CN-121987152-A - Senile dementia risk home monitoring method and system based on multi-modal behavior analysis

CN121987152ACN 121987152 ACN121987152 ACN 121987152ACN-121987152-A

Abstract

The invention belongs to the technical field of health monitoring, and particularly relates to a multi-modal behavior analysis-based senile dementia risk home monitoring method which comprises the following specific steps of monitoring human body gestures in real time, analyzing repeated actions, synchronously early warning abnormal movements, conducting voice interaction operation through a built-in AI dialogue terminal, conducting behavioral, emotion and physiological association modeling, executing cognitive decline dynamic assessment through cognitive function self-test, judging risk grades according to analysis and assessment results, executing a multi-level early warning mechanism based on risk grade information, automatically taking corresponding emergency measures, dynamically calibrating threshold values through integrating user habit data, designing a closed loop feedback mechanism to realize false alarm suppression, and synchronously generating a monthly health report to assist doctors to adjust intervention schemes. The invention can solve the problems of difficult capture of fine behavior change, subjective evaluation hysteresis, passive response-caused rescue delay, low screening frequency and limited scene in the traditional monitoring mode.

Inventors

XU BIAO
ZHAO MINGMING
TIAN GANG
HAN YUBO
WANG LIUYI

Assignees

河南科技大学第一附属医院

Dates

Publication Date: 20260508
Application Date: 20260227

Claims (10)

