An adaptive filtering and noise reduction algorithm and system for biological physiological signals

Abstract

This invention discloses an adaptive filtering and denoising algorithm and system for biophysiological signals, belonging to the technical fields of physiological signal processing, embedded signal denoising, and adaptive filtering algorithms. Addressing the problems of power frequency interference, baseline drift, motion artifacts, and white noise contamination during the acquisition of biosignals such as ECG, EEG, and EMG, this invention proposes a combined filtering scheme based on multi-level adaptive thresholds. Based on sliding window adaptive mean filtering, this invention combines an adaptive threshold wavelet decomposition and reconstruction algorithm to achieve graded suppression of different noise types. Simultaneously, a signal quality assessment module is introduced to dynamically adjust the filtering order and threshold parameters according to the real-time signal-to-noise ratio, avoiding over-filtering that leads to the loss of effective physiological features. This invention solves the industry pain points of traditional physiological signal filtering algorithms, such as poor adaptability to fixed parameters, denoising failure in strong noise environments, and easy destruction of effective signal waveforms. As a core functional enhancement module in the second tier of the patent matrix, this invention can seamlessly connect to the first tier sensing and acquisition layer, improving the quality of physiological data acquisition and adapting to portable health and wellness terminals, medical monitoring equipment, and brain-computer interface terminals for practical application.

Description

Technical Field

[0001] This invention relates to the fields of physiological signal processing, embedded signal denoising, adaptive filtering algorithms, and bioelectrical signal preprocessing, specifically to an adaptive filtering and denoising algorithm and system for biophysiological signals. This invention belongs to the front-end signal preprocessing and enhancement module of a multi-level unit dynamic linkage and collaborative control system, specifically adapted to biosignal acquisition scenarios such as ECG, EEG, and EMG. It can be used in portable health and wellness terminals, medical monitoring equipment, brain-computer interface terminals, and distributed sensing nodes, belonging to the interdisciplinary field of biosignal processing and embedded algorithms. Background Technology

[0002] Current biophysiological signal acquisition processes commonly suffer from problems such as power frequency interference, baseline drift, motion artifacts, and white noise contamination. Traditional filtering algorithms often employ fixed parameter designs, resulting in poor adaptability and ineffective denoising in high-noise environments, while also easily leading to the loss of effective physiological features and signal waveform distortion. Conventional mean filtering and wavelet filtering algorithms cannot dynamically adjust parameters based on real-time noise conditions, failing to denoise in low signal-to-noise ratio scenarios. Furthermore, they lack a closed-loop feedback mechanism for signal quality assessment, making it impossible to balance denoising effectiveness with feature integrity. Existing technologies are ill-suited for the low-computing-power, low-power operating conditions of portable embedded terminals, exhibiting high algorithm complexity and an inability to process physiological signals in real time, indicating a significant technological gap. Summary of the Invention

[0003] To address the shortcomings of existing biological signal filtering algorithms, such as poor adaptability to fixed parameters, failure to denoise in strong noise environments, easy destruction of effective signals, and inability to adapt to embedded low-computing-power environments, this invention proposes an adaptive filtering and denoising algorithm and system for biological physiological signals. This invention is based on sliding window adaptive mean filtering, combined with a multi-scale wavelet decomposition and reconstruction algorithm to achieve hierarchical suppression of different noise types; it introduces a noise type identification and adaptive threshold adjustment mechanism to dynamically adjust filtering parameters according to the real-time signal-to-noise ratio; and it forms a closed-loop feedback through signal quality assessment to balance denoising effect and feature integrity. The core innovations of this invention compared to traditional filtering algorithms are: achieving adaptive parameter adjustment to adapt to different noise intensity scenarios; employing a combined filtering scheme to target and suppress multiple noise types; adding a quality assessment closed loop to avoid over-filtering; and having low algorithm complexity, making it suitable for embedded low-computing-power environments.

