Adaptive multi-dimensional continuous signal quality monitoring method and system

Abstract

The application belongs to the technical field of signal processing and quality evaluation, and provides a self-adaptive multi-dimensional continuous signal quality monitoring method and system, which comprises the following steps: loading processing parameters matched with the current input continuous signal in a signal characteristic archive, combining the processing parameters to perform multi-stage cascaded signal processing on the current input continuous signal to obtain a preprocessed signal; performing multi-dimensional calculation on the preprocessed signal to obtain multi-dimensional quality indexes; loading weight coefficients corresponding to the multi-dimensional quality indexes in the signal characteristic archive and calculating a comprehensive quality score; performing independent judgment on each of the multi-dimensional quality indexes based on a preset strategy to obtain a point-by-point quality mask; and evaluating the quality of the current input signal according to the comprehensive quality score and the point-by-point quality mask. The application has good universality, strong multi-type noise processing capability, consistent evaluation standard and sufficient self-adaptive capability.

Description

Technical Field

[0001] This application relates to the field of signal processing and quality assessment technology, specifically to an adaptive multi-dimensional continuous signal quality monitoring method and system. Background Technology

[0002] In modern information acquisition systems, the quality of continuous signals (such as PPG and ECG in physiological signals, temperature / pressure signals in industrial monitoring, and CPU utilization / memory usage in equipment performance monitoring) directly affects the accuracy of subsequent analysis and decision-making. Therefore, monitoring the quality of continuous signals is very important.

[0003] However, existing continuous signal quality monitoring technologies have significant drawbacks, including: 1. High specialization and poor versatility: Traditional continuous signal quality monitoring methods are mostly designed for specific fields and cannot be reused across fields; 2. Fragmented evaluation standards: Different fields have significantly different definitions of signal quality, making it impossible to compare evaluation results across fields; 3. Insufficient ability to handle multiple types of noise: There is a lack of comprehensive processing mechanisms for composite noise such as baseline drift, impulse noise, and periodic interference; 4. Lack of online learning and dynamic optimization mechanisms: Fixed parameter algorithms are difficult to adapt to dynamic changes in signal characteristics; 5. Lack of interpretability and robustness of evaluation results: Most methods only provide a binary judgment of "qualified / unqualified," lacking fine-grained quantification and interpretable output, making it difficult to support refined decision-making.

[0004] In summary, existing continuous signal quality monitoring technologies suffer from technical problems such as poor versatility, insufficient noise processing capabilities, and a lack of online learning capabilities. Summary of the Invention

[0005] This application provides an adaptive multi-dimensional continuous signal quality monitoring method and system to solve at least one of the above-mentioned technical problems.

[0006] The first aspect of this application provides an adaptive multi-dimensional continuous signal quality monitoring method, including:

[0007] Load the processing parameters from the signal feature archive that match the current input continuous signal, and perform multi-level cascaded signal processing on the current input continuous signal in combination with the processing parameters to obtain the preprocessed signal;

[0008] Multi-dimensional calculations are performed on the preprocessed signal to obtain corresponding multi-dimensional quality indicators;

[0009] Load the weight coefficients corresponding to the quality indicators of each dimension from the signal feature archive and calculate the comprehensive quality score; independently judge each quality indicator based on the preset strategy to obtain the point-by-point quality mask;

[0010] The quality of the current input signal is evaluated based on the overall quality score and the point-by-point quality mask.

[0011] In some preferred embodiments, the processing parameters in the loaded signal feature archive that correspond to the currently input continuous signal are combined with the processing parameters to perform multi-stage cascaded signal processing on the currently input continuous signal to obtain a preprocessed signal, including:

[0012] Based on the type or sampling frequency of the currently input continuous signal, load the processing parameters that match the currently input continuous signal from the signal feature archive;

[0013] The current input continuous signal is subjected to three-stage cascaded signal processing based on the processing parameters to obtain a preprocessed signal.

[0014] In some preferred embodiments, the three-stage cascaded signal processing of the currently input continuous signal, combined with processing parameters, to obtain a preprocessed signal includes:

[0015] By combining the basic coefficient of the jump detection threshold, the current input continuous signal is corrected for jumps to obtain the jump-corrected signal;

[0016] By combining the baseline removal weighting coefficients, filter cutoff frequency, and filter order, and based on a preset morphological opening operation and low-pass filter fusion strategy, baseline drift removal is performed on the transition-corrected signal to obtain the drift-removed signal.

[0017] By combining the signal type with an adaptive threshold mapping table, dynamic noise suppression is performed on the drift-removed signal to obtain the preprocessed signal.

[0018] In some preferred embodiments, the step of combining the jump detection threshold base coefficient to perform signal jump correction on the currently input continuous signal to obtain the jump-corrected signal includes:

[0019] Calculate the absolute value of the difference sequence of the currently input continuous signal;

[0020] The jump threshold is dynamically calculated based on the absolute value of the signal differential sequence and the basic coefficient of the jump detection threshold.

[0021] When the absolute value of the signal difference sequence is greater than the jump threshold, the sampling point corresponding to the absolute value of the current signal difference sequence is identified as the jump point;

[0022] For each transition point, multiple rounds of detection are performed to obtain the median difference of the neighboring signals before and after each transition point to calculate the correction amount. Based on the correction amount, the amplitude of the signal segment after the transition point is compensated to obtain the transition-corrected signal.

[0023] In some preferred embodiments, the baseline drift removal is performed on the transition-corrected signal by combining the baseline removal weighting coefficients, the filter cutoff frequency, and the filter order, based on a preset morphological opening operation and low-pass filter fusion strategy, to obtain the drift-removed signal, including:

[0024] Morphological opening operation is used to extract the fast baseline components of the jump-corrected signal.

[0025] Based on the Butterworth low-pass filter with the preset filter cutoff frequency and filter order, the signal after jump correction is filtered to obtain the slow baseline component.

[0026] Based on the baseline removal weighting coefficient, the fast baseline component and the slow baseline component are weighted and fused to obtain the fused baseline component;

[0027] Subtract the jump-corrected signal from the fused baseline component to obtain the drift-free signal.

[0028] In some preferred embodiments, the step of combining the signal type with an adaptive threshold mapping table to perform dynamic noise suppression on the drift-removed signal to obtain a preprocessed signal includes:

[0029] Based on the signal type of the current input continuous signal, the corresponding noise scene judgment coefficient is loaded from the signal type and adaptive threshold mapping table;

[0030] Calculate the standard deviation of the initial segment of the signal after removing drift;

[0031] Calculate the peak-to-peak value of the signal after drift removal;

[0032] The noise scene judgment coefficient and the peak-to-peak value are calculated. The noise level is judged by comparing the product with the standard deviation. Different denoising processes are applied to the drift-removed signal according to the noise level to obtain the preprocessed signal.