1. The home monitoring method for senile dementia risk based on multi-modal behavior analysis is characterized by comprising the following specific steps: The human body posture is monitored in real time, repeated actions, orientation disorders and life skill degradation are analyzed, abnormal movements are synchronously early-warned, gait stability and heart rate variability are monitored, blood oxygen saturation and skin electric reaction are continuously collected, and voice interaction operation is carried out through a built-in AI dialogue terminal; After the data is subjected to marginal end primary screening, performing cloud multi-mode fusion analysis, performing behavior, emotion and physiological association modeling, performing cognitive decline dynamic assessment through cognitive function self-measurement, and judging the risk level according to analysis and assessment results; executing a multi-level early warning mechanism based on the risk level information, automatically taking corresponding emergency measures, and assisting the self-test of the cognitive function and the home security monitoring by executing an active intervention mechanism; The threshold value is dynamically calibrated by integrating user habit data, and a closed loop feedback mechanism is designed to realize false alarm suppression, so that a month health report is synchronously generated to assist a doctor in adjusting an intervention scheme.
2. The home monitoring method for senile dementia risk based on multi-modal behavior analysis according to claim 1, wherein the specific steps when the monitoring method is used for daily abnormal behavior monitoring function are as follows: The method comprises the steps of integrating visual sensing, environment assistance and physiological parameters to perform multi-source data acquisition, specifically comprising capturing three-dimensional space actions, tracking human body gestures in real time, judging space orientation abnormality by combining the assistance of an intelligent household device log, and performing cross verification on the wearable device and video stream data; Analyzing an action sequence to detect an abnormal circulation mode, tracking a motion track to identify repeated loitering behaviors, constructing a human-space relation model to judge directional force disorder, analyzing article misuse behaviors to detect life skill degradation; adopting a multi-stage early warning mechanism to perform risk assessment, judging low, medium and high risk levels and taking emergency measures; and generating a monthly behavior report, and labeling an abnormal frequency trend line.
3. The home monitoring method for senile dementia risk based on multi-modal behavior analysis according to claim 1, wherein the specific steps when the monitoring method is used for daily abnormal behavior monitoring function are as follows: Firstly, carrying out multisource data acquisition by fusing voice analysis, facial expression recognition and physiological parameters, specifically comprising the steps of locking a target user to capture dialogue content, acquiring facial micro-expressions, synchronously monitoring heart rate variability and skin-electric response through wearable equipment, and capturing physiological stress signals when emotion fluctuates; The method comprises the steps of inputting voice characteristics into an NLP model to identify an emotion state, cross-verifying emotion authenticity by physiological data and behavior data, identifying an abnormal mode by constructing a behavior-time-space relation map, synchronously analyzing intelligent home logs and UWB positioning data, quantifying social withdrawal, and carrying out abnormal behavior matching by comparing with typical symptom characteristics.
4. The home monitoring method for senile dementia based on multi-modal behavior analysis according to claim 3, wherein the risk assessment is carried out according to the abnormal behaviors, wherein the risk assessment comprises low risk when single emotion falls, medium risk when social withdrawal is carried out for 3 days and anxiety index > threshold value, and high risk when aggressive behavior and self-injury tendency occur; And dynamically calibrating an alarm threshold according to the baseline personality, and dynamically updating the probability at the same time to generate a lunar emotion fluctuation thermodynamic diagram and annotating the personality mutation node.
5. The home monitoring method for senile dementia risk based on multi-modal behavior analysis according to claim 1, wherein the specific steps when the home safety monitoring function is performed by using the monitoring method are as follows: Tracking the gravity center position, the motion track and the posture change of a human body in real time, detecting ground wet and slippery and obstacle by combining environment sensing, marking a high risk area, monitoring the gravity center deviation and gait unbalance precursor by adopting a wearable device, and capturing physiological stress response at the moment of falling by combining physiological parameters; And (3) identifying the 'sudden falling and static' action, detecting the weightlessness acceleration and the abnormal heart rate by the synchronous wearable equipment to trigger a secondary verification mechanism, and eliminating a false alarm scene by using an algorithm synchronously.
6. The home monitoring method for senile dementia risk based on multi-modal behavioral analysis of claim 5, wherein the home safety monitoring is low risk when environmental pre-warning occurs, medium risk when posture abnormality occurs but physiological parameters are stable, and high risk when posture mutation, physiological stress and no response occur; synchronously executing local intervention operation of voice reminding and emergency lighting; Based on user habit, self-adaptive optimization is realized, and multi-user data is integrated to improve the recognition precision of complex scenes.
7. The home monitoring method for senile dementia risk based on multi-modal behavior analysis according to claim 1, wherein the specific steps when the monitoring method is used for performing the cognitive function self-test function are as follows: The target user is locked to acquire voice information, and the micro-expression and limb actions of the user are fed back in an auxiliary mode by combining the visual equipment, so that the distraction degree is analyzed synchronously; Executing a core test project, fusing the three-word recall error rate, the calculated interruption times and the response delay characteristics, outputting a cognitive risk level, updating the risk probability, and marking the key index change according to a historical monthly cognitive ability track report; The method has the advantages that the occurrence of vocabulary recall omission is less than or equal to 1, but the occurrence of calculation errors is low risk, the occurrence of vocabulary recall omission is more than or equal to 2, or the occurrence of calculation errors is middle risk, and the occurrence of failure to complete any test item is high risk; And generating a cognitive ability curve graph in a month, and marking abnormal fluctuation nodes.
8. The multi-modal behavioral analysis-based senile dementia risk home monitoring method of claim 7, wherein the core test items include: The test of three-word recall, the test process is as follows, the system firstly reads three irrelevant words with a slow speed, requires the old to review in time, analyzes the accuracy and response delay of review, immediately guides the old to conduct other activities for 5 minutes, requires the review of words after 5 minutes, records the missing number and error replacement, detects and evaluates the memory decline degree, and analyzes the first review completeness; the test is calculated by continuously subtracting 7, the test process is as follows, the system starts to gradually subtracting 7 from 100, the old people answer orally, the voice recognition engine transfers text in real time and verifies the answer, and abnormal behaviors are synchronously captured during the process.
9. A multi-modal behavior analysis-based senile dementia risk home monitoring system for executing the monitoring method according to any one of claims 1-8, wherein the monitoring system comprises a perception layer, an edge calculation layer, a platform layer and an application layer; The sensing layer is used for collecting data and comprises a millimeter wave radar array arranged at a ceiling/corner, a multispectral camera combined with infrared light supplementing, wearable equipment with an IMU sensor, an intelligent ring for measuring blood oxygen and skin electricity reaction and an environment sensor; the edge computing layer is used for processing data in real time, and is provided with a light AI model and a multi-mode data fusion engine; the platform layer is used for data analysis and storage, and is provided with a multi-source database and an intelligent analysis engine; The application layer is used for providing functional services, and four core functional modules are developed inside, namely a daily behavior monitoring module which is used for identifying repetitive actions/life skill degradation through a behavior time sequence analyzer; the intelligent monitoring system comprises a first emotion analysis module, a third emotion safety monitoring module, a fourth emotion self-test interaction module, a third emotion analysis module, a fourth emotion analysis module and a third emotion analysis module, wherein the first emotion analysis module is used for detecting harmful delusions and apathy through a multi-mode emotion calculation engine; further, a hierarchical response system and an adaptive threshold management system have been developed.
10. The senile dementia risk house monitoring system based on the multi-modal behavioral analysis of claim 9, wherein the multi-source database stores corresponding action frequency and path track content according to the data type of behavioral sequences for identifying repeated actions/disorientation, stores corresponding HRV, GSR and sleep quality content according to the data type of physiological parameters for identifying apathy and anxiety analysis, and stores corresponding hydropower use and position record content according to the data type of environmental logs for assisting in abnormal behavioral analysis; The intelligent analysis engine analyzes the abnormal behavior sequence through an LSTM+transducer model, builds a human-object-space relationship through a GNN map model, detects article misuse, and updates the illness probability according to historical data through a Bayesian dynamic network.

