A multi-layer film flexible piezoelectric sensor signal adaptive noise reduction method

Abstract

The application discloses a multilayer film flexible piezoelectric sensor signal adaptive noise reduction method, and belongs to the technical field of piezoelectric sensor signal filtering. The application firstly synchronously samples multilayer output signals of a multilayer film flexible piezoelectric sensor; then, adjacent layer output signals are matched according to matching coefficients, and difference signals are obtained; a noise proportion coefficient is calculated based on short-time energy and baseline energy of the difference signals, and the difference signals are corrected in combination with dominant factors to obtain an adaptive noise estimation coefficient; next, an estimated noise component obtained by the adaptive noise estimation coefficient and the difference signals is subtracted from the adjacent layer fusion signals to obtain initial noise reduction signals of the adjacent layers; a trend consistency coefficient is calculated, a filtering strength is determined, and adaptive filtering is performed to obtain final noise reduction signals of the adjacent layers; and finally, the final noise reduction signals of the adjacent layers are weighted to obtain overall noise reduction output signals of the sensor. The application effectively improves filtering precision and realizes efficient noise reduction.

Description

Technical Field

[0001] This invention relates to the field of piezoelectric sensor signal filtering technology, specifically to an adaptive noise reduction method for multilayer flexible piezoelectric sensor signals. Background Technology

[0002] Multilayer flexible piezoelectric sensors, with their advantages of being lightweight, flexible, and easy to conformally attach, have broad application prospects in wearable health monitoring, intelligent robot tactile sensing, and structural health monitoring. These sensors are typically composed of multiple stacked piezoelectric material layers, each independently outputting an electrical signal. By jointly analyzing the signals from multiple layers, higher sensitivity and richer information acquisition can be achieved.

[0003] In practical applications, multilayer flexible piezoelectric sensors are susceptible to interference from various noise sources, including environmental electromagnetic interference, mechanical vibration coupling noise, baseline fluctuations caused by temperature drift, and triboelectric noise generated by the deformation of the flexible substrate. Existing common noise reduction methods employ low-pass or band-pass filters with fixed cutoff frequencies. Based on prior knowledge of the energy frequency band distribution of the piezoelectric signal, the multilayer output signals are filtered separately before fusion. This method assumes that the target signal and noise have separable distribution ranges in the frequency domain; therefore, out-of-band noise can be suppressed by using a preset filter passband.

[0004] However, due to significant uncertainties in the curvature of the contact surface, stress state, and environmental conditions of flexible piezoelectric sensors in practical applications, the spectral characteristics and intensity of noise often change in real time. The aforementioned noise reduction methods based on fixed filtering parameters are ill-suited to dynamically changing noise environments. Especially when noise overlaps with the target signal in the frequency domain, relying solely on a preset filter passband is insufficient to effectively distinguish between signal and noise, resulting in a high level of noise remaining in the denoised signal or excessive filtering of effective signal components, affecting the fidelity of the sensor output. Therefore, existing technologies suffer from low filtering accuracy. Summary of the Invention

[0005] To address the aforementioned shortcomings in the prior art, this invention provides an adaptive noise reduction method for multilayer flexible piezoelectric sensors, which solves the problem of low filtering accuracy in the prior art.

[0006] To achieve the above-mentioned objectives, the technical solution adopted by this invention is as follows: an adaptive noise reduction method for multilayer film flexible piezoelectric sensors, comprising the following steps: S1. Synchronously sample the multilayer outputs of the multilayer flexible piezoelectric sensor to obtain multiple output signals; S2. Match the output signals of adjacent layers according to the matching coefficient and obtain the differential signal; S3. Based on the short-time energy and baseline energy of the differential signal, calculate the noise proportion coefficient, and combine it with the dominant factor of the differential signal to correct the noise proportion coefficient, thereby obtaining the adaptive noise estimation coefficient. S4. Subtract the estimated noise component obtained by the adaptive noise estimation coefficient and the difference signal from the fused signal of the adjacent layers to obtain the initial denoising signal of the adjacent layers; S5. Based on the amplitude trend of the initial noise reduction signals of all adjacent layers, calculate the trend consistency coefficient and determine the filtering intensity, perform adaptive filtering, and obtain the final noise reduction signals of adjacent layers. S6. Weight the signals according to the amplitude ratio of the final noise reduction signals of adjacent layers to obtain the overall noise reduction output signal of the multilayer flexible piezoelectric sensor.

