Signal monitoring system based on human vital signs feature extraction

Abstract

The present invention discloses a signal monitoring system based on human vital sign feature extraction, including a vital sign data acquisition module, a preprocessing module, a vital sign monitoring model building module and a vital sign monitoring module. The present invention belongs to the field of data processing technology, and specifically refers to a signal monitoring system based on human vital sign feature extraction. This scheme introduces a feature response function to design an adaptive loss function, introduces an adaptive intra-class distance and a dynamic inter-class distance to improve the ability to distinguish between mild abnormalities and severe abnormalities, aggregates similar samples and distinguishes heterogeneous samples, thereby improving the monitoring accuracy of the vital sign status of signal monitoring; enhances the initial population diversity through multi-layer mapping, and performs parameter search based on hierarchical division, so that the constructed vital sign monitoring model converges quickly, thereby improving the accuracy of vital sign monitoring and achieving more stable monitoring results.

Description

Technical Field

[0001] The present invention relates to the technical field of data processing, and in particular to a signal monitoring system based on human vital sign feature extraction. Background Art

[0002] The vital signs monitoring system is a system used to collect, process and analyze key physiological parameters of the human body in real time. It evaluates the health status of the human body by collecting and analyzing the vital signs data of the human body. However, the general vital signs monitoring system has problems such as class imbalance and difficulty in classifying abnormal samples, resulting in low accuracy of vital signs status monitoring and poor distinction between different categories; the general vital signs monitoring system has problems such as insufficient system setting dynamics and poor flexibility in the adjustment process, which leads to poor monitoring effect of the final vital signs status. Summary of the invention

[0003] In view of the above situation, in order to overcome the defects of the prior art, the present invention provides a signal monitoring system based on human vital signs feature extraction. In view of the problems that general vital signs monitoring systems have class imbalance and difficulty in classifying abnormal samples, resulting in low accuracy in vital signs status monitoring and poor distinction between different categories, this scheme introduces a feature response function to design an adaptive loss function, introduces adaptive intra-class distance and dynamic inter-class distance to improve the ability to distinguish between mild abnormalities and severe abnormalities, aggregates samples of the same class and distinguishes samples of different classes, thereby improving the monitoring accuracy of the vital signs status of signal monitoring; in view of the problems that general vital signs monitoring systems have insufficient system setting dynamics and poor flexibility in the adjustment process, thereby resulting in poor final vital signs status monitoring effect, this scheme enhances the initial population diversity through multi-layer mapping, and performs parameter search based on hierarchical division, so that the constructed vital signs monitoring model converges quickly, thereby improving the accuracy of vital signs monitoring and achieving more stable monitoring results.

[0004] The technical solution adopted by the present invention is as follows: the signal monitoring system based on human vital signs feature extraction provided by the present invention comprises a vital signs data acquisition module, a preprocessing module, a vital signs monitoring model establishment module and a vital signs monitoring module;

[0005] The vital signs data collection module collects historical signal monitoring data;

[0006] The preprocessing module performs denoising, data conversion and data set division on the collected signal monitoring data;

[0007] The vital signs monitoring model establishment module designs an adaptive loss function by introducing a feature response function, introduces an adaptive intra-class distance and a dynamic inter-class distance, constructs a final loss function, initializes the population through multi-layer mapping, performs parameter search based on hierarchical division, and then completes parameter adjustment, thereby finally realizing the establishment of a vital signs monitoring model;

[0008] The vital sign monitoring module implements status monitoring of real-time vital sign data based on the established vital sign monitoring model.

[0009] Furthermore, in the vital signs data acquisition module, the historical signal monitoring data includes heart rate, respiratory rate, blood oxygen saturation, body temperature, blood pressure, heart rate variability, movement status, time and state monitoring results; the state monitoring results include normal, mild abnormality and severe abnormality; the state monitoring results are used as data labels.

[0010] Furthermore, in the preprocessing module, the data conversion is to convert the signal monitoring data into a vector form and perform normalization processing to obtain a time series data set; the data set division is to divide the time series data set into a test set and a training set.

[0011] Furthermore, the vital signs monitoring model building module includes a system framework overview unit, an adaptive loss function design unit, a model determination unit and an initial parameter adjustment unit, which specifically include the following contents:

[0012] System framework overview unit; The established vital signs monitoring model adopts a stacked multi-layer LSTM network. The first layer of LSTM is used to capture the preliminary time series features, and the second layer of LSTM is used to further abstract the time-dependent features. Finally, the time series features output by the LSTM are mapped to the classification results through the fully connected layer. The Softmax layer is used to output the classification results of the vital signs status, which are classified into normal, mildly abnormal and severely abnormal.

