Method for automatically identifying physiological state of human body

By using multimodal signal analysis and change point detection algorithms, the problem of insufficient accuracy in human physiological state recognition in existing technologies has been solved, enabling accurate monitoring of physiological states and recognition of transition processes in complex daily life.

CN118924264APending Publication Date: 2024-11-12GENERAL HOSPITAL OF PLA
View PDF 5 Cites 0 Cited by

Patent Information

Application Number
CN202411315138.0
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-09-20
Publication Date
2024-11-12

AI Technical Summary

Technical Problem

Existing technologies struggle to accurately identify the transitional processes of human physiological states in the complexities of daily life, especially during the transition from high-intensity exercise to rest, resulting in insufficient monitoring accuracy.

Method used

By employing multimodal signals (ECG, respiration, and triaxial acceleration) combined with change point detection algorithms (such as PELT), and through signal preprocessing, feature extraction, and state segmentation, different physiological states of the human body and their transition processes can be identified.

Benefits of technology

It enables accurate identification and real-time monitoring of human physiological states, improving the accuracy and reliability of state recognition.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118924264A_ABST
    Figure CN118924264A_ABST
Patent Text Reader

Abstract

The invention provides a method for automatically recognizing a physiological state, which comprises the following steps of: preprocessing electrocardiogram, respiration and three-axis acceleration data, and extracting statistical characteristics of heart rate, respiration rate and activity amount; based on statistical characteristic segmentation, identifying a change point in the time sequence data, and dividing continuous signal data into a plurality of sub-segments, each sub-segment representing a relatively stable state; and identifying the specific human body physiological state of each sub-section, wherein the specific human body physiological state comprises activity, micro-motion or resting, a transition process from the activity to the resting, a body position under the resting and a sleep state.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present application relates to the field of physiological state monitoring and analysis, and in particular to a method for automatically identifying the physiological state of a human body. Background Art

[0002] With the development of science and technology, biosignal monitoring devices such as wearable devices and smart watches have been widely used in health management and disease monitoring. By collecting and analyzing multimodal signals such as ECG, respiration and triaxial acceleration, the physical state and activity level of an individual can be effectively evaluated. However, due to the diversity and complexity of individuals' activities in daily life, how to accurately identify different states (such as activity, rest, sleep, etc.) plays a vital role in state recognition. In the existing technology, most methods are based only on single modal signals or simple statistical features, which are difficult to deal with complex time series changes and transitions between different states. For example, in the process of subjects switching from high-intensity exercise to stillness, although the amount of activity is significantly reduced, the heart rate and respiratory rate remain at a high level and gradually decrease. If this state transition process is not accurately identified and processed, and it is simply classified as a static state, it may lead to misleading analysis results, thereby affecting the accuracy of the overall monitoring. In addition, how to combine prior knowledge to improve the accuracy and reliability of state recognition is also a problem that needs to be solved urgently.

[0003] Patent application number 201610672525.9 discloses a method and a sports bracelet for automatically detecting the physiological state of the human body, but does not subdivide the amount of activity, body position, and transition state.

[0004] The patent with application number 202110619191.X discloses a physiological state recognition and driving safety warning system based on IPPG technology, but it only targets two abnormal physiological states in a driving scenario, and has not been extended to all daily life states. Summary of the invention

[0005] The present invention aims to provide a method for automatically identifying the physiological state of a human body, which can effectively identify different states of the human body.

[0006] The method of automatically identifying the physiological state of a human body of the present application comprises: Signal preprocessing and feature extraction steps: preprocess the collected ECG, respiration and triaxial acceleration data, extract the statistical features of heart rate, respiration rate and activity, and form corresponding time series data respectively; Change point detection and state division steps: identifying change points in time series data and dividing continuous signal data into several sub-segments; each sub-segment represents a relatively stable state, thereby realizing segmented processing of time series data; State recognition and classification steps: Analyze the sub-segments obtained by segmentation processing to identify the specific human body state.

