A risk assessment method and system based on multi-bioelectric signal fusion

Abstract

The present invention discloses a risk assessment method and system based on the fusion of multiple bioelectric signals, which relates to the technical field related to signal processing and analysis. The method includes: extracting biological labels to obtain the original multiple bioelectric signal sets; performing noise reduction processing on the original bioelectric signal sets to obtain multiple bioelectric signals; obtaining multiple biological features based on the multiple bioelectric signals; establishing a multi-head feature extractor to obtain multiple signal features; fusing the multiple signal features with the multiple biological features to generate a fusion vector; analyzing and calculating the features of the fusion vector to obtain the fusion features; constructing a risk assessor to obtain the damage risk; obtaining damage information based on the feedback control of the error value between the damage risk and the real-time data. The present invention solves the technical problems existing in the existing multi-bioelectric signal fusion risk assessment, such as inaccurate acquisition of multiple bioelectric signals and inaccurate feature extraction, which leads to poor fusion effect and inaccurate and unreliable risk assessment results.

Description

Technical Field

[0001] The present invention relates to the technical field related to signal processing and analysis, and in particular to a risk assessment method and system based on multi-bioelectric signal fusion. Background Art

[0002] With the rapid development of biosensor technology and the widespread acquisition of medical data, we can more easily obtain and analyze bioelectric signal data. Risk assessment of bioelectric signals has become an important research direction in the field of medical health, which can provide health status assessment and risk prediction. However, traditional single bioelectric signal assessment can only provide limited bioelectric signal information, and cannot obtain accurate and comprehensive bioelectric signal data and perform feature extraction on it. Moreover, the amount of bioelectric signal data is becoming increasingly large, and traditional risk assessment methods have difficulties and limitations when processing large amounts of complex data.

[0003] Therefore, the existing multi-bioelectric signal fusion risk assessment has technical problems such as inaccurate acquisition of multiple bioelectric signals and inaccurate feature extraction, which leads to poor fusion effect and inaccurate and unreliable risk assessment results. Summary of the Invention

[0004] Based on this, the embodiment of the present application provides a risk assessment method and system based on the fusion of multiple bioelectric signals, which solves the technical problems of the existing multi-bioelectric signal fusion risk assessment, such as inaccurate acquisition of multiple bioelectric signals and inaccurate feature extraction, resulting in poor fusion effect and inaccurate and unreliable risk assessment results, thereby achieving the technical effect of more accurate and reliable risk assessment results.

[0005] A first aspect of an embodiment of the present application provides a risk assessment method based on multi-bioelectric signal fusion, the method comprising:

[0006] Extracting multiple biological tags and obtaining original multiple bioelectrical signal sets corresponding to the multiple biological tags;

[0007] Performing noise reduction processing on the original multiple bioelectrical signal set to obtain multiple bioelectrical signals;

[0008] Extracting biometric features based on the multiple bioelectric signals to obtain multiple biometric features;

[0009] Establishing a multi-head feature extractor to extract signal features from the multiple bioelectric signals to obtain multiple signal features;

[0010] fusing the multiple signal features with the multiple biometric features to generate a fusion vector;

[0011] Analyzing and calculating features of the fusion vector through a first fully connected layer and a second fully connected layer to obtain fusion features;

[0012] Constructing a risk assessor to perform risk assessment on the fusion features and obtain damage risk;

[0013] Error analysis is performed based on the damage risk and real-time data of corresponding characteristics, and feedback control is performed based on the error value to obtain damage information.

[0014] A second aspect of the embodiments of the present application provides a risk assessment system based on multi-bioelectric signal fusion, the system comprising:

[0015] an original multi-bioelectric signal set acquisition module, the original multi-bioelectric signal set acquisition module being used to extract multi-biological tags and obtain the original multi-bioelectric signal set corresponding to the multi-biological tags;

[0016] a multi-bioelectric signal acquisition module, configured to perform noise reduction processing on the original multi-bioelectric signal set to acquire multi-bioelectric signals;

[0017] a multi-biometric feature acquisition module, configured to extract biometric features based on the multiple bioelectric signals to obtain multiple biometric features;

[0018] A multi-signal feature acquisition module, wherein the multi-signal feature acquisition module is used to establish a multi-head feature extractor, perform signal feature extraction on the multiple bioelectric signals, and obtain multi-signal features;

[0019] a fusion vector acquisition module, configured to fuse the multiple signal features with the multiple biometric features to generate a fusion vector;

[0020] A fusion feature acquisition module, configured to analyze and calculate features of the fusion vector through a first fully connected layer and a second fully connected layer to obtain a fusion feature;

[0021] An injury risk acquisition module, wherein the injury risk acquisition module is used to construct a risk assessor, perform risk assessment on the fusion features, and obtain injury risk;

[0022] The damage information feedback control module is used to perform error analysis based on the damage risk and real-time data of corresponding characteristics, and perform feedback control based on the error value to obtain damage information.