[0033] In some preferred embodiments, the multi-dimensional calculation of the preprocessed signal to obtain multi-dimensional quality indicators specifically includes:

[0034] Multiple time-domain and depth feature dimension calculations are performed on the preprocessed signal to obtain multiple time-domain quality indicators and depth feature quality indicators.

[0035] In some preferred embodiments, the weight coefficients corresponding to the quality indicators of each dimension in the loaded signal feature archive are used to calculate the comprehensive quality score; based on a preset strategy, each dimension of quality indicators is independently judged to obtain a point-by-point quality mask.

[0036] This involves loading the weight coefficients corresponding to the quality indicators of each dimension from the signal feature archive, and calculating the comprehensive quality score, including:

[0037] Based on the signal type of the currently input continuous signal, load the weight coefficients corresponding to the quality indicators of each dimension from the signal feature archive;

[0038] The quality indicators of each dimension are weighted and integrated with their corresponding weight coefficients to obtain a comprehensive quality score.

[0039] In some preferred embodiments, the weight coefficients corresponding to the quality indicators of each dimension in the loaded signal feature archive are used to calculate the comprehensive quality score; based on a preset strategy, each dimension of quality indicators is independently judged to obtain a point-by-point quality mask.

[0040] Among them, based on a preset strategy, each dimension of quality indicators is independently judged to obtain a point-by-point quality mask, including:

[0041] Based on the signal type of the currently input continuous signal, load the judgment thresholds corresponding to the quality indicators of each dimension from the signal feature archive;

[0042] Construct a pointwise quality mask M(n) of the same length as the preprocessed signal. For each valid evaluation unit sample point in the pointwise quality mask, make a judgment based on the following rules in conjunction with the judgment threshold:

[0043] Only when all the indicators in all dimensions meet the requirements can all sample points in the unit be marked as qualified in terms of quality.

[0044] If any dimension of the indicator fails to meet the conditions, the sample point in the corresponding unit will be marked as unqualified.

[0045] As can be seen from the above, the beneficial effects of the embodiments of this application compared with the prior art include at least the following:

[0046] This application embodiment loads processing parameters from a signal feature archive that are compatible with the currently input continuous signal. Therefore, it can process continuous signals from various fields, achieving cross-domain reuse and good versatility. Furthermore, since the processing parameters are combined to perform multi-level cascaded signal processing on the currently input continuous signal, it can effectively improve the noise processing capability and thus effectively improve the signal quality.

[0047] Meanwhile, in this embodiment, by setting a comprehensive quality score and a point-by-point quality mask to evaluate the quality of the current input signal, the comprehensive quality score provides a global quality overview, suitable for rapid decision-making and inter-system comparisons, while the point-by-point quality mask provides a detailed view of the local quality, suitable for applications that need to identify specific periods of degradation in the signal, such as data cleaning and confidence interval analysis. It can be seen that the two complement each other, taking into account both the global and local aspects, and constitute a complete quality assessment of the current input signal, providing a reliable basis for decision-making.

[0048] A second aspect of this application provides an adaptive multi-dimensional continuous signal quality monitoring system, comprising:

[0049] The preprocessing module is used to load processing parameters from the signal feature archive that correspond to the current input continuous signal, and to perform multi-level cascaded signal processing on the current input continuous signal in combination with the processing parameters to obtain the preprocessed signal.

[0050] The multi-dimensional index module is used to perform multi-dimensional calculations on the preprocessed signal to obtain corresponding multi-dimensional quality indices.

[0051] The calculation module is used to load the weight coefficients corresponding to the quality indicators of each dimension from the signal feature archive and calculate the comprehensive quality score; based on the preset strategy, the quality indicators of each dimension are judged independently to obtain the point-by-point quality mask.

[0052] The evaluation module is used to evaluate the quality of the current input signal based on the overall quality score and the point-by-point quality mask.

[0053] A third aspect of this application provides a terminal including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the method described in the first aspect.

[0054] A fourth aspect of this application provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps of the method described in the first aspect.

[0055] The fifth aspect of this application provides a computer program product that, when run on a terminal, causes the terminal to perform the steps of the method described in the first aspect.

[0056] It should be understood that the beneficial effects of the second to fifth aspects mentioned above can be found in the relevant descriptions in the first aspect mentioned above, and will not be repeated here. Attached Figure Description

[0057] To more clearly illustrate the technical solutions in the embodiments of this application, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0058] Figure 1 This is a flowchart illustrating an adaptive multi-dimensional continuous signal quality monitoring method provided in one embodiment of this application;

[0059] Figure 2 This is a flowchart illustrating an adaptive multi-dimensional continuous signal quality monitoring method provided in another embodiment of this application;

[0060] Figure 3 This is a schematic diagram of the structure of the adaptive multi-dimensional continuous signal quality monitoring system provided in the embodiments of this application;

[0061] Figure 4 This is a schematic diagram of the structure of a terminal provided in an embodiment of this application. Detailed Implementation

[0062] In the following description, specific details such as particular system architectures and techniques are set forth for illustrative purposes and not for limitation, in order to provide a thorough understanding of the embodiments of this application. However, those skilled in the art will understand that this application may also be implemented in other embodiments without these specific details. In other instances, detailed descriptions of well-known systems, apparatuses, circuits, and methods have been omitted so as not to obscure the description of this application with unnecessary detail.

[0063] It should be understood that, when used in this specification and the appended claims, the term "comprising" indicates the presence of the described features, integrals, steps, operations, elements and / or components, but does not exclude the presence or addition of one or more other features, integrals, steps, operations, elements, components and / or collections thereof.

[0064] It should be understood that the terminology used in this specification is for the purpose of describing particular embodiments only and is not intended to limit the scope of the application. As used in this specification and the appended claims, the singular forms “a,” “an,” and “the” are intended to include the plural forms unless the context clearly indicates otherwise.

[0065] It should be understood that the term "and / or" as used in this application specification and the appended claims refers to any combination of one or more of the associated listed items and all possible combinations, and includes such combinations.

[0066] It should be understood that the sequence number of each step in this embodiment does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of this application embodiment.

[0067] To illustrate the technical solution described in this application, specific embodiments are provided below.