[0007] Furthermore, S2 includes the following sub-steps: S21, the first The amplitude of each time step in the historical neighborhood window of the layer's output signal is squared and accumulated to obtain the contrast energy in the denominator. It is a positive integer; S22, the first The historical neighborhood window at each time step in the output signal of the layer is compared with the first... The historical neighborhood windows of each time step in the output signal of the layer are multiplied by the amplitude at the corresponding time step, and the multiplication results are accumulated to obtain the molecular contrast energy; S23. The ratio of the numerator contrast energy to the denominator contrast energy is used as the matching coefficient at the corresponding time. S24, Using the matching coefficient and the first Multiply the amplitudes of the output signals of the layer at corresponding times to obtain the first... Layer matching signals; S25, adopting the... The output signal of the layer minus the first The matching signal of the layer is used to obtain the differential signal after dynamic matching.

[0008] Furthermore, S3 includes the following sub-steps: S31. Obtain short-time energy for each differential signal; S32. Update the baseline energy of short-time energy using a recursive method; S33. Calculate the noise proportion factor based on the short-time energy and baseline energy; S34. Calculate the dominant factor of the differential signal based on the differential signal; S35. The noise proportion coefficient is corrected by using the differential signal dominance factor to obtain the adaptive noise estimation coefficient.

[0009] Furthermore, the formula for calculating the noise proportion factor in S33 is as follows: , in, For the first The differential signal of the layer Noise ratio at any given time For the first The differential signal of the layer Baseline energy at time, For the first The differential signal of the layer Short-term energy at any given moment For the denominator parameter, The time number, To obtain the maximum value, It is a positive integer.

[0010] Furthermore, the process of calculating the dominant factor of the differential signal in S34 includes: for the first... The window average energy is calculated at each time step in the differential signal of the layer to obtain the differential energy, and then the differential energy is calculated for the layer 1. Layer and first The average energy of the output signal of each layer is calculated at each time step, resulting in the energy of two channels. The two channel energies are then added together to obtain the total energy of the two layers. The ratio of the differential energy to the total energy of the two layers is used as the dominant factor of the differential signal. It is a positive integer; The process of obtaining the adaptive noise estimation coefficients in S35 includes: subtracting the product of the scaling factor and the corresponding differential signal dominance factor from 1 to obtain the correction factor; and multiplying the noise scaling factor by the correction factor to obtain the adaptive noise estimation coefficients.

[0011] Furthermore, S4 includes the following sub-steps: S41, the first The matching signal of the layer and the first The output signals of the layers are added together and averaged to obtain the fused signal of adjacent layers; S42. The estimated noise components are obtained by multiplying the adaptive noise estimation coefficients with the differential signal. S43. Subtract the estimated noise component from the fused signal of the adjacent layer to obtain the initial denoised signal of the adjacent layer.

[0012] Furthermore, S5 includes the following sub-steps: S51. Starting from each moment in the initial denoising signal of each adjacent layer, obtain the average amplitude of the historical neighborhood window; S52. Based on the amplitude and corresponding average amplitude of the initial noise reduction signal of each adjacent layer at each time step, obtain the trend direction at each time step. S53. Count the number of rising, falling and unchanged trends at the same moment in the initial noise reduction signals of all adjacent layers, and calculate the trend consistency coefficient. S54. Obtain the filter strength based on the trend consistency coefficient; S55. Filter the initial noise reduction signals of adjacent layers according to the filtering intensity to obtain the final noise reduction signals of adjacent layers.

[0013] Furthermore, the S52 process includes: in At that time, assign +1 to the trend direction. At that time, assign -1 to the trend direction. At that time, assign 0 to the trend direction, and perform statistics. In the initial denoising signals of adjacent layers, the number of +1 occurrences at the same time yields the number of occurrences in the upward trend direction; the number of -1 occurrences at the same time yields the number of occurrences in the downward trend direction; and the number of 0 occurrences at the same time yields the number of occurrences in the unchanged trend direction. For the first In the initial noise reduction signal of the adjacent layers, the first Amplitude at time, For the first In the initial noise reduction signal of the adjacent layers, the first The average amplitude of the historical neighborhood window at any given time. It is a positive integer. The time number, Number of layers; The process of calculating the trend consistency coefficient in S53 includes: at any given moment, selecting the maximum value from the number of upward trend directions, the number of downward trend directions, and the number of unchanged trend directions, and then comparing the maximum value with... The ratio is used as the percentage of the maximum direction. When the percentage of the maximum direction is greater than or equal to 0.5, the percentage of the maximum direction is used as the trend consistency coefficient. When the percentage of the maximum direction is less than 0.5, the trend consistency coefficient is set to 0.