[0004] Furthermore, the window length of the sliding window mean filter can be dynamically adjusted according to the signal noise intensity, balancing noise reduction effect and signal real-time performance. Furthermore, the number of multi-scale wavelet decomposition layers and threshold parameters can be adaptively adjusted according to the noise type identification results to improve the targeted noise reduction effect; Furthermore, the signal quality assessment module uses multi-dimensional indicators to comprehensively score the signal, triggering dynamic adjustments to the filter parameters to form a closed-loop control. Furthermore, the algorithm of this invention can be deployed independently on the front-end acquisition terminal, or it can be embedded as a module in a multi-level linkage control system to achieve full-link signal purification.

[0005] The beneficial effects of this invention are as follows: 1. Parameter adaptive adjustment: The filter parameters can be dynamically adjusted according to the real-time noise intensity to adapt to different acquisition scenarios; 2. Targeted suppression of multiple noises: It has a good suppression effect on power frequency interference, baseline drift, motion artifacts and white noise; 3. Feature integrity protection: Through closed-loop feedback of quality assessment, it avoids distortion of effective signals caused by over-filtering; 4. Low computing power adaptation: The algorithm has low complexity and is adapted to the low power consumption and low computing power operating conditions of embedded terminals. 5. Strong architectural compatibility: It can seamlessly connect to the sensing and acquisition layer, improving the accuracy of subsequent physiological data analysis; 6. Strong commercial applicability: It can be licensed as a standalone filtering algorithm IP, adapted to various physiological monitoring devices, and has a low threshold for industrialization. Attached Figure Description Figure 1 This is a timing diagram of the biophysiological signal adaptive filtering and noise reduction algorithm of the present invention. It adopts a vertical pure planar patent standard process architecture. The main lines are as follows: signal preprocessing, sliding window adaptive mean filtering, multi-scale wavelet decomposition, noise component filtering, signal reconstruction, quality assessment and feedback adjustment. A quality assessment judgment branch is set. If the assessment is qualified, the noise reduction signal is output. If the assessment is unqualified, the filtering parameters are automatically reset and the process is repeated. The overall process is closed-loop from beginning to end, and the logic is clear and neat. Detailed Implementation

[0006] The present invention will be further described in detail below with reference to specific embodiments.

[0007] This embodiment provides an adaptive filtering and noise reduction system for biophysiological signals, applied to a portable electrocardiogram (ECG) monitoring terminal, and adapted to the front-end signal preprocessing scenario of a multi-level unit dynamic linkage and collaborative control system. The hardware controller in this embodiment uses the STM32L4 low-power embedded chip, equipped with the adaptive filtering algorithm of this invention, to perform real-time noise reduction processing on the acquired ECG signals.

[0008] At the algorithm execution level, the ECG signal acquired at the front end first undergoes analog-to-digital conversion and DC bias removal preprocessing; then it enters the sliding window adaptive mean filtering module, which dynamically adjusts the window length according to the signal noise intensity to suppress high-frequency white noise; next, the mean-filtered signal is subjected to multi-scale wavelet decomposition, with the number of decomposition levels adjusted to 3-4 levels based on the noise type identification results; after soft thresholding to remove power frequency interference and baseline drift components, the effective components are reconstructed using wavelets to obtain the denoised ECG signal; finally, through the signal quality assessment module, the signal-to-noise ratio improvement and R-wave recognition accuracy are calculated. When the comprehensive score is lower than the preset threshold, the parameter adjustment amount is automatically calculated and the filtering order and threshold coefficient are reset to form a closed-loop control.

[0009] In this embodiment, under high-noise motion scenarios, the signal-to-noise ratio of ECG signals can be improved by more than 12dB, the R-wave recognition accuracy is improved to more than 98.5%, and the processing time of a single algorithm is no more than 15ms, making it suitable for the real-time processing requirements of embedded terminals. This invention can be widely adapted to various physiological monitoring devices, brain-computer interface terminals, and distributed sensing nodes, effectively improving the quality of physiological signal acquisition and possessing extremely high industrialization and promotion value.