[0068] like Figure 1 As shown in the embodiment of this application, an adaptive multi-dimensional continuous signal quality monitoring method is provided, which includes:

[0069] Step 1: Load the processing parameters from the signal feature archive that match the current input continuous signal, and perform multi-level cascaded signal processing on the current input continuous signal in combination with the processing parameters to obtain the preprocessed signal;

[0070] Step 2: Perform multi-dimensional calculations on the preprocessed signal to obtain the corresponding multi-dimensional quality indicators;

[0071] Step 3: Load the weight coefficients corresponding to the quality indicators of each dimension from the signal feature archive and calculate the comprehensive quality score; independently judge the quality indicators of each dimension based on the preset strategy to obtain the point-by-point quality mask.

[0072] Step 4: Evaluate the quality of the current input signal based on the overall quality score and the point-by-point quality mask.

[0073] Obviously, by loading processing parameters from the signal feature archive that are compatible with the currently input continuous signal, this embodiment of the application can process continuous signals from various fields, achieving cross-domain reuse and good versatility. Furthermore, since the processing parameters are combined to perform multi-level cascaded signal processing on the currently input continuous signal, the multi-noise processing capability can be effectively improved, thereby effectively improving signal quality.

[0074] Meanwhile, by setting a comprehensive quality score and a point-by-point quality mask to evaluate the quality of the current input signal, the comprehensive quality score provides a global quality overview, suitable for rapid decision-making and inter-system comparisons, while the point-by-point quality mask provides a detailed view of the local quality, suitable for applications that need to identify specific periods of degradation in the signal, such as data cleaning and confidence interval analysis. It can be seen that the two complement each other, taking into account both the global and local aspects, and constitute a complete quality assessment of the current input signal, providing a reliable basis for decision-making.

[0075] In some embodiments, step 1, loading processing parameters from the signal feature archive that correspond to the currently input continuous signal, and performing multi-stage cascaded signal processing on the currently input continuous signal in combination with the processing parameters to obtain a preprocessed signal, may include the following steps:

[0076] Step 11, based on the type of the currently input continuous signal or the sampling frequency f s Load processing parameters that match the currently input continuous signal from the signal feature archive;

[0077] The types of continuous signals can be physiological signals (such as PPG and ECG), industrial environmental monitoring signals (such as temperature and pressure), equipment performance monitoring signals (such as CPU utilization and memory usage), etc. The specific settings can be configured as needed and are not limited here.

[0078] In this embodiment, the processing parameters include the basic coefficient k of the transition detection threshold, the baseline removal weighting coefficients (α, β), and the filter cutoff frequency f. cut The filter order N, as well as the mapping table between signal type and adaptive threshold, can be set as needed and are not limited here.

[0079] To facilitate understanding, the following examples illustrate the types of continuous signals and their corresponding processing parameters pre-stored in a signal feature archive:

[0080] Physiological signals (e.g., PPG, ECG): α=0.8, β=0.2 (emphasizing fast baseline removal), filter cutoff frequency f cut =0.5Hz (Qualcomm);

[0081] Industrial environmental monitoring signals (e.g., temperature, pressure): α=0.3, β=0.7 (focusing on slow baseline removal), f cut =1Hz (low-pass);

[0082] Equipment performance monitoring signals (e.g., CPU utilization): α=0.5, β=0.5, f cut =5Hz (bandpass).

[0083] As can be seen, by calling the processing parameters that match the current input continuous signal from the signal feature database according to the type or sampling frequency of the current input continuous signal, a processing strategy that matches the current input continuous signal can be realized without manual intervention, thereby achieving adaptive adjustment.

[0084] Step 12: Perform multi-stage cascaded signal processing on the current input continuous signal based on the processing parameters to obtain a preprocessed signal, which may specifically include:

[0085] Step 121: Combine the basic coefficient of the jump detection threshold to perform signal jump correction on the current input continuous signal to obtain the jump-corrected signal;

[0086] Step 122: Combining the baseline removal weighting coefficients, filter cutoff frequency, and filter order, and based on the preset morphological opening operation and low-pass filter fusion strategy, baseline drift removal is performed on the transition-corrected signal to obtain the drift-removed signal.

[0087] Step 123: Combine the signal type with the adaptive threshold mapping table to perform dynamic noise suppression on the drift-removed signal to obtain the preprocessed signal.

[0088] Clearly, by employing a three-tiered cascaded signal processing architecture of signal jump correction, baseline drift removal, and dynamic noise suppression, adaptive preprocessing of continuous signals can ensure efficient suppression of continuous signal noise, thereby effectively improving signal quality.

[0089] In this embodiment, step 121, combining the basic coefficient of the jump detection threshold, performs signal jump correction on the currently input continuous signal to obtain the jump-corrected signal, which may include the following:

[0090] Step 1211, assume the current input continuous signal is To calculate the absolute value of the difference sequence of the current input continuous signal, the following formula can be used:

[0091]

[0092] In the above formula, Represents the absolute value of the signal difference sequence. This represents the currently input continuous signal. Indicates the index of the signal sampling point;

[0093] Step 1212: Based on the absolute value of the signal differential sequence and the basic coefficient of the transition detection threshold, dynamically calculate the transition threshold. The calculation formula can be as follows:

[0094]

[0095] In the above formula, Indicates the threshold for the transition. This represents the basic coefficient for the jump detection threshold. This indicates taking the 99th percentile of the sequence;

[0096] Step 1213: When the absolute value of the signal difference sequence is greater than the transition threshold, the sampling point corresponding to the absolute value of the current signal difference sequence is identified as the transition point, that is:

[0097] > At that time, the judgment of the first Each sampling point is a transition point;

[0098] Step 1214: For each transition point, perform multiple rounds of detection (usually 2-3 rounds, depending on the sampling frequency). Calculate the correction amount C by obtaining the median difference between the neighboring signals before and after each transition point. Based on the correction amount, perform amplitude compensation on the signal segment after the transition point, and output the transition-corrected signal, i.e.:

[0099] x1(n) = x(n) + C

[0100] In the above formula, x1(n) represents the signal after jump correction, and C represents the correction amount.