[0023] One or more technical solutions provided in this application have at least the following technical effects or advantages:

[0024] The present application extracts multiple biological labels, obtains the original multiple bioelectric signal sets corresponding to the multiple biological labels, performs noise reduction processing on the original multiple bioelectric signal sets, obtains multiple bioelectric signals, extracts biometric features based on the multiple bioelectric signals, obtains multiple biometric features, establishes a multi-head feature extractor, extracts signal features from the multiple bioelectric signals, obtains multiple signal features, fuses the multiple signal features with the multiple biometric features, generates a fusion vector, analyzes and calculates features of the fusion vector through the first fully connected layer and the second fully connected layer, obtains fusion features, constructs a risk assessor, performs risk assessment on the fusion features, obtains damage risk, performs error analysis based on the damage risk and real-time data of the corresponding features, performs feedback control based on the error value to obtain damage information, solves the technical problems of inaccurate acquisition of multiple bioelectric signals and feature extraction in the existing multi-bioelectric signal fusion risk assessment, resulting in poor fusion effect and inaccurate and unreliable risk assessment results, thereby achieving a technical effect of more accurate and reliable risk assessment results. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0026] Figure 1 A flow chart of a risk assessment method based on multi-bioelectric signal fusion provided in an embodiment of the present application;

[0027] Figure 2 A schematic diagram of the process of extracting biometric features in a risk assessment method based on multi-bioelectric signal fusion provided in an embodiment of the present application;

[0028] Figure 3 A schematic diagram of a system structure for risk assessment based on multi-bioelectric signal fusion provided in an embodiment of the present application;

[0029] Explanation of the accompanying drawings: original multi-bioelectric signal set acquisition module 10, multi-bioelectric signal acquisition module 20, multi-biometric feature acquisition module 30, multi-signal feature acquisition module 40, fusion vector acquisition module 50, fusion feature acquisition module 60, damage risk acquisition module 70, damage information feedback control module 80. DETAILED DESCRIPTION

[0030] The embodiment of the present application provides a risk assessment method based on the fusion of multiple bioelectric signals, which solves the technical problems existing in the existing multi-bioelectric signal fusion risk assessment, such as inaccurate acquisition of multiple bioelectric signals and inaccurate feature extraction, resulting in poor fusion effect and inaccurate and unreliable risk assessment results.

[0031] To make the purpose, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only some embodiments of the present application, not all embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of this application.

[0032] It should be noted that the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions. For example, a process, method, system, product or server that includes a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or modules that are not clearly listed or are inherent to these processes, methods, products or devices.

[0033] Example 1

[0034] like Figure 1 As shown, an embodiment of the present application provides a method for risk assessment based on multi-bioelectric signal fusion, the method comprising:

[0035] Specifically, risk assessment based on the fusion of multiple bioelectric signals refers to the collection, processing and analysis of different types of bioelectric signals from multiple bioelectric signal sources, including electrocardiogram (ECG), electroencephalogram (EEG), electromyogram (EMG), etc. By integrating the temporal and spatial characteristics, frequency characteristics, mutual relationships and other information characteristics of these bioelectric signals, combined with advanced data processing and analysis and model building technologies, it can provide a more comprehensive and accurate understanding of the individual's physiological state and function, and provide a more comprehensive health status assessment and risk prediction.

[0036] Extracting multiple biological tags and obtaining original multiple bioelectrical signal sets corresponding to the multiple biological tags;

[0037] Preferably, multiple biological tags are extracted to obtain the original multiple biological electrical signal sets corresponding to the multiple biological tags, wherein the multiple biological tags refer to multiple physiological parameters or biological signal tags measured or monitored in the organism, such as electrocardiogram, electroencephalogram, electromyogram, electrooculogram, blood pressure, blood oxygen saturation, respiratory rate, etc. The electrocardiogram records the electrical activity of the heart and is used to evaluate heart function and diagnose heart disease. The electroencephalogram measures the potential changes on the scalp and is used to study brain activity, sleep and epilepsy. The electromyogram records the electrical activity of the muscle and is used to evaluate muscle function, muscle lesions and neuromuscular diseases. The electrooculogram records eye movements and changes in eye potential. The blood pressure measures the arterial pressure during heart contraction and relaxation and is used to evaluate cardiovascular health. Blood oxygen saturation is used to evaluate respiratory and circulatory function, and respiratory rate is used to evaluate respiratory function and respiratory diseases. The raw bioelectric signal set refers to a collection of unprocessed bioelectric signal data obtained from detection in a living organism, which can be measurement values ​​from different physiological parameters. Raw bioelectric signals are usually recorded by bioelectric signal acquisition equipment, such as bioelectric amplifiers, bioelectric collectors, bioelectric signal amplifiers, etc. Raw bioelectric signals may contain a lot of unnecessary or interfering information, such as noise, motion artifacts, etc. Before subsequent analysis and processing, the raw bioelectric signals usually need to be preprocessed, such as filtering, denoising, downsampling, etc., to reduce interference and improve the data quality of multiple bioelectric signals.