[0014] Furthermore, the formula for calculating the filter strength in S54 is: , in, For the first Filter strength at time 10:00 To achieve the minimum filter strength, To achieve the maximum filter strength, For the first The trend consistency coefficient at any given time. Used as time number.

[0015] Furthermore, S6 includes the following sub-steps: S61. The sum of the absolute values ​​of the instantaneous amplitudes of the final noise-reduced signals of all adjacent layers is taken as the total amplitude. S62. Use the ratio of the absolute value of the instantaneous amplitude of the final denoised signal of each adjacent layer to the total amplitude as a weighting coefficient. S63. Multiply the weighting coefficients by the amplitude of the final noise reduction signal of the adjacent layers and sum them to obtain the overall noise reduction output signal of the multilayer flexible piezoelectric sensor.

[0016] The beneficial effects of this invention are as follows: 1. This invention introduces matching coefficients to match signals from adjacent layers and obtain differential signals, effectively extracting noise difference information between layers. Furthermore, by combining the short-time energy, baseline energy, and dominant factor of the differential signal, adaptive noise estimation coefficients are calculated and corrected in real time, enabling dynamic adjustment of noise estimation to follow changes in the actual noise environment. Compared to existing methods with fixed filtering parameters, this invention does not require a preset passband and can still achieve accurate noise component estimation and suppression even when noise and signal frequency domains overlap or noise characteristics of different layers are inconsistent. This avoids problems such as noise residue or excessive filtering of effective signals, significantly improving noise reduction accuracy.

[0017] 2. After obtaining the initial denoising signals from adjacent layers, this invention further calculates the trend consistency coefficient based on the amplitude trends of all adjacent layer signals, and determines the filtering strength accordingly for adaptive filtering. This mechanism can automatically adjust the smoothing degree based on the local consistency characteristics of the signal itself, applying stronger filtering in areas with strong noise to suppress residual interference, and maintaining weaker filtering in areas with real signals to preserve the original waveform characteristics, thereby effectively reducing signal distortion during the denoising process and ensuring the fidelity of the sensor output waveform.

[0018] 3. Finally, the present invention performs weighted fusion based on the amplitude ratio of the final denoised signals of adjacent layers to obtain the overall denoised output signal. This results in layers with greater signal contribution having higher weight in the fusion result, while layers with greater signal attenuation have a correspondingly lower contribution. This further optimizes the output signal quality at the global level and improves the output reliability and availability of the sensor in complex environments with multiple noises and interferences. Attached Figure Description

[0019] Figure 1 A flowchart of an adaptive noise reduction method for a multilayer flexible piezoelectric sensor signal; Figure 2 This is a schematic diagram of the structure of a multilayer flexible piezoelectric sensor. Detailed Implementation

[0020] The specific embodiments of the present invention are described below to enable those skilled in the art to understand the present invention. However, it should be understood that the present invention is not limited to the scope of the specific embodiments. For those skilled in the art, various changes are obvious as long as they are within the spirit and scope of the present invention as defined and determined by the appended claims. All inventions utilizing the concept of the present invention are protected.

[0021] like Figure 1 As shown, an adaptive noise reduction method for signals from a multilayer flexible piezoelectric sensor includes the following steps: S1. Synchronously sample the multilayer outputs of the multilayer flexible piezoelectric sensor to obtain multiple output signals; S2. Match the output signals of adjacent layers according to the matching coefficient and obtain the differential signal; S3. Based on the short-time energy and baseline energy of the differential signal, calculate the noise proportion coefficient, and combine it with the dominant factor of the differential signal to correct the noise proportion coefficient, thereby obtaining the adaptive noise estimation coefficient. S4. Subtract the estimated noise component obtained by the adaptive noise estimation coefficient and the difference signal from the fused signal of the adjacent layers to obtain the initial denoising signal of the adjacent layers; S5. Based on the amplitude trend of the initial noise reduction signals of all adjacent layers, calculate the trend consistency coefficient and determine the filtering intensity, perform adaptive filtering, and obtain the final noise reduction signals of adjacent layers. S6. Weight the signals according to the amplitude ratio of the final noise reduction signals of adjacent layers to obtain the overall noise reduction output signal of the multilayer flexible piezoelectric sensor.