[0101] In this embodiment, step 122, combining the baseline removal weighting coefficients, filter cutoff frequency, and filter order, and based on a preset morphological opening operation and low-pass filter fusion strategy, performs baseline drift removal on the transition-corrected signal to obtain the drift-removed signal, and may include the following steps:

[0102] Step 1221: Use morphological opening operation to perform fast baseline component extraction on the transition-corrected signal to obtain the fast baseline components. The calculation formula can be as follows:

[0103]

[0104] In the above formula, Indicates the fast baseline component, S represents the structuring element (length). ), opening(·) denotes the morphological opening operation;

[0105] Step 1222: Based on the preset filter cutoff frequency and filter order of the Butterworth low-pass filter, filter the transition-corrected signal to obtain the slow baseline component. ;

[0106] Step 1223: Based on the baseline removal weighting coefficients, the fast baseline component and the slow baseline component are weighted and fused to obtain the fused baseline component. The calculation formula can be as follows:

[0107]

[0108] In the above formula, This indicates the fusion of baseline components, and ;

[0109] Step 1224: Subtract the jump-corrected signal from the fused baseline component to obtain the drift-removed signal, i.e.:

[0110]

[0111] In the above formula, This represents the signal after drift removal.

[0112] In this embodiment, step 123, which combines the signal type with the adaptive threshold mapping table to perform dynamic noise suppression on the drift-removed signal to obtain the preprocessed signal, may include the following steps:

[0113] Step 1231: Load the corresponding noise scene judgment coefficients from the signal type and adaptive threshold mapping table according to the signal type of the currently input continuous signal, where:

[0114] The noise scene judgment coefficient is preset according to the signal type, for example:

[0115] For physiological signals, noise is mostly baseline drift and impulse interference, with large peak-to-peak fluctuations. In this case, the noise scene judgment coefficient for the corresponding physiological signal can be configured to 0.8 to sensitively identify high noise. For equipment performance monitoring signals, noise is sudden pulses with moderate peak-to-peak fluctuations. In this case, the noise scene judgment coefficient for the corresponding equipment performance monitoring signal can be configured to 0.1.

[0116] Step 1232: Calculate the standard deviation σ of the initial segment of the signal after removing drift. noise Specifically, it can be:

[0117] The signal start segment is defined as the first 5 seconds of sampled data (or the first round(f)). s (×5) sampling points, take the larger of the two values); σ noise The standard deviation of the sliding window of the initial segment signal (window size round(f)) s Calculate the mean of (×0.5).

[0118] Step 1233: Calculate the peak-to-peak value of the signal after drift removal, which can be calculated using the following formula:

[0119]

[0120] In the above formula, Indicates peak-to-peak value;

[0121] Steps 1, 2, 3, and 4: Calculate the product of the noise scene judgment coefficient and the peak-to-peak value. Determine the noise level by comparing the product with the standard deviation. Based on the judgment result, apply different denoising processes to the drift-removed signal to obtain the preprocessed signal x. pre (n), for example:

[0122] When the product is determined to be less than the standard deviation, it is considered a low-noise scenario. The signal after drift removal is then subjected to 3-point median filtering to obtain a preprocessed signal. When the product is determined to be not less than the standard deviation, it is considered a high-noise scenario. The signal after drift removal is then subjected to db4 wavelet denoising to obtain a preprocessed signal.

[0123] In some embodiments, step 2 involves performing multi-dimensional calculations on the preprocessed signal to obtain multi-dimensional quality indicators, which can specifically include:

[0124] Multiple time-domain and depth feature dimension calculations are performed on the preprocessed signal to obtain multiple time-domain quality indicators and depth feature quality indicators.

[0125] Clearly, by constructing a multi-dimensional evaluation calculation of "time-domain indicators + deep features", we can improve the modeling and quantification capabilities of complex signal quality.

[0126] In this embodiment, multiple temporal dimension calculations and depth feature dimension calculations are performed on the preprocessed signal to obtain multiple temporal quality indicators and depth feature quality indicators, which can be specifically:

[0127] The preprocessed signal is calculated in four time-domain dimensions (volatility, periodicity, signal-to-noise ratio, and trend consistency) and in the depth feature dimension, resulting in four time-domain quality indicators and a depth feature quality indicator: volatility index, periodicity index, signal-to-noise ratio index, and trend consistency index.

[0128] Among them, volatility indicators The calculations may include the following:

[0129] The standard deviation of the sliding window is used to measure the fluctuation of the preprocessed signal, that is:

[0130]

[0131] In the above formula, For window size, The preprocessed signal is the signal value within the window, μ is the mean value within the window, and V represents the degree of fluctuation. The smaller the value, the more stable the signal.

[0132] Normalizing the volatility level V yields a volatility index. Its calculation formula can be:

[0133]

[0134] Among them, cyclical indicators The calculations may include the following:

[0135] Peak detection is performed on the preprocessed signal. If the number of detected peaks is ≥2, the coefficient of variation of the peak interval in the preprocessed signal is calculated. The calculation formula is as follows:

[0136] ,

[0137] In the above formula, Represents the coefficient of variation. The standard deviation of the peak interval, This represents the mean of the peak intervals;

[0138] The periodic index is obtained based on the difference between the preset value and the coefficient of variation. In this embodiment, the preset value is 1, and its calculation formula can be:

[0139]

[0140] In the above formula, Indicates a periodic indicator. The larger the value, the more regular the signal period;

[0141] Among them, the signal-to-noise ratio index The calculations may include the following:

[0142] The ratio of signal power to noise power can be calculated using the following formula:

[0143]

[0144] In the above formula, The power of the effective signal in the preprocessed signal. The power of residual noise in the preprocessed signal. This represents the ratio of signal power to noise power; the larger the value, the better the separation between signal and noise.

[0145] The signal-to-noise ratio (SNR) is obtained by normalizing the ratio of signal power to noise power. The formula for calculating SNR is:

[0146]

[0147] Among them, the trend consistency index The calculations may include the following:

[0148] The reasonableness of a trend can be assessed by evaluating the baseline change rate of the preprocessed signal. Specifically, this can be done by:

[0149] Savitzky-Golay filtering is used to extract the baseline trend from the preprocessed signal;

[0150] The baseline trend of the preprocessed signal is evenly divided into 5 segments, and the fitting slope k is used for each segment. i ;

[0151] The trend consistency index is calculated based on the standard deviation and mean of the fitted slope. The formula for its calculation is as follows:

[0152]

[0153] In the above formula, This represents the standard deviation of the fitted slope for each segment. This represents the mean of the fitted slope for each segment, and represents the trend consistency index. The larger the value, the more consistent the trend stability of the signal.

[0154] Among them, the depth feature quality index F deep The calculations may include the following:

[0155] The preprocessed signal is input into a lightweight one-dimensional CNN structure (3 convolutional layers + 1 global average pooling layer), which outputs a 128-dimensional depth feature vector. This vector is then mapped to a one-dimensional scalar through a fully connected layer (with the sigmoid activation function), thus obtaining the depth feature quality index, which may include the following:

[0156] The preprocessed signal is divided into multiple signal segments of fixed duration (e.g., 5 seconds), with each segment having a length L = round(f) s ×5), and normalize the amplitude of each segment to make it range between [-1,1].