[0038] Performing noise reduction processing on the original multiple bioelectrical signal set to obtain multiple bioelectrical signals;

[0039] Specifically, the original multi-bioelectric signal set needs to be denoised to obtain multi-bioelectric signals, which may include but is not limited to mean filtering, median filtering, wavelet transform, adaptive filtering, Kalman filtering, etc. Among them, mean filtering is to average multiple bioelectric signals at each time point to reduce the influence of random noise. It is used for situations where the noise is small and normally distributed. Median filtering is to sort multiple bioelectric signals at each time point by numerical value and select the median as the filtered value, which can effectively remove noise pulses. Wavelet transform is to use wavelet signals to decompose multiple bioelectric signals into sub-bands of different frequencies, and then filter each sub-band. It has advantages in processing non-stationary signals and multi-scale signals. Adaptive filtering is to dynamically adjust the filter parameters according to the statistical characteristics of the signal to adapt to different signal characteristics. It has good adaptability when processing different bioelectric signals. Kalman filtering is based on the state space model. It filters and predicts bioelectric signals through real-time measurement of the signal and estimation of system dynamics. It has good effect when processing signals with random noise.

[0040] Furthermore, the original multi-bioelectric signal set is subjected to noise reduction processing to obtain multi-bioelectric signals, and the method includes:

[0041] Reducing the signal-to-noise ratio of the original multiple bioelectrical signal sets through a filter to obtain a first electrical signal set;

[0042] The first electrical signal set is normalized according to a normalization amplitude threshold to obtain the multiple bioelectrical signals.

[0043] Preferably, the signal-to-noise ratio of the original multiple bioelectric signal sets is reduced by a filter to obtain a first electrical signal set, and then the data of the first electrical signal set is normalized according to a normalized amplitude threshold to obtain multiple bioelectric signals, wherein the signal-to-noise ratio refers to the ratio or degree of difference between the bioelectric signal and the noise, which is usually used to measure the relative strength and clarity between the useful information data contained in the bioelectric signal and the background noise. Reducing the signal-to-noise ratio of the original multiple bioelectric signal sets refers to reducing the difference between the bioelectric signal and the noise, for example, by filtering or noise reduction algorithm, so that the bioelectric signal is easier to identify, extract and analyze, thereby obtaining more accurate and reliable results. The first electrical signal set is the electrical signal data obtained after noise reduction of the multiple bioelectric signal sets. According to the data, a suitable normalized amplitude threshold is determined. It can be set according to the characteristics and requirements of the bioelectric signal and is between 0 and 1. For example, it can be selected according to the ratio of the maximum value to the minimum value of the bioelectric signal or according to the range of change of the bioelectric signal. Each bioelectric signal data in the first electric signal set is compared with the set normalized amplitude threshold. If the amplitude of the data point is greater than the normalized amplitude threshold, the bioelectric signal data point is set to 1. If the amplitude of the data point is less than or equal to the normalized amplitude threshold, the data point is set to 0. All bioelectric signal data in the first electric signal set are normalized and thresholded to obtain binary multi-bioelectric signals, which form a multi-bioelectric signal set and perform feature extraction or pattern recognition, etc., for further analysis and processing.

[0044] Extracting biometric features based on the multiple bioelectric signals to obtain multiple biometric features;

[0045] Preferably, biometric feature extraction is performed based on multiple bioelectric signals to obtain multiple biometric features, wherein common methods of feature extraction include time domain features, frequency domain features, energy features or spectral features, etc. The extracted bioelectric signal features are screened, combined or reduced in dimension through statistical analysis, feature selection algorithms, principal component analysis and other methods to obtain more representative and effective multiple biometric features for subsequent vector representation.

[0046] Further, such as Figure 2 As shown, the method of extracting biometric features based on the multiple bioelectric signals to obtain multiple biometric features includes:

[0047] Obtaining an electrical signal identifier of a training load factor based on the multiple bioelectric signals;

[0048] Obtaining electrical signal parameters corresponding to the electrical signal identifier;

[0049] The biological detection index is obtained according to the clinical database, and the multi-biological characteristics are obtained by combining the electrical signal identifier and the electrical signal parameter.