[0022] like Figure 2 As shown, the multilayer flexible piezoelectric sensor includes: multiple piezoelectric sensor units and multiple isolation layers. The isolation layers are used to separate the piezoelectric sensor units, so that each piezoelectric sensor unit outputs a piezoelectric signal. The piezoelectric sensor units are vertically distributed.

[0023] In this embodiment, S2 includes the following sub-steps: S21, the first The amplitude of each time step in the historical neighborhood window of the layer's output signal is squared and accumulated to obtain the contrast energy in the denominator. It is a positive integer; S22, the first The historical neighborhood window at each time step in the output signal of the layer is compared with the first... The historical neighborhood windows of each time step in the output signal of the layer are multiplied by the amplitude at the corresponding time step, and the multiplication results are accumulated to obtain the molecular contrast energy; S23. The ratio of the numerator contrast energy to the denominator contrast energy is used as the matching coefficient at the corresponding time. S24, Using the matching coefficient and the first Multiply the amplitudes of the output signals of the layer at corresponding times to obtain the first... Layer matching signals; S25, adopting the... The output signal of the layer minus the first The matching signal of the layer is used to obtain the differential signal after dynamic matching.

[0024] In this embodiment, the formula for calculating the matching coefficient in S23 is: , in, For the first The output signal of the layer Matching coefficient at time step For the first The output signal of the layer Amplitude at time, For the first The output signal of the layer Amplitude at time, For the denominator parameter, The length of the historical neighborhood window. and This is the time number.

[0025] The expression for the differential signal obtained after dynamic matching in S25 is: , in, For the first The differential signal of the layer Amplitude at time, For the first The output signal of the layer Amplitude at time, For the first The output signal of the layer The amplitude at any given moment.

[0026] This invention achieves real-time adaptive matching of interlayer sensitivity through sliding window energy comparison, which can automatically compensate for sensitivity differences in flexible sensors caused by bending, aging or manufacturing errors during use without prior calibration; the differential signal constructed on this basis can effectively cancel out the components of the same mechanical signal while retaining the independent noise components of each layer.

[0027] In this embodiment, S3 includes the following sub-steps: S31. Obtain short-time energy for each differential signal; S32. Update the baseline energy of short-time energy using a recursive method; S33. Calculate the noise proportion factor based on the short-time energy and baseline energy; S34. Calculate the dominant factor of the differential signal based on the differential signal; S35. The noise proportion coefficient is corrected by using the differential signal dominance factor to obtain the adaptive noise estimation coefficient.

[0028] In this embodiment, the formula for calculating short-time energy in S31 is: , in, For the first The differential signal of the layer Short-term energy at any given moment For the first The differential signal of the layer Amplitude at time, This represents the length of the historical neighborhood window. In this embodiment, when the sampling rate is 1000Hz, L is set to 100, corresponding to a time length of 0.1 seconds.

[0029] The formula for calculating the baseline energy in S32 is: , in, For the first The differential signal of the layer Baseline energy at time, For the first The differential signal of the layer Short-term energy at any given moment For recursive weights, For the first The differential signal of the layer Baseline energy at time, .

[0030] The formula for calculating the noise proportion factor in S33 is: , in, For the first The differential signal of the layer Noise ratio at any given time For the first The differential signal of the layer Baseline energy at time, For the first The differential signal of the layer Short-term energy at any given moment For the denominator parameter, The time number, To obtain the maximum value, It is a positive integer.

[0031] This invention uses a sliding window to calculate short-time energy. It can sensitively capture instantaneous fluctuations in signals; then, the baseline energy is obtained through exponential recursive smoothing. A long-term memory of stable background noise was established; a noise proportion coefficient was constructed. When instantaneous energy Significantly higher than baseline When the time coefficient approaches 1, it indicates that the current differential signal is dominated by noise components. When the coefficient approaches or falls below the baseline, it tends to be 0, indicating that the noise level is low or the system is in a silent state.

