasible interval sequence, each pre-clipping reconstructed value is pruned, and the quantization noise-smoothed sequence is output in the order of sampling number.

2. The method for quantization noise reduction of a high-frequency sampling sensor according to claim 1, characterized in that, The acquisition of the original sampling sequence, sampling period, and quantization step size, and the establishment and storage of the original quantization feasible interval sequence according to each discrete sample value, specifically includes: The original sampling sequence output by the high-frequency sampling sensor is acquired. The original sampling sequence output by the high-frequency sampling sensor consists of multiple discrete sampling values ​​arranged in the order of sampling. The sampling period, quantization step size, total number of sampling points, and sampling sequence number corresponding to each discrete sampling value are read. The numerical unit of each discrete sampling value is the sensor output unit. For each discrete sample value corresponding to a sampling number, subtract half a quantization step size from the current discrete sample value to obtain the lower boundary of the original quantization feasible interval, and add half a quantization step size to the current discrete sample value to obtain the upper boundary of the original quantization feasible interval. The lower boundary and the upper boundary of the original quantization feasible interval form the original quantization feasible interval corresponding to the current discrete sample value. The original sampling sequence and the original quantization feasible interval sequence composed of each original quantization feasible interval are saved in the order of sampling sequence number.

3. The method for quantization noise reduction of a high-frequency sampling sensor according to claim 2, characterized in that, The step of extracting peak blocks in chronological order, obtaining representative peak sampling numbers, filtering adjacent representative peak sampling number pairs, and forming an effective period boundary sequence and effective period length specifically includes: Within the internal sampling sequence of the original sampling sequence, all consecutive integer blocks that cannot be extended forward or backward are extracted in ascending time order as peak blocks. The starting sampling number of each peak block is not less than the sampling number corresponding to the second discrete sampling value, and the ending sampling number of each peak block is not greater than the sampling number corresponding to the second-to-last discrete sampling value. The discrete sampling value at the starting sampling number is greater than the discrete sampling value at the sampling number preceding the starting sampling number, and the discrete sampling value at the ending sampling number is greater than the discrete sampling value at the sampling number following the ending sampling number. For each peak block, check all integer sampling numbers from the start sampling number of the current peak block to the end sampling number of the current peak block, and retain the consecutive integer blocks whose discrete sampling values ​​at each sampling number are equal to the discrete sampling value at the start sampling number of the current peak block as peak blocks; For each peak block, obtain the arithmetic mean of the starting sampling number and the ending sampling number of the current peak block, and use the arithmetic mean as the representative peak sampling number of the current peak block; Traverse the representative peak sampling number pairs that are adjacent in time, calculate the interval between the next representative peak sampling number and the previous representative peak sampling number, and retain the adjacent representative peak sampling number pairs with an interval greater than or equal to four sampling numbers. The adjacent representative peak sampling number pairs are renumbered in ascending order of time. Each pair of adjacent representative peak sampling number is used as the start and end sampling number of an effective period. The effective period boundary sequence is formed according to the renumbered start and end sampling numbers of each effective period. For each valid period, the length of the current valid period is obtained by subtracting the starting sampling number of the current valid period from the ending sampling number of the current valid period.

4. The method for quantization noise reduction of a high-frequency sampling sensor according to claim 3, characterized in that, The process of establishing a normalized phase for each effective period corresponding to the sampling point, obtaining the number of phase grid points per period, and forming a standard phase sequence of phase grid points specifically includes: Set the full cycle phase. For each valid cycle, starting from the beginning sampling number of the current valid cycle and ending at the sampling number before the end sampling number of the current valid cycle, obtain the number offset of each sampling number relative to the beginning sampling number of the current valid cycle. Calculate the ratio of the number offset to the length of the current valid cycle and multiply the ratio result by the full cycle phase to obtain the normalized phase of the sampling point corresponding to the corresponding sampling number. Obtain the shortest effective period length among all effective period lengths. If one-third of the shortest effective period length is less than four, then take four as the number of phase grid points per period. If one-third of the shortest effective period length is not less than four, then take one-third of the shortest effective period length as the number of phase grid points per period. The phase grid number is set according to the number of phase grid points in each cycle. Starting from the zeroth phase grid point and ending at the last phase grid point, for each phase grid point number, the phase of the complete cycle is multiplied by the current phase grid point number and then divided by the number of phase grid points in each cycle to obtain the standard phase of the current phase grid point. The phase grid points are then arranged according to their numbers to form a standard phase sequence.