[0050] Preferably, the electrical signal identifier of the training load factor is obtained based on multiple bioelectric signals, and then the electrical signal parameters corresponding to the electrical signal identifier are obtained. Then, the biological detection index is obtained based on the clinical database, and multiple biological features are obtained in combination with the electrical signal identifier and the electrical signal parameters. Specifically, obtaining the electrical signal index of the training load factor based on multiple bioelectric signals refers to using the collected multiple bioelectric signals to identify and judge the electrical signal characteristic index of the training load factor, wherein the training load factor may include factors that affect the biological system, such as exercise load, cognitive load, psychological pressure, etc. The electrical signal identifier is an electrical signal identifier representing the characteristics of the training load factor extracted by processing the bioelectric signal, reflecting the electrical signal changes of the biological system under different training load factors, such as electrocardiogram data such as heart rate, rhythm, arrhythmia, and electrocardiogram morphological changes. By analyzing the electrocardiogram, parameters such as heart rate and heart rate variability can be extracted as electrical signal identifiers to judge the response of the cardiovascular system to the training load. The electrical signal parameters are numerical quantities obtained by signal processing and feature extraction based on the collected bioelectric signal data, representing certain characteristics or changes of the electrical signal, reflecting the biological system under different training load factors. Electrical signal characteristics under training load, for example, for electrocardiogram signals, electrical signal parameters may include heart rate, which reflects the frequency of heart contractions, RR interval, which is the time interval between two adjacent electrocardiogram R waves, used to analyze cardiac activity characteristics such as heart rate variability, and P wave and T wave amplitudes, which are used to analyze the changing patterns of cardiac activity. A clinical database refers to a database in which medical institutions collect and store clinical data, which may include various types of data such as clinical cases, medical images, and health examination records. When obtaining biometric indicators, reliable clinical databases should be sought, such as medical institution case databases, health databases of national or international public health institutions, etc., to ensure that the source of clinical data is credible. Then, the multiple biometric characteristics are obtained by combining electrical signal identifiers and electrical signal parameters. For example, the electrical signal identifier of the training load factor is obtained using the electrocardiogram and the heart rate and heart rate variability are extracted as electrical signal parameters. Combining these two parameter indicators can obtain multiple biometric indicator characteristics, including cardiovascular health characteristics reflecting good cardiovascular health status, autonomic nervous system regulation characteristics reflecting the regulatory ability of the autonomic nervous system, and exercise load adaptation characteristics evaluating the adaptability of the cardiovascular system of an individual organism under different exercise loads.

[0051] Establishing a multi-head feature extractor to extract signal features from the multiple bioelectric signals to obtain multiple signal features;

[0052] Preferably, a multi-head feature extractor is established to perform signal feature extraction on multiple bioelectric signals and obtain multi-signal features. The multi-head feature extractor can process multiple bioelectric signals at the same time and perform feature extraction to extract multi-signal features therefrom. Multi-signal features refer to signal features with information content extracted from multiple bioelectric signals, providing more comprehensive electrical signal data of the biological system.

[0053] Furthermore, a multi-head feature extractor was established, which previously included:

[0054] Establishing the multi-head feature extractor, wherein each head corresponds to a bioelectric signal;

[0055] The multi-head feature extractor is trained on multiple bioelectric signal records to complete the establishment of the multi-head feature extractor.

[0056] Preferably, a multi-head feature extractor is established, and the bioelectric signal records are trained according to the multi-head feature extractor to complete the establishment of the multi-head feature extractor. Specifically, a model architecture of the multi-head feature extractor is established, and multiple bioelectric signal recording data are used as training data, which are input into the initial model for training. Multiple different bioelectric signals are processed and features are extracted to complete the establishment of the multi-head feature extractor, wherein each head corresponds to the extraction of a bioelectric signal.

[0057] Furthermore, signal feature extraction is performed on the multiple bioelectric signals to obtain multiple signal features, and the method includes:

[0058] Calculating the mean and variance of the input data of each head in the multi-head feature extractor, and inputting the data into the multi-head feature extractor;

[0059] Extracting negative values ​​from the output data of the multi-head feature extractor and mapping them to linear units for correction;

[0060] The output data is mapped to a pooling unit for feature reduction analysis to obtain the multi-signal features.