[0032] In this embodiment, the process of calculating the dominant factor of the differential signal in S34 includes: for the first The window average energy is calculated at each time step in the differential signal of the layer to obtain the differential energy, and then the differential energy is calculated for the layer 1. Layer and first The average energy of the output signal of each layer is calculated at each time step, resulting in the energy of two channels. The two channel energies are then added together to obtain the total energy of the two layers. The ratio of the differential energy to the total energy of the two layers is used as the dominant factor of the differential signal. It is a positive integer; The formula for calculating the dominance factor of the differential signal is: ,in, For the first The differential signal of the layer The dominant factor of the difference signal at time step, For the first The differential signal of the layer Amplitude at time, For the first The output signal of the layer Amplitude at time, For the first The output signal of the layer Amplitude at time, For the denominator parameter, The time number, For the layer number, The expectation is the mean of the squared values ​​within the historical neighborhood window.

[0033] This invention utilizes differential signal dominance factors This characterizes the proportion of the residual real mechanical signal in the differential signal. When the matching coefficient... When accurate, the differential signal is dominated by noise. Smaller; when matching deviation causes signal residue, Significantly increased.

[0034] The process of obtaining the adaptive noise estimation coefficients in S35 includes: subtracting the product of the scaling factor and the corresponding differential signal dominance factor from 1 to obtain the correction factor, and multiplying the noise proportion factor and the correction factor to obtain the adaptive noise estimation coefficients. The formula for calculating the adaptive noise estimation coefficients is: ,in, For the first The differential signal of the layer Adaptive noise estimation coefficients at time step For the first The differential signal of the layer The dominant factor of the difference signal at time step, This is the scaling factor. When there is a significant amount of residual real signal in the differential signal ( When the value is large, automatically reduce To avoid mistaking effective mechanical components for noise cancellation; when the differential signal is pure ( When using this method, the original noise percentage coefficient remains unchanged, thus fully preserving the noise reduction capability.

[0035] In this embodiment, the scaling factor The value range is [0.1, 1], and this proportionality coefficient is used to control the dominant factor of the differential signal. noise proportion factor The correction strength.

[0036] In this embodiment, S4 includes the following sub-steps: S41, the first The matching signal of the layer and the first The output signals of the layers are added together and averaged to obtain the fused signal of adjacent layers: ,in, For the first In the fused signal of adjacent layers, the th Amplitude at time, For the first The output signal of the layer Amplitude at time, For the first The output signal of the layer Amplitude at time, For the first The output signal of the layer Matching coefficient at time step For the first In the matching signal of the layer, the first Amplitude at any given moment; S42. The estimated noise components are obtained by multiplying the adaptive noise estimation coefficients with the differential signal. S43. Subtract the estimated noise component from the fused signal of the adjacent layer to obtain the initial denoised signal of the adjacent layer.

[0037] The expression for obtaining the initial denoised signal of the adjacent layer is: , in, For the first In the initial noise reduction signal of the adjacent layers, the first Amplitude at time, For the first In the fused signal of adjacent layers, the th Amplitude at time, For the first The differential signal of the layer Adaptive noise estimation coefficients at time step For the first The differential signal of the layer The amplitude at any given moment.

[0038] This invention will The matching signal of the layer and the first The output signal of the layer is used to obtain the fused signal of the adjacent layers, and then adaptive noise estimation coefficients are applied. For differential signals Weighting is performed to obtain the estimated noise components. ,because Taking into account the proportion of noise energy ( Larger signals will be reduced more significantly, and the degree of signal residue ( (Reduce noise as needed if it is large), enabling adaptive matching of the estimated noise. The actual noise level in the signal is determined. The estimated noise component is subtracted from the fused signal to obtain the initial denoised signal. .because Residual factors of the signal Control – The more residual real signal in the differential signal, the better. The smaller the value, the better the matching coefficient. When temporary misalignment occurs due to bending or transient impact, the noise reduction intensity is automatically reduced, preserving some noise rather than damaging the effective mechanical signal. For an N-layer stacked array, S4 generates N-1 initial noise reduction signals from adjacent layers in parallel.

[0039] In this embodiment, S5 includes the following sub-steps: S51. Starting from each moment in the initial denoising signal of each adjacent layer, obtain the average amplitude of the historical neighborhood window; S52. Based on the amplitude and corresponding average amplitude of the initial noise reduction signal of each adjacent layer at each time step, obtain the trend direction at each time step. S53. Count the number of rising, falling and unchanged trends at the same moment in the initial noise reduction signals of all adjacent layers, and calculate the trend consistency coefficient. S54. Obtain the filter strength based on the trend consistency coefficient; S55. Filter the initial noise reduction signals of adjacent layers according to the filtering intensity to obtain the final noise reduction signals of adjacent layers.