[0013] Adaptive loss function design unit; specifically includes:

[0014] Define the classification loss; the formula used is as follows:

[0015] ;

[0016] ;

[0017] ;

[0018] Where LX is the classification loss; N is the total number of samples; C is the total number of health status categories; s is the temperature scaling factor; is the characteristic function of the category corresponding to the i-th sample, that is, the characteristic representation obtained after the signal monitoring data is input into the model; α is the balance factor; is the weight of the category corresponding to the i-th sample; is the predicted probability distribution of the category corresponding to the i-th sample; is the feature response function, which represents the activation value of the model on the category; a is the balance coefficient, h is the hidden state of the model output; k is a constant term used to adjust the output range of the model; are all the trainable parameters of the model; is the characteristic parameter vector of the jth category;

[0019] Define the metric loss; introduce the adaptive intra-class distance ICD. The smaller the adaptive intra-class distance, the closer the clustering of similar samples is, expressed as: ; Dynamic inter-class distance CD is introduced. The larger the dynamic inter-class distance, the higher the distinction between different categories, which can be expressed as: ; Among them, z i is the feature vector of the i-th sample; is the mean vector of the category corresponding to the i-th sample; is the standard deviation of the category corresponding to the i-th sample; is the regulating factor; and are the mean vectors of the j1th and jth categories respectively; and are the inter-class covariance matrices of the j1th and jth categories respectively; is the Kullback-Leibler divergence between the distributions of the j1th category and the jth category j; β and γ are adjustment weight coefficients; Tr(·) is the trace of the matrix;

[0020] Design of total loss function: The total loss function during training is composed of classification loss and metric loss. The total loss function is defined as:

[0021] ;

[0022] Where Ltt is the total loss function; and is the loss weight coefficient;

[0023] Model judgment unit; when the loss of the vital sign monitoring model for the training set converges, the training of the vital sign monitoring model is completed; a prediction threshold is set in advance, and when the prediction accuracy of the trained vital sign monitoring model for the test set is higher than the prediction threshold, the establishment of the vital sign monitoring model is completed; otherwise, it goes to the initial parameter adjustment unit;

[0024] Initial parameter adjustment unit; specifically including:

[0025] Constructing parameter adjustment space; constructing parameter adjustment space based on temperature scaling coefficient, balance factor, characteristic parameter vector, adjustment factor, adjustment weight coefficient, loss weight coefficient, balance coefficient and constant term;

[0026] Initialize the parameter search population; Initialize the parameter search population individual position, the formula used is as follows:

[0027] ;

[0028] ;

[0029] In the formula, z i1+1,j2 and z i1,j2 are the mapping coefficients of the j2th dimension of the i1+1th individual and the i1th individual respectively; α1 is the mapping parameter; is the position of the j2th dimension of the i1th individual; 0 means the 0th search, i.e. initialization; and are the upper and lower limits of the parameter adjustment space of the j2th dimension respectively; is the mapping coefficient; rr is a random number between 0 and 1;

[0030] Hierarchical division; the prediction accuracy of the vital sign monitoring model based on individual position training is used as the individual fitness value; the hierarchical division is dynamically adjusted according to the ranking of the individual fitness value, and the formula used is as follows:

[0031] ;

[0032] In the formula, ll i1 is the individual level; t is the current search number; T is the maximum search number; bf and wf are the maximum and minimum fitness values ​​of the population respectively; f i1 is the individual fitness value; is rounded down;

[0033] Parameter search; parameter search selection is performed based on the individual level; the formula used is as follows:

[0034] ;

[0035] ;

[0036] ;

[0037] Where r, r1, r2, r3, r4 and r5 are all random numbers between 0 and 1 and are independent of each other; is the individual position with the highest fitness value; is the adjustment parameter; is the position when the individual's historical fitness value is the highest; F is the search direction mark; is the acquisition intensity; α2 is the control scaling factor; is the distance between the individual and the optimal individual;

[0038] Adjustment judgment; when there is an individual fitness value higher than the prediction threshold, the adjustment ends, and the individual position is the adjusted parameter setting of the vital sign monitoring model; if the maximum number of iterations is reached, it goes to the initialization parameter search population; otherwise it goes to the hierarchical division.