[0061] Preferably, the mean and variance of the input data of each head in the multi-head feature extractor are calculated and input into the multi-head feature extractor, and then the negative values ​​in the output data of the multi-head feature extractor are extracted and mapped to the linear unit for correction, and then the output data is mapped to the pooling unit for feature reduction analysis. Specifically, calculating the mean and variance of the input data of each head is a common preprocessing method to ensure that the input data of different bioelectric signals have similar scales and distributions, so as to improve the stability of the multi-head feature extractor model. The bioelectric signal data after the mean and variance calculation is used as input and input into the multi-head feature extractor, and each head is responsible for processing and extracting the corresponding The characteristics of the signal are obtained from the multi-head feature extractor, and the output data corresponding to each head is obtained, which may contain positive and negative values. The negative values ​​in the output data are then mapped to the linear unit for correction, and the negative values ​​are set to 0 to ensure the non-negativity of the output data, avoiding the adverse effects caused by the existence of negative values. The output data is mapped to the pooling unit for feature reduction analysis to obtain a simpler representation of multi-signal features, thereby reducing the dimension of the bioelectric signal data and extracting and retaining important biological feature information. Among them, the pooling unit is used to reduce the dimension of the data and extract the main features in the convolutional neural network, and enhance the translation invariance of the model, such as the maximum pooling operation and the average pooling operation, and finally obtain multi-signal features.

[0062] fusing the multiple signal features with the multiple biometric features to generate a fusion vector;

[0063] Specifically, when fusing multi-signal features with multi-biometric features, the feature vectors corresponding to the multi-signal features and multi-biometric features are extracted respectively to ensure that the multi-signal feature vectors and the multi-biometric feature vectors have the same dimension. For example, the feature vectors are adjusted to the same dimension through dimensionality reduction methods. Commonly used feature fusions include splicing fusion, weighted fusion, and dot product fusion. Splicing fusion is to splice two feature vectors in sequence to obtain a longer feature vector. Weighted fusion is to perform a weighted summation of the different weights used for the two feature vectors to obtain a fused feature vector. Dot product fusion is to perform a dot product operation on the two feature vectors to obtain a scalar value as the fused feature representation, and finally generate a fusion vector of multi-signal features and multi-biometric features.

[0064] Analyzing and calculating features of the fusion vector through a first fully connected layer and a second fully connected layer to obtain fusion features;

[0065] Preferably, the fusion vector is analyzed and feature calculated by the first fully connected layer and the second fully connected layer, including obtaining the fusion feature according to the first fully connected layer function and the second fully connected layer function in the first fully connected layer and the second fully connected layer, inputting the fusion vector into the first fully connected layer for feature calculation, and for each neuron, using weights and biases to perform a linear transformation on the input fusion vector to calculate a new feature representation, and then using the output feature of the first fully connected layer as input data and inputting it into the second fully connected layer for feature calculation, and similarly performing a linear transformation on each neuron using weights and biases to obtain the final fusion feature, wherein the first fully connected layer function and the second fully connected layer function are as follows:

[0066] V1=ReLU(W1z+b1),

[0067] V2 = ReLU(W2V1+b2);

[0068] Among them, W1, b1, represent the parameters of the first fully connected layer, W2, b2 represent the parameters of the second fully connected layer, V1 represents the fusion features of the first fully connected layer, V2 represents the fusion features of the second fully connected layer, and z represents the fusion vector.

[0069] Constructing a risk assessor to perform risk assessment on the fusion features and obtain damage risk;

[0070] Preferably, a risk assessor is constructed to perform risk assessment on the fusion features and obtain the damage risk. The risk assessor is a tool based on machine learning that uses existing multi-bioelectric signal data for modeling and analysis, identifies the association between different features and risks, and performs risk assessment and prediction. Damage risk refers to the possibility of damage or impact in a biological system under specific circumstances. The multiple damage risk calculation functions in the risk assessor are as follows:

[0071] p1=Softmax(ReLU(W3V2+b3)W4+b4),

[0072] p2=Softmax(ReLU(W5V2+b5)W6+b6),

[0073] p3=Softmax(ReLU(W7V2+b7)W8+b8),

[0074] p4=Softmax(ReLU(W9V2+b9)W 10 +b 10 );

[0075] Among them, p1 represents the first damage risk, W3, W4, b3, and b4 represent the parameters of the first damage risk calculation function, p2 represents the second damage risk, W5, W6, b5, and b6 represent the parameters of the second damage risk calculation function, p3 represents the third damage risk, W7, W8, b7, and b8 represent the parameters of the third damage risk calculation function, p4 represents the fourth damage risk, W9, W 10 、b9、b 10 Parameters characterizing the first damage risk calculation function.

[0076] Error analysis is performed based on the damage risk and real-time data of corresponding characteristics, and feedback control is performed based on the error value to obtain damage information.