Description

Senile dementia risk home monitoring method and system based on multi-modal behavior analysis Technical Field The invention belongs to the technical field of health monitoring, and particularly relates to a home monitoring method and system for senile dementia risk based on multi-modal behavior analysis. Background The senile dementia risk home monitoring technology mainly surrounds the dimensionalities of abnormal behaviors, physiological indexes, environmental risks and the like, early warning is realized through non-invasive sensing and intelligent analysis technologies, a landed mature scheme covers multiple directions of behavior recognition, positioning tracking, electrocardio/sleep monitoring and the like, active operation of a patient is not needed, the cognitive ability reduction characteristic of a senile dementia patient is adapted, real-time warning and long-term behavior data tracking are realized, and family members and doctors are assisted in evaluating the illness state. Problems of the prior art: The traditional monitoring is limited in that the traditional monitoring depends on manual observation or a simple camera, and early symptoms such as repetitive actions (such as repeated salt release), disorientation (lost) and the like are difficult to identify, so that the omission rate is high; The subjective evaluation hysteresis is that the traditional monitoring is limited in that the subjective feedback of family members is relied on, and psychological changes such as harmful delusions, apathy and the like are difficult to capture in time; The passive response causes rescue delay, namely the traditional monitoring is limited in that an emergency button needs to be actively triggered and cannot be operated after falling; the screening frequency is low, the scene is limited, the traditional monitoring is limited in that medical staff is required to go to the gate or to be detected in a hospital, the period is long, and the environmental pressure influences the result. Disclosure of Invention The invention aims to provide a multi-modal behavior analysis-based senile dementia risk home monitoring method and system, which can solve the problems of difficult capture of fine behavior change, subjective evaluation hysteresis, rescue delay caused by passive response, low screening frequency and limited scene in the traditional monitoring mode. The technical scheme adopted by the invention is as follows: The home monitoring method for senile dementia risk based on multi-modal behavior analysis comprises the following specific steps: The human body posture is monitored in real time, repeated actions, orientation disorders and life skill degradation are analyzed, abnormal movements are synchronously early-warned, gait stability and heart rate variability are monitored, blood oxygen saturation and skin electric reaction are continuously collected, and voice interaction operation is carried out through a built-in AI dialogue terminal; After the data is subjected to marginal end primary screening, performing cloud multi-mode fusion analysis, performing behavior, emotion and physiological association modeling, performing cognitive decline dynamic assessment through cognitive function self-measurement, and judging the risk level according to analysis and assessment results; executing a multi-level early warning mechanism based on the risk level information, automatically taking corresponding emergency measures, and assisting the self-test of the cognitive function and the home security monitoring by executing an active intervention mechanism; The threshold value is dynamically calibrated by integrating user habit data, and a closed loop feedback mechanism is designed to realize false alarm suppression, so that a month health report is synchronously generated to assist a doctor in adjusting an intervention scheme. When the monitoring method is used for carrying out the daily behavior abnormality monitoring function, the specific steps are as follows: The method comprises the steps of integrating visual sensing, environment assistance and physiological parameters to perform multi-source data acquisition, specifically comprising capturing three-dimensional space actions, tracking human body gestures in real time, judging space orientation abnormality by combining the assistance of an intelligent household device log, and performing cross verification on the wearable device and video stream data; Analyzing an action sequence to detect an abnormal circulation mode, tracking a motion track to identify repeated loitering behaviors, constructing a human-space relation model to judge directional force disorder, analyzing article misuse behaviors to detect life skill degradation; adopting a multi-stage early warning mechanism to perform risk assessment, judging low, medium and high risk levels and taking emergency measures; and generating a monthly behavior report, and labeling an abnormal frequency trend line. When t