5. The method for quantization noise reduction of a high-frequency sampling sensor according to claim 4, characterized in that, The establishment of the ring phase difference processing rule processes the phase difference between the normalized phase and the standard phase of the phase grid. Within each effective period, the grid point with the smallest ring phase difference processing result is selected for sampling. The sampled values ​​of the foldback grid points and the ring phase deviation are then obtained. Specifically, this includes: A ring-shaped phase difference processing rule is established based on the complete cycle phase. Half of the complete cycle phase is the positive half-cycle phase. The negative half-cycle phase is obtained by taking the negative positive half-cycle phase, and the negative complete cycle phase is obtained by taking the negative complete cycle phase. When the input phase difference is greater than the positive half-cycle phase and less than the complete cycle phase, the input phase difference is subtracted from the complete cycle phase as the ring-shaped phase difference processing result. When the input phase difference is within the closed interval between the negative half-cycle phase and the positive half-cycle phase, the input phase difference is used as the ring-shaped phase difference processing result. When the input phase difference is not less than the negative complete cycle phase and less than the negative half-cycle phase, the input phase difference is added to the complete cycle phase as the ring-shaped phase difference processing result. Within each effective period, for each phase grid point, the phase difference between the normalized phase of each sampling point corresponding to the sampling number in the current effective period and the standard phase of the current phase grid point is obtained. The ring phase difference processing result is obtained according to the ring phase difference processing rule. The ring phase difference processing results are compared according to their numerical values, and the sampling number with the smallest ring phase difference processing result is selected as the grid point sampling number at the current phase grid point in the current effective period. Read the discrete sampled value of the grid point sampling sequence number at the current phase grid point in the current effective period in the original sampling sequence, and use the read discrete sampled value as the foldback grid point sampled value at the current phase grid point in the current effective period; For the grid point sampling sequence number of the current effective period at the current phase grid point, the normalized phase of the sampling point corresponding to the grid point sampling sequence number of the current effective period at the current phase grid point is subtracted from the standard phase of the current phase grid point, and the corresponding ring phase difference processing result is obtained according to the ring phase difference processing rule. The corresponding ring phase difference processing result is used as the ring phase deviation of the current effective period at the current phase grid point.

6. The method for quantization noise reduction of a high-frequency sampling sensor according to claim 5, characterized in that, The process of calculating the initial center value based on the sampled values ​​of the foldback grid points, obtaining the center values ​​of the previous adjacent grid point and the center values ​​of the next adjacent grid point, and forming a local phase slope specifically includes: For each phase grid point, collect the sampled values ​​of the back-turn grid points at the current phase grid point for all valid periods, calculate the arithmetic mean of the collected back-turn grid point sampled values, and use the arithmetic mean as the initial center value of the current phase grid point; For each phase grid point, obtain the center value of the next adjacent grid point. The center value of the next adjacent grid point of non-last phase grid points is taken as the initial center value of the next adjacent grid point of the current phase grid point, and the center value of the next adjacent grid point of the last phase grid point is taken as the initial center value of the zero phase grid point. For each phase grid point, obtain the center value of the previous adjacent grid point. The center value of the previous adjacent grid point of the zero phase grid point is taken as the initial center value of the last phase grid point, and the center value of the previous adjacent grid point of the non-zero phase grid point is taken as the initial center value of the previous adjacent grid point of the current phase grid point. For each phase grid point, the center value difference is obtained by subtracting the center value of the previous adjacent grid point from the center value of the next adjacent grid point of the current phase grid point. The phase interval scale is obtained by dividing four times pi by the number of phase grid points per cycle. The local phase slope of the current phase grid point is obtained by dividing the center value difference by the phase interval scale.

7. The method for quantization noise reduction of a high-frequency sampling sensor according to claim 6, characterized in that, The process of establishing the lower and upper boundaries of back projection based on the sampled values ​​of the return grid points, the quantization step size, the local phase slope, and the annular phase deviation, and converging to form the lower boundary of the interval convergence, the upper boundary of the interval convergence, and the boundary median of the phase grid points, specifically includes: For each effective period and each phase grid point, the sampled value of the foldback grid point at the current phase grid point in the current effective period is reduced by half a quantization step to obtain the first intermediate boundary value. The local phase slope of the current phase grid point is multiplied by the annular phase deviation of the current effective period at the current phase grid point to obtain the back projection correction value. The back projection correction value is subtracted from the first intermediate boundary value to obtain the back projection lower boundary of the current effective period at the current phase grid point. For each effective period and each phase grid point, add half a quantization step size to the sampled value of the foldback grid point at the current phase grid point in the current effective period to obtain the second intermediate boundary value. Multiply the local phase slope of the current phase grid point by the annular phase deviation at the current phase grid point in the current effective period to obtain the back projection correction value. Subtract the back projection correction value from the second intermediate boundary value to obtain the back projection upper boundary of the current effective period at the current phase grid point. For each phase grid point, collect the back projection lower boundaries of all valid periods at the current phase grid point, and take the back projection lower boundary with the highest value as the interval convergence lower boundary of the current phase grid point; For each phase grid point, collect the back projection upper boundaries of all valid periods at the current phase grid point, and take the back projection upper boundary with the lowest value as the interval convergence upper boundary of the current phase grid point. For each phase grid point, add the lower boundary of the interval convergence of the current phase grid point to the upper boundary of the interval convergence of the current phase grid point, and then divide by two to obtain the boundary median of the current phase grid point.