[0007] Preferably, the human body state includes a sleeping state and a waking state; In the identification of sleep states, the sleep staging algorithm is used to detect the point of falling asleep and divide it into rapid eye movement, light sleep and deep sleep; In the recognition of the awake state, the active state, transition state, micro-motion state, and resting state are identified according to the amount of activity.

[0008] Preferably, PELT is used to implement change point identification in time series data.

[0009] Preferably, the PELT algorithm uses a radial basis function model for detecting mean and variance changes, and the penalty term parameter pen is 2.

[0010] Preferably, the activity is obtained based on the vector modulus of the three-axis acceleration data.

[0011] Preferably, for the awake state, if the average activity in each sub-segment obtained by segmentation processing is higher than 0.35g, it is divided into a high-activity state; if the average activity is between 0.03g and 0.35g, it is divided into a micro-movement state; for a period of time after the end of the high-activity state, it is divided into a transition state; the resting state occurs after the transition state or the micro-movement state and the average activity during the period is lower than 0.03g, where g is the acceleration of gravity.

[0012] Preferably, for the resting state, if the component of gravity in the x-axis direction in the three-axis acceleration data is greater than 0.6g, it is classified as a standing or sitting position; if the absolute value of the component of gravity in the y-axis direction in the three-axis acceleration is greater than 0.5g or the absolute value of the component in the z-axis direction is greater than 0.7g, it is classified as a lying position.

[0013] Preferably, the duration of the transition state T t The duration of high activity state T h Related to, meet T t (T h )=min(T h ,5)+5.

[0014] The method of automatically identifying the physiological state of the human body in the present application, through the joint analysis of multimodal signals, combines the change point detection algorithm with the data-driven model, and accurately identifies each sub-state and its transition process, thereby improving the accuracy and real-time performance of the recognition. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] Figure 1 A classification diagram of the physiological states of the human body; Figure 2 is the PELT segmentation result of the original time series; Figure 3 is the activity volume diagram; Figure 4 This is a comparison chart of the effects before and after applying the method for identifying the physiological state of the human body. DETAILED DESCRIPTION

[0016] 1. First, the collected ECG, respiration and triaxial acceleration data are preprocessed. ECG and respiration signals are bandpass filtered and denoised to remove noise and baseline drift. Triaxial acceleration data is low-pass filtered to eliminate high-frequency noise. Next, statistical features are extracted from the preprocessed signals, including heart rate, respiration rate and activity level. Among them, heart rate and respiration rate are calculated from the peak intervals of ECG and respiration signals, respectively, and activity level is obtained by calculating the square root of the sum of the squares of the triaxial acceleration signals (i.e., vector modulus).

[0017] 2. Use the PELT (Pruned Exact Linear Time) algorithm for change point detection. The PELT algorithm can efficiently detect change points in signal sequences and divide the original time series data into multiple sub-segments. In order to improve the sensitivity of change point detection, the radial basis function model is used to detect changes in the mean and variance, and the penalty parameter pen is set to 2 to control the risk of overfitting. Each sub-segment represents a relatively stable state, providing a basis for subsequent state identification.

[0018] 3. In the state recognition process, the divided sub-segments are analyzed based on the extracted statistical features and prior knowledge.

[0019] a. Division of active state and resting state: For the awake state, the average activity within each subsegment was calculated. If the average activity was above 0.35 g, the subsegment was classified as active.

[0020] After the active state ends, if the activity level remains low but has not yet reached a stable resting state, the subsegment is classified as a transition state. The duration of the transition state is T t Related to the duration of the previous activity state, specifically satisfying T t (T h )=min(T h ,5)+5, where T h Indicates the duration of high activity state.

[0021] When the activity is between 0.03g and 0.35g, the sub-segment is classified as the micromotion state.

[0022] When the amount of activity is stable and below 0.03 g, the subsegment is classified as resting.

[0023] b. Identification of sleep state: Through the sleep detection algorithm, the falling asleep point is detected and sleep states such as rapid eye movement, light sleep and deep sleep are divided.