[0039] Furthermore, the vital signs monitoring module is based on the established vital signs monitoring model, and collects heart rate, respiratory rate, blood oxygen saturation, body temperature, blood pressure, heart rate variability, movement status and time data of vital signs signal monitoring in real time, and inputs them into the vital signs monitoring model after preprocessing, and outputs the status monitoring results output by the model as monitoring results. When the output status monitoring results are seriously abnormal, an early warning is issued to relevant personnel.

[0040] The beneficial effects achieved by the present invention using the above scheme are as follows:

[0041] (1) Aiming at the problems of class imbalance and difficulty in classifying abnormal samples in general vital signs monitoring systems, which lead to low accuracy in vital signs status monitoring and poor distinction between different categories, this scheme introduces a feature response function to design an adaptive loss function, and introduces adaptive intra-class distance and dynamic inter-class distance to improve the ability to distinguish between mild and severe abnormalities, so that samples of the same type can be aggregated and samples of different types can be distinguished, thereby improving the accuracy of monitoring the vital signs status of signal monitoring.

[0042] (2) In view of the problem that general vital signs monitoring systems have insufficient system setting dynamics and poor adjustment process flexibility, which leads to poor monitoring effect of the final vital signs status, this scheme enhances the initial population diversity through multi-layer mapping and performs parameter search based on hierarchical division, so that the constructed vital signs monitoring model converges quickly, thereby improving the accuracy of vital signs monitoring and achieving more stable monitoring results. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] Figure 1 A schematic diagram of the flow chart of the signal monitoring system based on human vital signs feature extraction provided by the present invention;

[0044] Figure 2 Schematic diagram of the process of building modules for the vital signs monitoring model;

[0045] Figure 3 Schematic diagram of the process of initial parameter adjustment unit.

[0046] The accompanying drawings are used to provide further understanding of the present invention and constitute a part of the specification. They are used to explain the present invention together with the embodiments of the present invention and do not constitute a limitation of the present invention. DETAILED DESCRIPTION

[0047] 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 invention. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments; based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0048] In the description of the present invention, it should be understood that terms such as “upper”, “lower”, “front”, “back”, “left”, “right”, “top”, “bottom”, “inside” and “outside” indicating directions or positional relationships are based on the directions or positional relationships shown in the accompanying drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific direction, be constructed and operated in a specific direction, and therefore should not be understood as limiting the present invention.

[0049] Example 1, see Figure 1 The signal monitoring system based on human vital signs feature extraction provided by the present invention comprises a vital signs data acquisition module, a preprocessing module, a vital signs monitoring model establishment module and a vital signs monitoring module;

[0050] The vital sign data acquisition module acquires historical signal monitoring data; and sends the data to the preprocessing module;

[0051] The preprocessing module receives the data sent by the vital signs data acquisition module; performs denoising, data conversion and data set division on the collected signal monitoring data; and sends the data to the vital signs monitoring model establishment module;

[0052] The vital signs monitoring model establishment module receives the data sent by the preprocessing module; designs an adaptive loss function by introducing a feature response function, introduces an adaptive intra-class distance and a dynamic inter-class distance, constructs a final loss function, initializes the population through multi-layer mapping, performs parameter search based on hierarchical division, and then completes parameter adjustment, and finally realizes the establishment of a vital signs monitoring model; and sends the data to the vital signs monitoring module;

[0053] The vital sign monitoring module receives data sent by the vital sign monitoring model building module; and implements status monitoring of real-time vital sign data based on the built vital sign monitoring model.

[0054] Example 2, see Figure 1This embodiment is based on the above embodiment. In the vital sign data acquisition module, the historical signal monitoring data includes heart rate, respiratory rate, blood oxygen saturation, body temperature, blood pressure, heart rate variability, motion state, time and state monitoring results; the state monitoring results include normal, mild abnormality and severe abnormality; the state monitoring results are used as data labels; the heart rate, respiratory rate and motion state are collected by millimeter wave radar.

[0055] Example 3, see Figure 1 This embodiment is based on the above embodiment. In the preprocessing module, data conversion is to convert the signal monitoring data into a vector form and perform normalization processing to obtain a time series data set; data set division is to divide the time series data set into a test set and a training set.