[0040] In this embodiment, the process of S52 includes: in At that time, assign +1 to the trend direction. At that time, assign -1 to the trend direction. At that time, assign 0 to the trend direction, and perform statistics. In the initial denoising signals of adjacent layers, the number of +1 occurrences at the same time yields the number of occurrences in the upward trend direction; the number of -1 occurrences at the same time yields the number of occurrences in the downward trend direction; and the number of 0 occurrences at the same time yields the number of occurrences in the unchanged trend direction. For the first In the initial noise reduction signal of the adjacent layers, the first Amplitude at time, For the first In the initial noise reduction signal of the adjacent layers, the first The average amplitude of the historical neighborhood window at any given time. It is a positive integer. The time number, Number of layers; The process of calculating the trend consistency coefficient in S53 includes: at any given moment, selecting the maximum value from the number of upward trend directions, the number of downward trend directions, and the number of unchanged trend directions, and then comparing the maximum value with... The ratio is used as the percentage of the maximum direction. When the percentage of the maximum direction is greater than or equal to 0.5, the percentage of the maximum direction is used as the trend consistency coefficient. When the percentage of the maximum direction is less than 0.5, the trend consistency coefficient is set to 0.

[0041] In this embodiment, the formula for calculating the maximum directional proportion is: , in, For the first The maximum directional percentage at any given time For the first The number of upward trend directions at any given moment. For the first The number of downward trend directions at any given time. For the first The number of constant trend directions at any given time. For the number of layers, Used as time number.

[0042] In this embodiment, the formula for calculating the filter strength in S54 is: , in, For the first Filter strength at time 10:00 To achieve the minimum filter strength, To achieve the maximum filter strength, For the first The trend consistency coefficient at any given time. Used as time number.

[0043] In this embodiment, Take 0.1, The filter strength is set to 0.9, and then continuously and linearly adjusted between 0.1 and 0.9.

[0044] The expression for filtering in S55 is: , in, For the first In the final noise reduction signal of the adjacent layers, the first Amplitude at time, For the first In the initial noise reduction signal of the adjacent layers, the first Amplitude at time, For the first Filter strength at time 10:00 For the first In the final noise reduction signal of the adjacent layers, the first The amplitude at any given moment.

[0045] This invention statistically analyzes the trend direction distribution of all N-1 channels at the same time. When a real mechanical signal occurs, all channels exhibit a consistent trend, with the largest proportion in the direction of maximum change and a high trend consistency coefficient. Approaching 1; when the signal is dominated by noise, the trend direction of each channel is randomly distributed, with the largest direction accounting for less than 0.5. It was set to 0. According to... Dynamically calculate filter strength ,when Use maximum filter strength To achieve the weakest filtering, quickly track signal changes and preserve effective amplitude, when Use minimum filter strength This invention achieves the strongest filtering to suppress random noise. It employs an independent filtering method for each channel, with the filtering strength related to the trend consistency coefficient. When the trend is consistent, the current valid signal is retained; when the trend is inconsistent, historical signals are given priority.

[0046] In this embodiment, S6 includes the following sub-steps: S61. The sum of the absolute values ​​of the instantaneous amplitudes of the final noise-reduced signals of all adjacent layers is taken as the total amplitude. S62. Use the ratio of the absolute value of the instantaneous amplitude of the final denoised signal of each adjacent layer to the total amplitude as a weighting coefficient. S63. Multiply the weighting coefficients by the amplitude of the final noise reduction signal of the adjacent layers and sum them to obtain the overall noise reduction output signal of the multilayer flexible piezoelectric sensor.

[0047] The expression for the overall noise-reduced output signal of the multilayer flexible piezoelectric sensor is as follows: , in, The first part of the overall noise reduction output signal of the multilayer flexible piezoelectric sensor The amplitude at any given moment.

[0048] This invention uses the absolute values ​​of the instantaneous amplitudes of the final denoised signals from adjacent layers for normalized weighted fusion, which adaptively allocates the fusion weights for each layer: layers with larger signal amplitudes and more significant effective components are assigned higher weighting coefficients; layers with smaller amplitudes and higher noise proportions are assigned lower weighting coefficients. By weighting and accumulating multi-layer signals in this way, the effective piezoelectric response signal can be highlighted, residual noise and inter-layer redundant interference can be suppressed, resulting in a higher signal-to-noise ratio and clearer characteristics in the overall denoised output signal of the multi-layer flexible piezoelectric sensor.