[0077] Preferably, error analysis is performed based on the real-time data of the injury risk and the corresponding characteristics, and feedback control is performed based on the error value to obtain the injury risk, wherein the real-time data of the corresponding characteristics refers to the real-time multi-bioelectric signal data extracted by monitoring. Error analysis is performed on the injury risk and the corresponding real-time data, for example, the error value is calculated by comparison using absolute error, root mean square error, etc., and then feedback control is continuously performed based on the error value to optimize the risk assessment model and obtain more accurate injury risk information.

[0078] In summary, the embodiments of the present application have at least the following technical effects:

[0079] The embodiment of the present application extracts multiple biological labels, obtains the original multiple bioelectric signal sets corresponding to the multiple biological labels, performs noise reduction processing on the original multiple bioelectric signal sets, obtains multiple bioelectric signals, extracts biometric features based on the multiple bioelectric signals, obtains multiple biometric features, establishes a multi-head feature extractor, extracts signal features from the multiple bioelectric signals, obtains multiple signal features, fuses the multiple signal features with the multiple biometric features, generates a fusion vector, analyzes and calculates features of the fusion vector through the first fully connected layer and the second fully connected layer, obtains fusion features, constructs a risk assessor, performs risk assessment on the fusion features, obtains injury risk, performs error analysis based on the injury risk and real-time data of the corresponding features, performs feedback control based on the error value to obtain injury information, solves the technical problem of inaccurate acquisition of multiple bioelectric signals and feature extraction in the existing multi-bioelectric signal fusion risk assessment, resulting in poor fusion effect and inaccurate and unreliable risk assessment results, thereby achieving a technical effect of more accurate and reliable risk assessment results.

[0080] Example 2

[0081] Based on the same inventive concept as the risk assessment method based on multi-bioelectric signal fusion in the aforementioned embodiment, Figure 3As shown, the present application provides a risk assessment system based on multi-bioelectric signal fusion. The system and method embodiments in the present application are based on the same inventive concept, wherein the system includes:

[0082] An original multi-bioelectric signal set acquisition module 10, which is used to extract multiple biomarkers and obtain the original multi-bioelectric signal sets corresponding to the multiple biomarkers;

[0083] a multi-bioelectric signal acquisition module 20, configured to perform noise reduction processing on the original multi-bioelectric signal set to obtain multi-bioelectric signals;

[0084] A multi-biometric feature acquisition module 30, configured to extract biometric features based on the multiple bioelectric signals to obtain multiple biometric features;

[0085] A multi-signal feature acquisition module 40 is used to establish a multi-head feature extractor to extract signal features from the multiple bioelectric signals to obtain multi-signal features;

[0086] a fusion vector obtaining module 50, configured to fuse the multiple signal features with the multiple biometric features to generate a fusion vector;

[0087] A fusion feature acquisition module 60 is configured to analyze and calculate features of the fusion vector through a first fully connected layer and a second fully connected layer to obtain a fusion feature;

[0088] An injury risk obtaining module 70 is used to construct a risk assessor, perform risk assessment on the fusion features, and obtain injury risk;

[0089] The damage information feedback control module 80 is used to perform error analysis based on the damage risk and real-time data of corresponding characteristics, and perform feedback control based on the error value to obtain damage information.

[0090] Furthermore, the multi-bioelectric signal acquisition module 20 is further configured to perform the following method:

[0091] Reducing the signal-to-noise ratio of the original multiple bioelectrical signal sets through a filter to obtain a first electrical signal set;

[0092] The first electrical signal set is normalized according to a normalization amplitude threshold to obtain the multiple bioelectrical signals.

[0093] Furthermore, the multi-biometric feature acquisition module 30 is further configured to perform the following method:

[0094] Obtaining an electrical signal identifier of a training load factor based on the multiple bioelectric signals;

[0095] Obtaining electrical signal parameters corresponding to the electrical signal identifier;

[0096] The biological detection index is obtained according to the clinical database, and the multi-biological characteristics are obtained by combining the electrical signal identifier and the electrical signal parameter.

[0097] Furthermore, the multi-signal feature acquisition module 40 is further configured to perform the following method:

[0098] Establishing the multi-head feature extractor, wherein each head corresponds to a bioelectric signal;

[0099] The multi-head feature extractor is trained on multiple bioelectric signal records to complete the establishment of the multi-head feature extractor.

[0100] Furthermore, the multi-signal feature acquisition module 40 is further configured to perform the following method:

[0101] Calculating the mean and variance of the input data of each head in the multi-head feature extractor, and inputting the data into the multi-head feature extractor;

[0102] Extracting negative values ​​from the output data of the multi-head feature extractor and mapping them to linear units for correction;

[0103] The output data is mapped to a pooling unit for feature reduction analysis to obtain the multi-signal features.