[0056] Example 4, see Figure 1 and Figure 2 This embodiment is based on the above embodiment. The vital sign monitoring model establishment module includes a system framework overview unit, an adaptive loss function design unit, a model determination unit and an initial parameter adjustment unit, and specifically includes the following contents:

[0057] System framework overview unit; The established vital signs monitoring model adopts a stacked multi-layer LSTM network. The first layer of LSTM is used to capture the preliminary time series features, and the second layer of LSTM is used to further abstract the time-dependent features. Finally, the time series features output by the LSTM are mapped to the classification results through the fully connected layer. The Softmax layer is used to output the classification results of the vital signs status, which are classified into normal, mildly abnormal and severely abnormal.

[0058] Adaptive loss function design unit; specifically includes:

[0059] Define the classification loss; the formula used is as follows:

[0060] ;

[0061] ;

[0062] ;

[0063] Where LX is the classification loss; N is the total number of samples; C is the total number of health status categories; s is the temperature scaling factor; is the characteristic function of the category corresponding to the i-th sample, that is, the characteristic representation obtained after the signal monitoring data is input into the model; α is the balance factor; is the weight of the category corresponding to the i-th sample; is the predicted probability distribution of the category corresponding to the i-th sample; is the feature response function, which represents the activation value of the model on the category; a is the balance coefficient, h is the hidden state of the model output; k is a constant term used to adjust the output range of the model; are all the trainable parameters of the model; is the characteristic parameter vector of the jth category;

[0064] Define the metric loss; introduce the adaptive intra-class distance ICD. The smaller the adaptive intra-class distance, the closer the clustering of similar samples is, expressed as: ; Dynamic inter-class distance CD is introduced. The larger the dynamic inter-class distance, the higher the distinction between different categories, which can be expressed as: ; Among them, z i is the feature vector of the i-th sample; is the mean vector of the category corresponding to the i-th sample; is the standard deviation of the category corresponding to the i-th sample; is the regulating factor; and are the mean vectors of the j1th and jth categories respectively; and are the inter-class covariance matrices of the j1th and jth categories respectively; is the Kullback-Leibler divergence between the distributions of the j1th category and the jth category j; β and γ are adjustment weight coefficients; Tr(·) is the trace of the matrix;

[0065] Design of total loss function: The total loss function during training is composed of classification loss and metric loss. The total loss function is defined as:

[0066] ;

[0067] Where Ltt is the total loss function; and is the loss weight coefficient;

[0068] Model judgment unit; when the loss of the vital sign monitoring model for the training set converges, the training of the vital sign monitoring model is completed; a prediction threshold is set in advance, and when the prediction accuracy of the trained vital sign monitoring model for the test set is higher than the prediction threshold, the establishment of the vital sign monitoring model is completed; otherwise, it goes to the initial parameter adjustment unit;

[0069] Initial parameter adjustment unit.

[0070] By performing the above operations, in order to address the problems of low accuracy in vital sign status monitoring and poor distinction between different categories caused by class imbalance and difficulty in classifying abnormal samples in general vital sign monitoring systems, this scheme introduces a feature response function to design an adaptive loss function, and introduces adaptive intra-class distance and dynamic inter-class distance to improve the ability to distinguish between mild and severe abnormalities, so that samples of the same type can be aggregated and samples of different types can be distinguished, thereby improving the accuracy of monitoring the vital sign status of signal monitoring.

[0071] Example 5, see Figure 2 and Figure 3 This embodiment is based on the above embodiment. In the vital sign monitoring model establishment module, the initial parameter adjustment unit specifically includes the following contents:

[0072] Constructing parameter adjustment space; constructing parameter adjustment space based on temperature scaling coefficient, balance factor, characteristic parameter vector, adjustment factor, adjustment weight coefficient, loss weight coefficient, balance coefficient and constant term;

[0073] Initialize the parameter search population; Initialize the parameter search population to ensure that the individual positions are evenly distributed and improve the traversability and diversity of the population. The formula used is as follows:

[0074] ;

[0075] ;

[0076] In the formula, z i1+1,j2 and z i1,j2 are the mapping coefficients of the j2th dimension of the i1+1th individual and the i1th individual respectively; α1 is the mapping parameter; is the position of the j2th dimension of the i1th individual; 0 means the 0th search, i.e. initialization; and are the upper and lower limits of the parameter adjustment space of the j2th dimension respectively; is the mapping coefficient; rr is a random number between 0 and 1;

[0077] Hierarchical division; the prediction accuracy of the vital sign monitoring model based on individual position training is used as the individual fitness value; the hierarchical division is dynamically adjusted according to the ranking of the individual fitness value, and the formula used is as follows:

[0078] ;