[0049] This invention first synchronously samples multi-layer signals, using matching coefficients to match signals from adjacent layers and obtain differential signals, thereby highlighting noise components unrelated to the signal. Then, it calculates a noise proportion coefficient based on the short-time energy and baseline energy of the differential signals, and introduces a differential signal dominance factor for correction, obtaining adaptive noise estimation coefficients that allow noise estimation to follow environmental changes in real time. Next, it subtracts the estimated noise components obtained from the adaptive noise estimation coefficients and differential signals from the fused signals of adjacent layers to obtain the initial denoised signal. Then, it determines the filtering strength by calculating a trend consistency coefficient, performs adaptive filtering, further suppressing residual noise and preserving effective signal details. Finally, it weights and fuses the signals according to the amplitude proportions of each adjacent layer to output the overall denoised signal. The entire process does not require a preset fixed filter passband and can dynamically adjust the denoising strategy according to the actual time-varying characteristics of the signal and noise. Even when noise and target signal frequency domains overlap and noise intensity varies across layers, it can effectively separate signal and noise, thus significantly improving filtering accuracy and signal fidelity.

[0050] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. An adaptive noise reduction method for signals from a multilayer flexible piezoelectric sensor, characterized in that, Includes the following steps: S1. Synchronously sample the multilayer outputs of the multilayer flexible piezoelectric sensor to obtain multiple output signals; S2. Match the output signals of adjacent layers according to the matching coefficient and obtain the differential signal; S3. Based on the short-time energy and baseline energy of the differential signal, calculate the noise proportion coefficient, and combine it with the dominant factor of the differential signal to correct the noise proportion coefficient, thereby obtaining the adaptive noise estimation coefficient. S4. Subtract the estimated noise component obtained by the adaptive noise estimation coefficient and the difference signal from the fused signal of the adjacent layers to obtain the initial denoising signal of the adjacent layers; S5. Based on the amplitude trend of the initial noise reduction signals of all adjacent layers, calculate the trend consistency coefficient and determine the filtering intensity, perform adaptive filtering, and obtain the final noise reduction signals of adjacent layers. S6. Weight the signals according to the amplitude ratio of the final noise reduction signals of adjacent layers to obtain the overall noise reduction output signal of the multilayer flexible piezoelectric sensor.

2. The adaptive noise reduction method for multilayer flexible piezoelectric sensor signals according to claim 1, characterized in that, S2 includes the following steps: S21, the first The amplitude of each time step in the historical neighborhood window of the layer's output signal is squared and accumulated to obtain the contrast energy in the denominator. It is a positive integer; S22, the first The historical neighborhood window at each time step in the output signal of the layer is compared with the first... The historical neighborhood windows of each time step in the output signal of the layer are multiplied by the amplitude at the corresponding time step, and the multiplication results are accumulated to obtain the molecular contrast energy; S23. The ratio of the numerator contrast energy to the denominator contrast energy is used as the matching coefficient at the corresponding time. S24, Using the matching coefficient and the first Multiply the amplitudes of the output signals of the layer at corresponding times to obtain the first... Layer matching signals; S25, adopting the... The output signal of the layer minus the first The matching signal of the layer is used to obtain the differential signal after dynamic matching.

3. The adaptive noise reduction method for multilayer flexible piezoelectric sensor signals according to claim 1, characterized in that, S3 includes the following steps: S31. Obtain short-time energy for each differential signal; S32. Update the baseline energy of short-time energy using a recursive method; S33. Calculate the noise proportion factor based on the short-time energy and baseline energy; S34. Calculate the dominant factor of the differential signal based on the differential signal; S35. The noise proportion coefficient is corrected by using the differential signal dominance factor to obtain the adaptive noise estimation coefficient.

4. The adaptive noise reduction method for multilayer flexible piezoelectric sensor signals according to claim 3, characterized in that, The formula for calculating the noise proportion coefficient in S33 is as follows: ， in, For the first The differential signal of the layer Noise ratio at any given time For the first The differential signal of the layer Baseline energy at time, For the first The differential signal of the layer Short-term energy at any given moment For the denominator parameter, The time number, To obtain the maximum value, It is a positive integer.