[0104] Furthermore, the fusion feature acquisition module 60 is further configured to perform the following method:

[0105] The fusion feature is obtained according to the first fully connected layer function and the second fully connected layer function in the first fully connected layer and the second fully connected layer, wherein the first fully connected layer function and the second fully connected layer function are shown as follows:

[0106] V1=ReLU(W1z+b1),

[0107] V2 = ReLU(W2V1+b2);

[0108] Among them, W1, b1, represent the parameters of the first fully connected layer, W2, b2 represent the parameters of the second fully connected layer, V1 represents the fusion features of the first fully connected layer, V2 represents the fusion features of the second fully connected layer, and z represents the fusion vector.

[0109] Furthermore, the damage risk obtaining module 70 is further configured to execute the following method:

[0110] The multiple damage risk calculation functions in the risk assessor are shown below:

[0111] p1=Softmax(ReLU(W3V2+b3)W4+b4),

[0112] p2=Softmax(ReLU(W5V2+b5)W6+b6),

[0113] p3=Softmax(ReLU(W7V2+b7)W8+b8),

[0114] p4=Softmax(ReLU(W9V2+b9)W 10 +b 10 );

[0115] Among them, p1 represents the first damage risk, W3, W4, b3, and b4 represent the parameters of the first damage risk calculation function, p2 represents the second damage risk, W5, W6, b5, and b6 represent the parameters of the second damage risk calculation function, p3 represents the third damage risk, W7, W8, b7, and b8 represent the parameters of the third damage risk calculation function, p4 represents the fourth damage risk, W9, W 10 、b9、b 10 Parameters characterizing the first damage risk calculation function.

[0116] It should be noted that the above-mentioned order of the embodiments of the present application is for descriptive purposes only and does not represent the superiority or inferiority of the embodiments. The above description is of specific embodiments of this specification. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps described in the claims can be performed in an order different from that in the embodiments and still achieve the desired results. In addition, the processes depicted in the accompanying drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0117] The above description is only a preferred embodiment of the present application and is not intended to limit the present application. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present application should be included in the scope of protection of the present application.

[0118] This specification and drawings are merely illustrative of the present application and are intended to cover any and all modifications, variations, combinations, or equivalents within the scope of this application. Obviously, those skilled in the art may make various modifications and variations to this application without departing from the scope of this application. Thus, this application is intended to include such modifications and variations as fall within the scope of this application and its equivalents.

Claims

1. A risk assessment method based on multi-bioelectric signal fusion, characterized in that: The method comprises: Extracting multiple biological tags and obtaining original multiple bioelectrical signal sets corresponding to the multiple biological tags; Performing noise reduction processing on the original multiple bioelectrical signal set to obtain multiple bioelectrical signals; Extracting biometric features based on the multiple bioelectric signals to obtain multiple biometric features; Establishing a multi-head feature extractor to extract signal features from the multiple bioelectric signals to obtain multiple signal features; fusing the multiple signal features with the multiple biometric features to generate a fusion vector; Analyzing and calculating features of the fusion vector through a first fully connected layer and a second fully connected layer to obtain fusion features; Constructing a risk assessor to perform risk assessment on the fusion features and obtain damage risk; Perform error analysis based on the real-time data of the damage risk and corresponding characteristics, and perform feedback control based on the error value to obtain damage information; Performing noise reduction processing on the original multiple bioelectrical signal set to obtain multiple bioelectrical signals includes: Reducing the signal-to-noise ratio of the original multiple bioelectrical signal sets through a filter to obtain a first electrical signal set; Normalizing the first electrical signal set according to a normalized amplitude threshold to obtain the multiple bioelectrical signals; Extracting biometric features based on the multiple bioelectric signals to obtain multiple biometric features includes: Obtaining an electrical signal identifier of a training load factor based on the multiple bioelectric signals; Obtaining electrical signal parameters corresponding to the electrical signal identifier; Acquire biological detection indicators according to a clinical database, and obtain the multi-biological features by combining the electrical signal identifier and the electrical signal parameter; Building a multi-head feature extractor, previously included: Establishing the multi-head feature extractor, wherein each head corresponds to a bioelectric signal; Training multiple bioelectric signal records according to the multi-head feature extractor to complete the establishment of the multi-head feature extractor; Extracting signal features from the multiple bioelectric signals to obtain multiple signal features includes: Calculating the mean and variance of the input data of each head in the multi-head feature extractor, and inputting the data into the multi-head feature extractor; Extracting negative values ​​from the output data of the multi-head feature extractor and mapping them to linear units for correction; Mapping the output data to a pooling unit for feature reduction analysis to obtain the multi-signal features; Analyzing and calculating features of the fusion vector through the first fully connected layer and the second fully connected layer to obtain fusion features, including: The fusion feature is obtained according to the first fully connected layer function and the second fully connected layer function in the first fully connected layer and the second fully connected layer, wherein the first fully connected layer function and the second fully connected layer function are shown as follows: V1=ReLU(W1z+b1), V2 = ReLU(W2V1+b2); Among them, W1, b1, represent the parameters of the first fully connected layer, W2, b2 represent the parameters of the second fully connected layer, V1 represents the fusion features of the first fully connected layer, V2 represents the fusion features of the second fully connected layer, and z represents the fusion vector.