[0024] 4. In the resting state, the body position is identified based on the direction and magnitude of the gravity component in the three-axis acceleration signal: a. If the component of gravity in the x-axis direction is greater than 0.6g, the subsegment is classified as standing or sitting.

[0025] b. If the absolute value of the component of gravity in the y-axis direction is greater than 0.5g or the absolute value of the component in the z-axis direction is greater than 0.7g, the subsegment is classified as a prone state.

[0026] Through the above-mentioned implementation modes, the present invention can realize automatic analysis of multimodal physiological signals, accurately identify different states and body positions, and provide precise data support and decision-making basis for personalized health management.

[0027] Unless otherwise defined, all technical and / or scientific terms used in this application have the same meaning as commonly understood by one of ordinary skill in the art to which the invention relates. The materials, methods and examples mentioned in this application are illustrative only and not limiting.

[0028] Although the present invention has been described in conjunction with specific embodiments, within the spirit of the invention of this application, those skilled in the art may make appropriate substitutions, modifications and changes, and such substitutions, modifications and changes still fall within the scope of protection of this application.

Claims

1. A method for automatically identifying a physiological state of a human body, comprising: Signal preprocessing and feature extraction steps; Preprocess the collected ECG, respiration and triaxial acceleration data, extract the statistical features of heart rate, respiration rate and activity, and form corresponding time series data respectively; Change point detection and state division steps: identifying change points in the corresponding time series data and dividing the continuous signal data into several sub-segments; each sub-segment represents a relatively stable state, thereby realizing segmented processing of time series data; State recognition and classification steps: Analyze the sub-segments obtained by segmentation processing to identify the specific human body state.

2. The method for automatically identifying the physiological state of a human body according to claim 1, characterized in that: The human body states include sleeping state and waking state; In the identification of sleep states, the sleep staging algorithm is used to detect the point of falling asleep and divide it into rapid eye movement, light sleep and deep sleep; In the recognition of the awake state, the active state, transition state, micro-motion state, and resting state are identified according to the amount of activity.

3. The method for automatically identifying the physiological state of a human body according to claim 1, characterized in that: Using PELT to identify change points in time series data.

4. The method for automatically identifying the physiological state of a human body according to claim 3, characterized in that: The PELT algorithm uses a radial basis function model for detecting mean and variance changes, with a penalty parameter pen of 2.

5. The method for automatically identifying the physiological state of a human body according to claim 1, characterized in that: The activity is obtained based on the vector modulus of the three-axis acceleration data.

6. The method for automatically identifying the physiological state of a human body according to claim 2, characterized in that: For the awake state, if the average activity in each subsegment obtained by segmentation processing is higher than 0.35g, it is divided into a high-activity state; if the average activity is between 0.03g and 0.35g, it is divided into a micro-movement state; for a period of time after the end of the high-activity state, it is divided into a transition state; the resting state occurs after the transition state or the micro-movement state and the average activity during the period is less than 0.03g, where g is the acceleration of gravity.

7. The method for automatically identifying the physiological state of a human body according to claim 2, characterized in that: For the resting state, if the component of gravity in the x-axis direction in the three-axis acceleration data is greater than 0.6g, it is classified as a standing or sitting position. If the absolute value of the component of gravity in the y-axis direction in the three-axis acceleration data is greater than 0.5g or the absolute value of the component in the z-axis direction is greater than 0.7g, it is classified as a lying position.

8. The method for automatically identifying the physiological state of a human body according to claim 2, characterized in that: The duration of the transition state is T t The duration of high activity state T h Related to, meet T t (T h )=min(T h ,5)+5.

Citation Information

Patent Citations

  • Sleep analysis method and device

    CN104706318A

  • Apparatus and program for analyzing living body

    JP2008067892A

  • Smoking estimation device

    JP2013162975A

  • Body movement analysis method and apparatus

    US20110231101A1

  • Stress estimation apparatus, stress estimation method, and computer readable recording medium

    US20220167915A1