[0157] A lightweight one-dimensional convolutional neural network (1D-CNN) is constructed for deep feature extraction. Its specific structure includes:

[0158] First convolutional layer: Uses 64 convolutional kernels of length 7 with a stride of 1, followed by ReLU activation function and max pooling (pooling size 2).

[0159] The second convolutional layer uses 128 convolutional kernels of length 5 with a stride of 1, followed by a ReLU activation function and max pooling (pooling size 2).

[0160] The third convolutional layer uses 256 convolutional kernels of length 3 with a stride of 1, followed by a ReLU activation function;

[0161] Global feature aggregation: Finally, a global average pooling layer is applied to aggregate the variable-length feature maps into a fixed-length 256-dimensional feature vector;

[0162] Feature dimensionality reduction and scalarization: First, a 256-dimensional vector is mapped to a 128-dimensional deep feature vector through a fully connected layer, and then it is transformed into a scalar deep feature quality index through a fully connected layer (output dimension 1, activation function Sigmoid).

[0163] Each signal segment corresponds to a deep feature quality index, which is used for subsequent comprehensive quality score calculation. This value can capture complex patterns and contextual information in the signal that are difficult to represent by traditional time-domain indicators.

[0164] Clearly, by introducing 1D-CNN deep features, the accuracy of quality modeling for complex noise and non-stationary signals can be effectively improved, especially suitable for scenarios such as physiological signals and weak industrial signals.

[0165] In some embodiments, step 3 involves loading the weight coefficients corresponding to the quality indicators of each dimension from the signal feature archive and calculating the comprehensive quality score; and independently judging each quality indicator based on a preset strategy to obtain a point-by-point quality mask.

[0166] The process of loading the signal feature archive and calculating the weighting coefficients corresponding to the quality indicators of each dimension, and then calculating the overall quality score, may include the following steps:

[0167] Step 301: Load the weight coefficients corresponding to the quality indicators of each dimension from the signal feature archive according to the signal type of the current input continuous signal;

[0168] The weighting coefficient can be expressed as follows:

[0169] { , , , , }, and the weight coefficients satisfy: ;

[0170] The weight coefficients are obtained by training a random forest regression model: taking multi-domain signal samples (physiological signals, industrial signals, equipment signals) as input and the signal quality scores labeled by domain experts as labels, the weight coefficients are optimized by minimizing the prediction error (mean square error), and finally forming weight configuration templates for different signal types, which are stored in the signal feature archive.

[0171] Step 302: Weight and fuse the quality indicators of each dimension with their corresponding weighting coefficients to obtain the comprehensive quality score, which can be expressed by the following formula:

[0172]

[0173] In the above formula, Q represents the overall quality score, and Q∈[0,1]. The larger the Q value, the better the signal quality.

[0174] In this embodiment of the application, each dimension of quality indicators is independently judged based on a preset strategy to obtain a point-by-point quality mask, which may specifically include the following steps:

[0175] Step 311: Based on the signal type of the currently input continuous signal, load the judgment thresholds corresponding to each dimension of quality indicators from the signal feature archive, whereby...

[0176] The judgment thresholds include: low volatility threshold V low High volatility threshold V high Low threshold SNR min Minimum periodicity threshold P minLow threshold for trend consistency T min and depth feature qualification threshold F min ;

[0177] Step 312: Construct a pointwise quality mask M(n) of the same length as the preprocessed signal. For each valid evaluation unit (usually global or an analysis window) in the pointwise quality mask, determine the quality based on the following rules, combined with the decision threshold:

[0178] Only when all the indicators in all dimensions meet the requirements are all the sample points in the unit marked as qualified (M(n)=1);

[0179] If any dimension of the indicator fails to meet the conditions, the sample point in the corresponding unit is marked as unqualified (M(n)=0);

[0180] Specifically, the indicator requirements for each dimension can be set as follows:

[0181] 1.V norm ≥V low And V norm ≤V high (Normal volatility);

[0182] 2. P ≥ P min (Periodic qualification);

[0183] 3. SNR norm ≥SNR min (Signal-to-noise ratio is acceptable);

[0184] 4. T≥T min (Consistent trend);

[0185] 5.F deep ≥F min (Depth features qualified, F) min It was derived through statistical analysis of signal samples from multiple domains.

[0186] Clearly, pointwise quality masks provide a quality pass / fail marker for a signal at the sample level, that is, they provide fine-grained evaluation at the sample level, ensuring that any severe degradation in a single dimension can be effectively identified, making them suitable for refined data analysis.

[0187] In some embodiments, step 4 involves evaluating the quality of the current input signal based on the overall quality score and the point-by-point quality mask, wherein:

[0188] The overall quality score provides a global, weighted average overview of quality, suitable for rapid decision-making and inter-system comparisons;

[0189] Point-by-point quality masks provide a detailed view of local quality and are suitable for applications that need to identify specific periods of degradation in a signal, such as data cleaning and confidence interval analysis.

[0190] Clearly, by combining the overall quality score and the point-by-point quality mask, a complete quality assessment output can be formed.

[0191] The above scheme is suitable for independent evaluation of a single continuous signal, i.e., single-mode quality evaluation. However, in practice, there are often multiple synchronous signals, i.e., multi-mode signals. In order to achieve quality detection of multi-mode signals, such as... Figure 2 As shown, embodiments of this application also include:

[0192] Step 5: Determine if multiple synchronization signals exist. If multiple synchronization signals exist, indicating the presence of multimodal signals, perform a preset multimodal fusion evaluation to obtain a multimodal joint score.

[0193] Multimodal fusion evaluation may include the following steps:

[0194] Step 51: When multimodal signals exist, an attention mechanism is used to dynamically allocate the weights of each modal signal. Specifically, this can be done as follows:

[0195] The feature vector of each modality is input into a 2-layer fully connected network (MLP, input dimension 10, hidden layer dimension 20, output dimension 1) to calculate its attention score, and the weights are obtained by normalization using the Softmax function. The calculation formula can be as follows:

[0196]

[0197] In the above formula, F represents the weight of the i-th mode. i This represents the eigenvector of the i-th mode;

[0198] Step 52: Based on the calculated attention weights, the independent quality scores of each modality are weighted and fused to obtain the multimodal joint score. The calculation formula can be used as follows:

[0199]

[0200] In the above formula, Qi represents the overall quality score of each mode;

[0201] In addition, when multimodal signals are present, synchronous alignment and feature fusion splicing processing can be performed on the multimodal signals.