2. The method according to claim 1, wherein Construct a risk assessor to perform risk assessment on the fusion features to obtain the damage risk, including: The multiple damage risk calculation functions in the risk assessor are shown below: p1=Softmax(ReLU(W3V2+b3)W4+b4), p2=Softmax(ReLU(W5V2+b5)W6+b6), p3=Softmax(ReLU(W7V2+b7)W8+b8), p4=Softmax(ReLU(W9V2+b9)W 10 +b 10 ): Among them, p1 represents the first damage risk, W3, W4, b3, and b4 represent the parameters of the first damage risk calculation function, p2 represents the second damage risk, W5, W6, b5, and b6 represent the parameters of the second damage risk calculation function, p3 represents the third damage risk, W7, W8, b7, and b8 represent the parameters of the third damage risk calculation function, p4 represents the fourth damage risk, W9, W 10 、b9、b 10 Parameters characterizing the first damage risk calculation function.

3. A risk assessment system based on multi-bioelectric signal fusion, characterized in that: The system comprises: an original multi-bioelectric signal set acquisition module, the original multi-bioelectric signal set acquisition module being used to extract multi-biological tags and obtain the original multi-bioelectric signal set corresponding to the multi-biological tags; a multi-bioelectric signal acquisition module, configured to perform noise reduction processing on the original multi-bioelectric signal set to acquire multi-bioelectric signals; a multi-biometric feature acquisition module, configured to extract biometric features based on the multiple bioelectric signals to obtain multiple biometric features; A multi-signal feature acquisition module, wherein the multi-signal feature acquisition module is used to establish a multi-head feature extractor, perform signal feature extraction on the multiple bioelectric signals, and obtain multi-signal features; a fusion vector acquisition module, configured to fuse the multiple signal features with the multiple biometric features to generate a fusion vector; A fusion feature acquisition module, configured to analyze and calculate features of the fusion vector through a first fully connected layer and a second fully connected layer to obtain a fusion feature; An injury risk acquisition module, wherein the injury risk acquisition module is used to construct a risk assessor, perform risk assessment on the fusion features, and obtain injury risk; A damage information feedback control module is used to perform error analysis based on the damage risk and real-time data of corresponding characteristics, and to perform feedback control based on the error value to obtain damage information; The multi-bioelectric signal acquisition module is also used to perform the following method: Reducing the signal-to-noise ratio of the original multiple bioelectrical signal sets through a filter to obtain a first electrical signal set; Normalizing the first electrical signal set according to a normalized amplitude threshold to obtain the multiple bioelectrical signals; Furthermore, the multi-biometric feature acquisition module is further configured to perform the following method: Obtaining an electrical signal identifier of a training load factor based on the multiple bioelectric signals; Obtaining electrical signal parameters corresponding to the electrical signal identifier; Acquire biological detection indicators according to a clinical database, and obtain the multi-biological features by combining the electrical signal identifier and the electrical signal parameter; Furthermore, the multi-signal feature acquisition module is further configured to perform the following method: Establishing the multi-head feature extractor, wherein each head corresponds to a bioelectric signal; Training multiple bioelectric signal records according to the multi-head feature extractor to complete the establishment of the multi-head feature extractor; Furthermore, the multi-signal feature acquisition module is further configured to perform the following method: Calculating the mean and variance of the input data of each head in the multi-head feature extractor, and inputting the data into the multi-head feature extractor; Extracting negative values ​​from the output data of the multi-head feature extractor and mapping them to linear units for correction; Mapping the output data to a pooling unit for feature reduction analysis to obtain the multi-signal features; Furthermore, the fusion feature acquisition module is further configured to perform the following method: The fusion feature is obtained according to the first fully connected layer function and the second fully connected layer function in the first fully connected layer and the second fully connected layer, wherein the first fully connected layer function and the second fully connected layer function are shown as follows: V1=ReLU(W1z+b1), V2 = ReLU(W2V1+b2); Among them, W1, b1, represent the parameters of the first fully connected layer, W2, b2 represent the parameters of the second fully connected layer, V1 represents the fusion features of the first fully connected layer, V2 represents the fusion features of the second fully connected layer, and z represents the fusion vector.

Images