Claims

1. A bio-physiological signal adaptive filtering and noise reduction system, characterized by, include: The system comprises a signal acquisition and preprocessing module, a sliding window adaptive mean filtering module, a multi-scale wavelet decomposition and reconstruction module, a noise type identification and classification module, an adaptive threshold adjustment module, and a signal quality assessment and feedback control module. The signal acquisition and preprocessing module receives the raw biophysiological signals acquired from the front end and performs analog-to-digital conversion and signal amplification preprocessing. The sliding window adaptive mean filtering module uses dynamically adjusted window lengths for mean filtering to suppress white noise and high-frequency random noise in the signal. The multi-scale wavelet decomposition and reconstruction module decomposes the signal into multiple values ​​using multi-scale wavelet decomposition. The signal is decomposed into different frequency components to separate and filter out power frequency interference and baseline drift. The noise type identification and classification module identifies four main noise types—power frequency interference, baseline drift, motion artifacts, and white noise—based on the signal's time-domain and frequency-domain characteristics. The adaptive threshold adjustment module dynamically adjusts the filtering order, wavelet decomposition level, and threshold parameters according to the noise type identification results and real-time signal-to-noise ratio. The signal quality assessment and feedback control module evaluates the signal-to-noise ratio and feature integrity of the filtered signal, forming a closed-loop feedback control to avoid over-filtering.

2. A bio-physiological signal adaptive filtering and noise reduction method applied to the filtering and noise reduction system of claim 1, characterized in that, Includes the following steps: S1. Signal preprocessing: Analog-to-digital conversion, DC bias removal and normalization preprocessing are performed on the raw biological physiological signals; S2. Sliding window adaptive mean filtering: Employs dynamic window length mean filtering to suppress high-frequency white noise in the signal; S3. Multi-scale wavelet decomposition: Perform multi-scale wavelet decomposition on the mean-filtered signal to obtain different frequency components. S4. Noise Component Identification and Filtering: Based on the noise type identification results, threshold filtering is performed on power frequency interference and baseline drift components; S5. Signal reconstruction: The processed effective components are reconstructed by wavelet to obtain the denoised physiological signal. S6. Quality Assessment and Feedback Adjustment: Evaluate the signal-to-noise ratio and feature integrity of the filtered signal, dynamically adjust the filtering parameters for the next round, and form a closed-loop control.

3. The adaptive filtering and noise reduction system for biological physiological signals according to claim 1, characterized in that, The window length of the sliding window adaptive mean filtering module can be dynamically adjusted within the range of 5 points to the signal period length. The window length is positively correlated with the real-time detected noise intensity, with high noise intensity corresponding to a long window.

4. The adaptive filtering and noise reduction system for biological physiological signals according to claim 1, characterized in that, The multi-scale wavelet decomposition and reconstruction module adopts a compactly supported orthogonal wavelet basis, preferably the Daubechies wavelet basis, with a decomposition level of 3-5, which is adapted to the frequency distribution characteristics of biological signals.

5. The adaptive filtering and noise reduction system for biological physiological signals according to claim 1, characterized in that, The noise type identification and classification module classifies and identifies noise types based on signal zero-crossing rate, power spectral density, and peak distribution characteristics.

6. The adaptive filtering and noise reduction system for biological physiological signals according to claim 1, characterized in that, The adaptive threshold adjustment module adopts a soft threshold processing method to retain the edge features of the effective signal and suppress the residual energy of the noise component.

7. The adaptive filtering and noise reduction system for biological physiological signals according to claim 1, characterized in that, This system can connect to the sensing and acquisition layer of a multi-level unit dynamic linkage and collaborative control system to achieve front-end signal purification and improve the accuracy of subsequent analysis and processing.

8. The adaptive filtering and noise reduction method for biological physiological signals according to claim 2, characterized in that, The quality assessment indicators mentioned in step S6 include signal-to-noise ratio improvement, feature point recognition accuracy, and signal waveform distortion. When the comprehensive score is lower than the preset threshold, the signal quality assessment module calculates the parameter adjustment amount and automatically resets the filter order and threshold coefficient.

Images