[0202] The multimodal signal synchronization alignment processing can be as follows:

[0203] The system achieves precise synchronization of multiple signals based on hardware timestamps. For signals with different sampling rates, linear interpolation is used for resampling to unify them to the highest sampling rate preset by the system or a specified frequency (such as 250Hz), ensuring that the data points correspond one-to-one.

[0204] Feature fusion and stitching processing can be:

[0205] The feature vectors of each signal are concatenated. For example, for ECG and PPG dual-mode signals, the feature vector of the ECG signal is... ;

[0206] PPG signal feature vector ;

[0207] Feature splicing is as follows: (10 dimensions).

[0208] Clearly, the above scheme can achieve synchronous fusion evaluation of multi-source signals, solve the problem of inconsistent judgment of multi-modal signal quality, and thus improve the reliability of multi-dimensional decision-making.

[0209] In this embodiment of the application, the method further includes: calculating the pointwise quality mask of each modal signal respectively, and fusing the pointwise quality masks of each modal signal to obtain a fused quality mask;

[0210] Based on the multimodal joint score, the comprehensive quality score of each modality, and the fusion quality mask, the quality of the current input signal is evaluated, and a comprehensive decision is made to determine whether to trigger an early warning. Specifically, this can be done as follows:

[0211] The overall quality score is used for global status monitoring, triggering system-level alerts (e.g., when Q < 0.3), or ranking the quality of different signal sources.

[0212] Pointwise quality masks are used in downstream data analysis tasks, such as when calculating statistical features or training models, to improve the reliability of results by using only high-quality data segments that are qualified by the pointwise quality mask.

[0213] Multimodal joint quality score is used to output multi-source data segments that are "jointly qualified" (based on fusion mask filtering), which can directly provide "reliable data input" for downstream modeling and analysis, avoid interference from low-quality single signals or inconsistent multi-source signals on subsequent decisions, and indirectly improve the reliability of the entire monitoring and analysis system.

[0214] Furthermore, an early warning is triggered when both the multimodal joint score and the overall quality score are not greater than a preset first threshold, and an early warning is also triggered when the unqualified area of ​​the point-by-point quality mask is greater than a second threshold.

[0215] Clearly, by making comprehensive decisions based on multimodal joint scores, overall quality scores, and point-by-point quality masks to trigger early warnings, users can be promptly alerted.

[0216] In some embodiments, the method further includes: step 6, optimizing and updating the signal feature database through online incremental learning, which may specifically include the following steps:

[0217] Step 61: Continuously collect new sample data and manage the samples, which may specifically include:

[0218] A circular buffer is used to store newly acquired signal samples and manually fed quality labels;

[0219] Introduce a hard sample mining mechanism to manage samples, for example:

[0220] When the overall quality score Q of a sample is in the fuzzy range [0.4, 0.6] and more than 3 dimensional indicators are close to their judgment threshold ("close" is defined as the deviation of the indicator value from the corresponding threshold ≤ 5%), it is judged as a "difficult sample". During training, difficult samples are given a weight of 2.0 (ordinary samples have a weight of 1.0) to improve the model's ability to distinguish boundary cases.

[0221] Step 62: When any preset trigger condition is met by monitoring, fine-tuning is initiated. The trigger condition can be that the sample accumulation reaches 1000, the signal standard deviation change is >15% ("signal standard deviation change" is defined as: the absolute value of the difference between the standard deviation of the current 100 samples and the historical standard deviation mean in the archive is >15% of the historical mean), or user-initiated triggering, etc. The specific settings can be configured as needed and are not limited here.

[0222] Step 63: Fine-tune according to the preset incremental algorithm, for example:

[0223] The Elastic Weight Consolidation (EWC) algorithm is used to prevent catastrophic forgetting. It mainly fine-tunes the deep feature extractor and dynamic weight decision model, and optimizes the model parameters through lightweight gradient descent (learning rate 1e−5, 10 iterations).

[0224] Step 64: After fine-tuning, generate parameter update instructions. Based on these instructions, collaboratively update the signal feature archive and related running instances. Specifically, this can be done as follows:

[0225] After fine-tuning, the updated parameters (such as baseline removal weighting coefficients, filter cutoff frequency, weight coefficients, deep feature extractor weights, etc.) are synchronized in real time to the running instances of the three-level signal processing module and the quality assessment model through the system's internal message bus or parameter server.

[0226] The parameter synchronization triggering mechanism is as follows: After fine-tuning is completed, a parameter update instruction is generated immediately and pushed to all running module instances through the system's internal message bus (delay ≤ 100ms); During the synchronization process, the module adopts a "double buffering mechanism" - while loading new parameters, the old parameters continue to be used to process the signal, and the new parameters are switched to after the new parameters are verified (verified by 100 test samples that the prediction error rate of Q is ≤ 5%). This avoids signal processing interruption and ensures that the processing and evaluation strategies can dynamically adapt to changes in signal characteristics, achieving end-to-end self-adaptation.

[0227] Clearly, through online incremental learning, the signal feature database can dynamically adapt to new scenarios and noise patterns without the need for manual algorithm redevelopment.

[0228] In summary, the embodiments of this application employ a three-level cascaded signal processing architecture and dynamically optimize and update the signal feature archive. By combining time-domain statistical features, morphological features, periodic features, and depth features for multi-dimensional evaluation, it outputs quantized quality scores and point-by-point quality masks. Compared with existing technologies, this application is the first to achieve intelligent quality monitoring of various continuous signals (such as physiological signals, industrial monitoring signals, equipment performance signals, etc.). It solves the problems of poor versatility, insufficient multi-noise processing capabilities, and lack of online learning capabilities in existing technologies. It supports multiple signal types such as PPG, ECG, and CPU utilization, and has comprehensive multi-noise processing capabilities (jumping, baseline drift, and random noise collaborative suppression). It also features standardized evaluation, strong adaptability, and strong decision support capabilities, and can be widely used in fields such as medical health, industrial monitoring, and equipment diagnosis.

[0229] Based on the same inventive concept, embodiments of this application also provide an adaptive multi-dimensional continuous signal quality monitoring system, such as... Figure 3 As shown, the system includes:

[0230] Preprocessing module 1 is used to load processing parameters from the signal feature archive that correspond to the current input continuous signal, and to perform multi-level cascaded signal processing on the current input continuous signal in combination with the processing parameters to obtain the preprocessed signal.