[0079] In the formula, ll i1 is the individual level; t is the current search number; T is the maximum search number; bf and wf are the maximum and minimum fitness values ​​of the population respectively; f i1 is the individual fitness value; is rounded down;

[0080] Parameter search; parameter search selection is performed based on the individual level; the formula used is as follows:

[0081] ;

[0082] ;

[0083] ;

[0084] Where r, r1, r2, r3, r4 and r5 are all random numbers between 0 and 1 and are independent of each other; is the individual position with the highest fitness value; is the adjustment parameter; is the position when the individual's historical fitness value is the highest; F is the search direction mark; is the acquisition intensity; α2 is the control scaling factor; is the distance between the individual and the optimal individual;

[0085] Adjustment judgment; when there is an individual fitness value higher than the prediction threshold, the adjustment ends, and the individual position is the adjusted parameter setting of the vital sign monitoring model; if the maximum number of iterations is reached, it goes to the initialization parameter search population; otherwise it goes to the hierarchical division.

[0086] By performing the above operations, in order to solve the problem that general vital signs monitoring systems have insufficient system setting dynamics and poor adjustment process flexibility, which leads to poor final vital signs status monitoring effect, this scheme enhances the initial population diversity through multi-layer mapping and performs parameter search based on hierarchical division, so that the constructed vital signs monitoring model converges quickly, thereby improving the accuracy of vital signs monitoring and achieving more stable monitoring results.

[0087] Example 6, see Figure 1 This embodiment is based on the above embodiment. The vital signs monitoring module is based on the established vital signs monitoring model. It collects heart rate, respiratory rate, blood oxygen saturation, body temperature, blood pressure, heart rate variability, movement status and time data of vital signs signal monitoring in real time, and inputs them into the vital signs monitoring model after preprocessing. The status monitoring result output by the model is output as the monitoring result. When the output status monitoring result is seriously abnormal, an early warning is given to relevant personnel.

[0088] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device.

[0089] While the embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that many changes, modifications, substitutions and variations can be made to the embodiments without departing from the principles and spirit of the invention.

[0090] The present invention and its embodiments are described above, and such description is not restrictive. The drawings show only one embodiment of the present invention, and the actual structure is not limited thereto. In short, if ordinary technicians in the field are inspired by it, without departing from the purpose of the invention, they can design a structure and embodiment similar to the technical solution without creativity, which should belong to the protection scope of the present invention.