5. The adaptive noise reduction method for multilayer flexible piezoelectric sensor signals according to claim 3, characterized in that, The process of calculating the dominant factor of the differential signal in S34 includes: [The process is repeated here, so the translation ends abruptly.] The window average energy is calculated at each time step in the differential signal of the layer to obtain the differential energy, and then the differential energy is calculated for the layer 1. Layer and first The average energy of the output signal of each layer is calculated at each time step, resulting in the energy of two channels. The two channel energies are then added together to obtain the total energy of the two layers. The ratio of the differential energy to the total energy of the two layers is used as the dominant factor of the differential signal. It is a positive integer; The process of obtaining the adaptive noise estimation coefficients in S35 includes: subtracting the product of the scaling factor and the corresponding differential signal dominance factor from 1 to obtain the correction factor; and multiplying the noise scaling factor by the correction factor to obtain the adaptive noise estimation coefficients.

6. The adaptive noise reduction method for multilayer flexible piezoelectric sensor signals according to claim 2, characterized in that, S4 includes the following sub-steps: S41, the first The matching signal of the layer and the first The output signals of the layers are added together and averaged to obtain the fused signal of adjacent layers; S42. The estimated noise components are obtained by multiplying the adaptive noise estimation coefficients with the differential signal. S43. Subtract the estimated noise component from the fused signal of the adjacent layer to obtain the initial denoised signal of the adjacent layer.

7. The adaptive noise reduction method for multilayer flexible piezoelectric sensor signals according to claim 1, characterized in that, S5 includes the following steps: S51. Starting from each moment in the initial denoising signal of each adjacent layer, obtain the average amplitude of the historical neighborhood window; S52. Based on the amplitude and corresponding average amplitude of the initial noise reduction signal of each adjacent layer at each time step, obtain the trend direction at each time step. S53. Count the number of rising, falling and unchanged trends at the same moment in the initial noise reduction signals of all adjacent layers, and calculate the trend consistency coefficient. S54. Obtain the filter strength based on the trend consistency coefficient; S55. Filter the initial noise reduction signals of adjacent layers according to the filtering intensity to obtain the final noise reduction signals of adjacent layers.

8. The adaptive noise reduction method for multilayer flexible piezoelectric sensor signals according to claim 7, characterized in that, The process S52 includes: in At that time, assign +1 to the trend direction. At that time, assign -1 to the trend direction. At that time, assign 0 to the trend direction, and perform statistics. In the initial denoising signals of adjacent layers, the number of +1 occurrences at the same time yields the number of occurrences in the upward trend direction; the number of -1 occurrences at the same time yields the number of occurrences in the downward trend direction; and the number of 0 occurrences at the same time yields the number of occurrences in the unchanged trend direction. For the first In the initial noise reduction signal of the adjacent layers, the first Amplitude at time, For the first In the initial noise reduction signal of the adjacent layers, the first The average amplitude of the historical neighborhood window at any given time. It is a positive integer. The time number, Number of layers; The process of calculating the trend consistency coefficient in S53 includes: at any given moment, selecting the maximum value from the number of upward trend directions, the number of downward trend directions, and the number of unchanged trend directions, and then comparing the maximum value with... The ratio is used as the percentage of the maximum direction. When the percentage of the maximum direction is greater than or equal to 0.5, the percentage of the maximum direction is used as the trend consistency coefficient. When the percentage of the maximum direction is less than 0.5, the trend consistency coefficient is set to 0.

9. The adaptive noise reduction method for multilayer flexible piezoelectric sensor signals according to claim 7, characterized in that, The formula for calculating the filter strength in S54 is as follows: ， in, For the first Filter strength at time 10:00 To achieve the minimum filter strength, To achieve the maximum filter strength, For the first The trend consistency coefficient at any given time. Used as time number.

10. The adaptive noise reduction method for multilayer flexible piezoelectric sensor signals according to claim 1, characterized in that, S6 includes the following sub-steps: S61. The sum of the absolute values ​​of the instantaneous amplitudes of the final noise-reduced signals of all adjacent layers is taken as the total amplitude. S62. Use the ratio of the absolute value of the instantaneous amplitude of the final denoised signal of each adjacent layer to the total amplitude as a weighting coefficient. S63. Multiply the weighting coefficients by the amplitude of the final noise reduction signal of the adjacent layers and sum them to obtain the overall noise reduction output signal of the multilayer flexible piezoelectric sensor.

Images