[0231] Multidimensional index module 2 is used to perform multidimensional calculations on the preprocessed signal to obtain multidimensional quality indices.

[0232] Calculation module 3 is used to load the weight coefficients corresponding to the quality indicators of each dimension from the signal feature archive and calculate the comprehensive quality score; based on the preset strategy, the quality indicators of each dimension are judged independently to obtain the point-by-point quality mask;

[0233] Evaluation module 4 is used to evaluate the quality of the current input signal based on the overall quality score and the point-by-point quality mask.

[0234] Obviously, by loading processing parameters from the signal feature archive that are compatible with the currently input continuous signal, this embodiment of the application can process continuous signals from various fields, achieving cross-domain reuse and good versatility. Furthermore, since the processing parameters are combined to perform multi-level cascaded signal processing on the currently input continuous signal, the multi-noise processing capability can be effectively improved, thereby effectively improving signal quality.

[0235] Meanwhile, by setting a comprehensive quality score and a point-by-point quality mask to evaluate the quality of the current input signal, the comprehensive quality score provides a global quality overview, suitable for rapid decision-making and inter-system comparisons, while the point-by-point quality mask provides a detailed view of the local quality, suitable for applications that need to identify specific periods of degradation in the signal, such as data cleaning and confidence interval analysis. It can be seen that the two complement each other, taking into account both the global and local aspects, and constitute a complete quality assessment of the current input signal, providing a reliable basis for decision-making.

[0236] This application also provides a terminal, such as... Figure 4 As shown, the terminal of this embodiment includes: at least one processor 40 ( Figure 4 (Only one is shown in the diagram), memory 41, and computer program 42 stored in said memory 41 and executable on said at least one processor 40, which, when executed, implements the steps in any of the above method embodiments.

[0237] The terminal 4 can be a computing device such as a desktop computer, laptop, handheld computer, or cloud server. The terminal 4 may include, but is not limited to, a processor 40 and a memory 41. Those skilled in the art will understand that... Figure 4 This is merely an example of terminal 4 and does not constitute a limitation on terminal 4. It may include more or fewer components than shown, or combine certain components, or different components. For example, the terminal may also include input / output devices, network access devices, buses, etc.

[0238] The processor 40 can be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor can be a microprocessor or any conventional processor.

[0239] The memory 41 can be an internal storage unit of the terminal 4, such as a hard disk or memory of the terminal 4. The memory 41 can also be an external storage device of the terminal 4, such as a plug-in hard disk, smart media card (SMC), secure digital (SD) card, flash card, etc., equipped on the terminal 4. Furthermore, the memory 41 can include both internal storage units and external storage devices of the terminal 4. The memory 41 is used to store the computer program and other programs and data required by the terminal. The memory 41 can also be used to temporarily store data that has been output or will be output.

[0240] In specific implementations, the terminals described in the embodiments of this application include, but are not limited to, other portable devices such as mobile phones, laptop computers, or tablet computers with touch-sensitive surfaces (e.g., touchscreen displays and / or touchpads). It should also be understood that in some embodiments, the device is not a portable communication device, but a desktop computer with touch-sensitive surfaces (e.g., touchscreen displays and / or touchpads).

[0241] The terminal supports a variety of applications, such as one or more of the following: drawing applications, presentation applications, word processing applications, website creation applications, disc burning applications, spreadsheet applications, game applications, telephone applications, video conferencing applications, email applications, instant messaging applications, exercise support applications, photo management applications, digital camera applications, digital camcorder applications, web browsing applications, digital music player applications, and / or digital video player applications.

[0242] Various applications that can run on a terminal can use at least one common physical user interface device, such as a touch-sensitive surface. One or more functions of the touch-sensitive surface and the corresponding information displayed on the terminal can be adjusted and / or changed between and / or within applications. In this way, the terminal's common physical architecture (e.g., the touch-sensitive surface) can support various applications with user interfaces that are intuitive and transparent to the user.

[0243] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the above-described division of functional units and modules is merely an example. In practical applications, the above functions can be assigned to different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above. The functional units and modules in the embodiments can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit. Furthermore, the specific names of the functional units and modules are only for easy differentiation and are not intended to limit the scope of protection of this application. The specific working process of the units and modules in the above system can be referred to the corresponding process in the foregoing method embodiments, and will not be repeated here.

[0244] In the above embodiments, the descriptions of each embodiment have different focuses. For parts that are not described in detail or recorded in a certain embodiment, please refer to the relevant descriptions of other embodiments.

[0245] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0246] In the embodiments provided in this application, it should be understood that the disclosed devices / terminals and methods can be implemented in other ways. For example, the device / terminal embodiments described above are merely illustrative. For instance, the division of modules or units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the displayed or discussed mutual coupling or direct coupling or communication connection may be through some interfaces; the indirect coupling or communication connection between devices or units may be electrical, mechanical, or other forms.

[0247] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0248] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0249] If the integrated module / unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, all or part of the processes in the methods of the above embodiments can also be implemented by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium, and when executed by a processor, it can implement the steps of the various method embodiments described above. The computer program includes computer program code, which can be in the form of source code, object code, executable files, or certain intermediate forms. The computer-readable medium can include: any entity or device capable of carrying the computer program code, recording media, USB flash drives, portable hard drives, magnetic disks, optical disks, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signals, telecommunication signals, and software distribution media, etc. It should be noted that the content included in the computer-readable medium can be appropriately added or removed according to the requirements of legislation and patent practice in the jurisdiction. For example, in some jurisdictions, according to legislation and patent practice, computer-readable media do not include electrical carrier signals and telecommunication signals.

[0250] The methods described in this application can be implemented in whole or in part by a computer program product. When the computer program product is run on a terminal, the terminal executes the steps in the various method embodiments described above.

[0251] The above-described embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application, and should all be included within the protection scope of this application.