Claims

1. A signal monitoring system based on human vital signs feature extraction, characterized by: The system includes a vital sign data acquisition module, a preprocessing module, a vital sign monitoring model building module and a vital sign monitoring module; The vital signs data collection module collects historical signal monitoring data; The preprocessing module performs denoising, data conversion and data set division on the collected signal monitoring data; The vital signs monitoring model establishment module designs an adaptive loss function by introducing a feature response function, introduces an adaptive intra-class distance and a dynamic inter-class distance, constructs a final loss function, initializes the population through multi-layer mapping, performs parameter search based on hierarchical division, and then completes parameter adjustment, thereby finally realizing the establishment of a vital signs monitoring model; The vital signs monitoring module implements status monitoring of real-time vital signs data based on the established vital signs monitoring model; The vital signs monitoring model building module includes defining a metric loss; introducing an adaptive intra-class distance ICD. The smaller the adaptive intra-class distance, the closer the clustering of similar samples is, expressed as: ; Dynamic inter-class distance CD is introduced. The larger the dynamic inter-class distance, the higher the distinction between different categories, which can be expressed as: ; Among them, z i is the feature vector of the i-th sample; is the mean vector of the category corresponding to the i-th sample; is the standard deviation of the category corresponding to the i-th sample; is the regulating factor; and are the mean vectors of the j1th and jth categories respectively; and are the inter-class covariance matrices of the j1th and jth categories respectively; is the Kullback-Leibler divergence between the distributions of the j1th category and the jth category j; β and γ are adjustment weight coefficients; N is the total number of samples; C is the total number of health status categories; is the weight of the class corresponding to the i-th sample; Tr(·) is the trace of the matrix; The vital sign monitoring model establishment module includes a system framework overview unit, an adaptive loss function design unit, a model determination unit and an initial parameter adjustment unit, and specifically includes the following contents: System framework overview unit; The established vital signs monitoring model adopts a stacked multi-layer LSTM network. The first layer of LSTM is used to capture the preliminary time series features, and the second layer of LSTM is used to further abstract the time-dependent features. Finally, the time series features output by the LSTM are mapped to the classification results through the fully connected layer. The Softmax layer is used to output the classification results of the vital signs status, which are classified into normal, mildly abnormal and severely abnormal. Adaptive loss function design unit; specifically includes: Define the classification loss; the formula used is as follows: ； ； ； Where LX is the classification loss; N is the total number of samples; C is the total number of health status categories; s is the temperature scaling factor; is the characteristic function of the category corresponding to the i-th sample, that is, the characteristic representation obtained after the signal monitoring data is input into the model; α is the balance factor; is the weight of the category corresponding to the i-th sample; is the predicted probability distribution of the category corresponding to the i-th sample; is the feature response function, which represents the activation value of the model on the category; a is the balance coefficient, h is the hidden state of the model output; k is a constant term used to adjust the output range of the model; are all the trainable parameters of the model; is the characteristic parameter vector of the jth category; Define metric loss; Design of total loss function: The total loss function during training is composed of classification loss and metric loss. The total loss function is defined as: ； Where Ltt is the total loss function; and is the loss weight coefficient; Model judgment unit; when the loss of the vital sign monitoring model for the training set converges, the training of the vital sign monitoring model is completed; a prediction threshold is set in advance, and when the prediction accuracy of the trained vital sign monitoring model for the test set is higher than the prediction threshold, the establishment of the vital sign monitoring model is completed; otherwise, it goes to the initial parameter adjustment unit; Initial parameter adjustment unit; specifically including: Constructing parameter adjustment space; constructing parameter adjustment space based on temperature scaling coefficient, balance factor, characteristic parameter vector, adjustment factor, adjustment weight coefficient, loss weight coefficient, balance coefficient and constant term; Initialize the parameter search population; Initialize the parameter search population individual position, the formula used is as follows: ； ； In the formula, z i1+1,j2 and z i1,j2 are the mapping coefficients of the j2th dimension of the i1+1th individual and the i1th individual respectively; α1 is the mapping parameter; is the position of the j2th dimension of the i1th individual; 0 means the 0th search, i.e. initialization; and are the upper and lower limits of the parameter adjustment space of the j2th dimension respectively; is the mapping coefficient; rr is a random number between 0 and 1; Hierarchical division; the prediction accuracy of the vital sign monitoring model based on individual position training is used as the individual fitness value; the hierarchical division is dynamically adjusted according to the ranking of the individual fitness value, and the formula used is as follows: ； In the formula, ll i1 is the individual level; t is the current search number; T is the maximum search number; bf and wf are the maximum and minimum fitness values ​​of the population respectively; f i1 is the individual fitness value; is rounded down; Parameter search; parameter search selection is performed based on the individual level; the formula used is as follows: ； ； ； Where r, r1, r2, r3, r4 and r5 are all random numbers between 0 and 1 and are independent of each other; is the individual position with the highest fitness value; is the adjustment parameter; is the position when the individual's historical fitness value is the highest; F is the search direction mark; is the acquisition intensity; α2 is the control scaling factor; is the distance between the individual and the optimal individual; Adjustment judgment; when there is an individual fitness value higher than the prediction threshold, the adjustment ends, and the individual position is the adjusted parameter setting of the vital sign monitoring model; if the maximum number of iterations is reached, it goes to the initialization parameter search population; otherwise it goes to the hierarchical division.

2. The signal monitoring system based on human vital signs feature extraction according to claim 1 is characterized in that: In the vital signs data acquisition module, the historical signal monitoring data includes heart rate, respiratory rate, blood oxygen saturation, body temperature, blood pressure, heart rate variability, movement status, time and state monitoring results; the state monitoring results include normal, mild abnormality and severe abnormality; the state monitoring results are used as data labels.

3. The signal monitoring system based on human vital signs feature extraction according to claim 2 is characterized in that: The vital signs monitoring module is based on the established vital signs monitoring model, and collects heart rate, respiratory rate, blood oxygen saturation, body temperature, blood pressure, heart rate variability, movement status and time data of vital signs signal monitoring in real time. After preprocessing, it is input into the vital signs monitoring model, and the status monitoring result output by the model is output as the monitoring result. When the output status monitoring result is seriously abnormal, an early warning is issued to relevant personnel.

4. The signal monitoring system based on human vital signs feature extraction according to claim 3 is characterized in that: In the preprocessing module, the data conversion is to convert the signal monitoring data into a vector form and perform normalization processing to obtain a time series data set; the data set division is to divide the time series data set into a test set and a training set.

Images