8. The method for quantization noise reduction of a high-frequency sampling sensor according to claim 7, characterized in that, The process domain involved in phase folding and temporal reconstruction is constructed. Within the processing domain, the nearest phase grid point is selected based on the sampling sequence number. The pre-clipping reconstruction value is obtained by combining the boundary median, local phase slope, and ring phase difference processing results of the nearest phase grid point. For sampling sequences outside the processing domain, the corresponding discrete sampling values ​​are read as pre-clipping reconstruction values, forming a pre-clipping reconstruction sequence. Specifically, this includes: A processing domain for phase folding and timing reconstruction is constructed, and the sampling numbers from the start sampling number to the end sampling number that fall within any valid period are included in the processing domain for phase folding and timing reconstruction. For each sampling number in the processing domain involved in phase folding and timing reconstruction, the phase difference between the normalized phase of the sampling point corresponding to the current sampling number and the standard phase of each phase grid point is obtained. The ring phase difference processing result corresponding to each phase grid point is obtained according to the ring phase difference processing rule. The ring phase difference processing results are compared according to their numerical values. The phase grid point with the smallest ring phase difference processing result is selected as the nearest phase grid point. When there are two identical minimum ring phase difference processing result values, the phase grid point with the earlier phase grid point number is selected as the nearest phase grid point. For each sampling number in the processing domain involved in phase folding and temporal reconstruction, the boundary median and local phase slope of the nearest phase grid point are read. The phase difference between the normalized phase of the sampling point corresponding to the current sampling number and the standard phase of the nearest phase grid point is obtained. The ring phase difference processing result of the sampling point corresponding to the current sampling number relative to the nearest phase grid point is obtained by processing according to the ring phase difference processing rule. The ring phase difference processing result of the sampling point corresponding to the current sampling number relative to the nearest phase grid point is multiplied by the local phase slope of the nearest phase grid point and then added to the boundary median of the nearest phase grid point to obtain the pre-clipping reconstructed value of the sampling point corresponding to the current sampling number. For each sampling number outside the processing domain involved in phase folding and timing reconstruction, the discrete sampling value corresponding to the current sampling number in the original sampling sequence is directly read, and the discrete sampling value corresponding to the current sampling number in the original sampling sequence is used as the pre-clipping reconstruction value of the sampling point corresponding to the current sampling number. All pre-cropping reconstructed values ​​are arranged in chronological order of sampling number to form a pre-cropping reconstructed sequence.

9. The method for quantization noise reduction of a high-frequency sampling sensor according to claim 8, characterized in that, The process of pruning the reconstructed values ​​before pruning based on the original quantization feasible interval sequence and outputting the quantization noise-smoothed sequence in the order of sampling number specifically includes: Establish clipping rules: when the input value is below the lower clipping bound, output the lower clipping bound; when the input value is within the closed interval between the lower and upper clipping bounds, output the input value; when the input value is above the upper clipping bound, output the upper clipping bound. For each sampling number, subtract half a quantization step from the discrete sample value corresponding to the current sampling number to obtain the lower boundary of the original quantization feasible interval corresponding to the current sampling number. Use the lower boundary of the original quantization feasible interval corresponding to the current sampling number as the lower bound of the clipping. Add half a quantization step to the discrete sample value corresponding to the current sampling number to obtain the upper boundary of the original quantization feasible interval corresponding to the current sampling number. Use the upper boundary of the original quantization feasible interval corresponding to the current sampling number as the upper bound of the clipping. Clip the reconstructed value before clipping corresponding to the current sampling number according to the clipping processing rules to obtain the output value after quantization noise reduction corresponding to the current sampling number. Arrange all the quantized noise-smoothed output values ​​in the order of sampling number, and output the quantized noise-smoothed sequence.

Images