Claims

1. An adaptive multi-dimensional continuous signal quality monitoring method, characterized in that, include: The system loads processing parameters from the signal feature archive that correspond to the currently input continuous signal. These parameters are then used to perform multi-stage cascaded signal processing on the current input continuous signal to obtain a preprocessed signal, including: Based on the type or sampling frequency of the currently input continuous signal, load the processing parameters that match the currently input continuous signal from the signal feature archive; The current input continuous signal is subjected to three-stage cascaded signal processing based on the processing parameters to obtain a preprocessed signal; Multi-dimensional calculations are performed on the preprocessed signal to obtain corresponding multi-dimensional quality indicators; Load the weight coefficients corresponding to each dimension of quality indicators from the signal feature archive, and calculate the comprehensive quality score, including: Based on the signal type of the current input continuous signal, load the weight coefficients corresponding to each dimension of quality indicators from the signal feature archive; then, weight and fuse each dimension of quality indicators with the corresponding weight coefficients to obtain a comprehensive quality score. Based on a preset strategy, each dimension of quality indicators is judged independently to obtain a point-by-point quality mask, including: loading the judgment threshold corresponding to each dimension of quality indicators from the signal feature archive according to the signal type of the current input continuous signal. Construct a pointwise quality mask M(n) of the same length as the preprocessed signal. For each valid evaluation unit sample point in the pointwise quality mask, make a judgment based on the following rules in conjunction with the judgment threshold: Only when all the indicators in all dimensions meet the requirements can all sample points in the unit be marked as qualified in terms of quality. If any dimension of the indicator fails to meet the conditions, the sample point in the corresponding unit will be marked as unqualified. Specifically, the indicator requirements for each dimension are as follows: V norm ≥V low And V norm ≤V high ; P≥P min ; SNR norm ≥SNR min ; T≥T min ; F deep ≥F min； In the above formula, V low Indicates a low volatility threshold; V high Indicates a high threshold for volatility; V norm Indicates volatility index; P min Represents the minimum periodicity threshold; P represents the periodicity index; SNR min Indicates a low signal-to-noise ratio threshold; SNR norm T represents the signal-to-noise ratio; min Indicates a low threshold for trend consistency; T represents the trend consistency index; F min F represents the threshold for qualifying depth features. deep Indicates the quality index of depth features; The quality of the current input signal is evaluated based on the overall quality score and the point-by-point quality mask.

2. The method according to claim 1, characterized in that, The combined processing parameters are used to perform three-stage cascaded signal processing on the currently input continuous signal to obtain a preprocessed signal, including: By combining the basic coefficient of the jump detection threshold, the current input continuous signal is corrected for jumps to obtain the jump-corrected signal; By combining the baseline removal weighting coefficients, filter cutoff frequency, and filter order, and based on a preset morphological opening operation and low-pass filter fusion strategy, baseline drift removal is performed on the transition-corrected signal to obtain the drift-removed signal. By combining the signal type with an adaptive threshold mapping table, dynamic noise suppression is performed on the drift-removed signal to obtain the preprocessed signal.

3. The method according to claim 2, characterized in that, The step of combining the jump detection threshold base coefficient to perform signal jump correction on the currently input continuous signal to obtain the jump-corrected signal includes: Calculate the absolute value of the difference sequence of the currently input continuous signal; The jump threshold is dynamically calculated based on the absolute value of the signal differential sequence and the basic coefficient of the jump detection threshold. When the absolute value of the signal difference sequence is greater than the jump threshold, the sampling point corresponding to the absolute value of the current signal difference sequence is identified as the jump point; For each transition point, multiple rounds of detection are performed to obtain the median difference of the neighboring signals before and after each transition point to calculate the correction amount. Based on the correction amount, the amplitude of the signal segment after the transition point is compensated to obtain the transition-corrected signal.

4. The method according to claim 2, characterized in that, The method combines baseline removal weighting coefficients, filter cutoff frequency, and filter order, and based on a preset morphological opening operation and low-pass filtering fusion strategy, performs baseline drift removal on the transition-corrected signal to obtain the drift-removed signal, including: Morphological opening operation is used to extract the fast baseline component of the jump-corrected signal to obtain the fast baseline component; the jump-corrected signal is filtered according to the preset filter cutoff frequency and filter order of the Butterworth low-pass filter to obtain the slow baseline component. The fast baseline component and the slow baseline component are weighted and fused according to the baseline removal weighting coefficient to obtain the fused baseline component; the signal after jump correction is subtracted from the fused baseline component to obtain the signal after drift removal.

5. The method according to claim 2, characterized in that, The method of combining signal type and adaptive threshold mapping table to perform dynamic noise suppression on the drift-removed signal to obtain a preprocessed signal includes: Based on the signal type of the current input continuous signal, the corresponding noise scene judgment coefficient is loaded from the signal type and adaptive threshold mapping table; Calculate the standard deviation of the initial segment of the signal after removing drift; Calculate the peak-to-peak value of the signal after drift removal; The noise scene judgment coefficient and the peak-to-peak value are calculated. The noise level is judged by comparing the product with the standard deviation. Different denoising processes are applied to the drift-removed signal according to the noise level to obtain the preprocessed signal.

6. The method according to claim 1, characterized in that, The multi-dimensional calculation of the preprocessed signal yields multi-dimensional quality indicators, specifically: Multiple time-domain and depth feature dimension calculations are performed on the preprocessed signal to obtain multiple time-domain quality indicators and depth feature quality indicators.

7. An adaptive multi-dimensional continuous signal quality monitoring system, characterized in that, include: The preprocessing module loads processing parameters from the signal feature archive that correspond to the currently input continuous signal. It then performs multi-stage cascaded signal processing on the input continuous signal using these parameters to obtain a preprocessed signal, including: Based on the type or sampling frequency of the currently input continuous signal, load the processing parameters that match the currently input continuous signal from the signal feature archive; The current input continuous signal is subjected to three-stage cascaded signal processing based on the processing parameters to obtain a preprocessed signal; The multi-dimensional index module is used to perform multi-dimensional calculations on the preprocessed signal to obtain corresponding multi-dimensional quality indices. The calculation module loads the weight coefficients corresponding to each dimension of quality indicators from the signal feature archive and calculates the comprehensive quality score, including: Based on the signal type of the current input continuous signal, load the weight coefficients corresponding to each dimension of quality indicators from the signal feature archive; then, weight and fuse each dimension of quality indicators with the corresponding weight coefficients to obtain a comprehensive quality score. Based on a preset strategy, each dimension of quality indicators is judged independently to obtain a point-by-point quality mask, including: loading the judgment threshold corresponding to each dimension of quality indicators from the signal feature archive according to the signal type of the current input continuous signal. Construct a pointwise quality mask M(n) of the same length as the preprocessed signal. For each valid evaluation unit sample point in the pointwise quality mask, make a judgment based on the following rules in conjunction with the judgment threshold: Only when all the indicators in all dimensions meet the requirements can all sample points in the unit be marked as qualified in terms of quality. If any dimension of the indicator fails to meet the conditions, the sample point in the corresponding unit will be marked as unqualified. The evaluation module is used to evaluate the quality of the current input signal based on the overall quality score and the point-by-point quality mask.

Images