Search

CN-121982829-A - Event prediction and alarm generation method based on multi-mode physiological time sequence data

CN121982829ACN 121982829 ACN121982829 ACN 121982829ACN-121982829-A

Abstract

The invention relates to the technical field of electric digital data processing, in particular to an event prediction and alarm generation method based on multi-mode physiological time sequence data, which comprises the steps of judging whether a suspected falling trend occurs or not; determining a plurality of preliminary abnormal segments, determining physiological feature vectors corresponding to the preliminary abnormal segments, calculating a physiological comprehensive index, determining a plurality of final abnormal segments, calculating a physiological fall index and an environment fall index, determining event types corresponding to suspected fall trend based on the difference degree of the physiological fall index and the environment fall index, generating corresponding alarm prompts based on real prediction types, and adjusting a preset impact amplitude threshold. The method effectively solves the problems of low prediction accuracy, high false alarm rate and poor self-adaptive capacity of the system caused by single data mode and excessive dependence on static parameters in the prior art through the multi-mode time sequence data fusion and the dynamic threshold adjustment mechanism.

Inventors

  • Jia Qiuhe

Assignees

  • 山东博文医疗器械有限公司

Dates

Publication Date
20260505
Application Date
20260323

Claims (10)

  1. 1. The event prediction and alarm generation method based on the multi-mode physiological time sequence data is characterized by comprising the following steps of: acquiring body surface acceleration in the daily activity process of a target individual in the endowment nursing park in real time, and judging whether a suspected falling trend occurs or not according to the body surface acceleration and a preset impact amplitude threshold value; based on the judging result of the suspected falling trend, acquiring the attitude angular speed and the gait symmetry of a target individual in real time, and determining a plurality of preliminary abnormal segments according to the body surface acceleration, the attitude angular speed and the gait symmetry in a preset prediction period; Acquiring heart rate variability, respiratory rate and blood pressure of a target individual in each preliminary abnormal segment in real time, and determining physiological feature vectors corresponding to each preliminary abnormal segment based on respective mutation amplitudes of the blood pressure, the heart rate variability and the respiratory rate and synchronous mutation features of the heart rate variability and the respiratory rate; calculating a physiological comprehensive index based on each physiological characteristic vector, and determining a plurality of final abnormal segments based on a threshold comparison result of the physiological comprehensive index; acquiring a ground friction coefficient and illumination intensity in the active environment of a target individual in each final abnormal segment, and respectively calculating a physiological fall index and an environment slip index by combining the body surface acceleration and the attitude angular speed; Determining that the event type corresponding to the suspected falling trend is a true prediction type or an interference prediction type based on the difference degree of the physiological falling index and the environment slip index; Generating a corresponding alarm prompt based on the real prediction type; And adjusting the preset impact amplitude threshold value based on the time sequence distribution characteristics of the interference prediction type and the gait symmetry in the next preset observation period.
  2. 2. The method for event prediction and alarm generation based on multi-modal physiological time series data according to claim 1, wherein the process of determining whether a suspected fall trend occurs according to the body surface acceleration and the preset impact amplitude threshold value comprises: And when the body surface acceleration is larger than the preset impact amplitude threshold, judging that the suspected falling trend occurs.
  3. 3. The method for event prediction and alarm generation based on multi-modal physiological time series data according to claim 1, wherein the process of determining a plurality of preliminary abnormal segments according to the body surface acceleration, the posture angular velocity and the gait symmetry within a preset prediction period comprises: calculating and normalizing the standard deviation of the attitude angular speed to obtain a normalized attitude angular speed; Calculating and normalizing the variation coefficient of the gait symmetry to obtain normalized gait symmetry; Calculating the fluctuation difference of the root mean square of the body surface acceleration and normalizing to obtain normalized body surface acceleration; Based on preset balance weight, the normalized attitude angular speed, the normalized gait symmetry and the normalized body surface acceleration, weighting and fusing to obtain a balance index; screening the preset prediction period in which the balance index is larger than a preset balance threshold value to determine the preliminary abnormal segment.
  4. 4. The method for event prediction and alarm generation based on multi-modal physiological time series data as set forth in claim 3, wherein the process of determining physiological feature vectors corresponding to each preliminary abnormal segment based on respective mutation amplitudes of blood pressure, heart rate variability and respiratory rate, and simultaneous mutation features of heart rate variability and respiratory rate includes: determining a synchronous mutation degree based on the correlation coefficient of the heart rate mutation degree sequence and the respiratory frequency sequence in a preset sliding window; determining a co-fluctuation degree based on the covariance of the heart rate variability sequence and the respiratory rate sequence; Determining a blood pressure fluctuation range based on the maximum value of the blood pressure and a preset blood pressure baseline; Determining a heart rate variability mutation amplitude based on a standard deviation of the heart rate variability sequence; Determining respiratory rate mutation amplitude based on the difference between the mean value of the respiratory rate and a preset respiratory rate baseline; And determining physiological characteristic vectors corresponding to the primary abnormal segments based on the synchronous mutation degree, the cooperative fluctuation degree, the blood pressure fluctuation amplitude, the heart rate mutation degree mutation amplitude and the respiratory frequency mutation amplitude.
  5. 5. The method for event prediction and alarm generation based on multi-modal physiological time series data as set forth in claim 4 wherein the process of determining a number of final anomaly segments based on threshold comparison results of physiological composite index includes: And marking the preliminary abnormal segments with the physiological comprehensive index larger than a preset physiological comprehensive index threshold value as final abnormal segments in each preliminary abnormal segment so as to determine a plurality of final abnormal segments.
  6. 6. The method for generating event prediction and alarm based on multi-modal physiological time series data according to claim 5, wherein the process of obtaining the ground friction coefficient and the illumination intensity in the active environment of the target individual in each of the final abnormal segments and calculating the physiological fall index and the environment slip index respectively in combination with the body surface acceleration and the attitude angular velocity comprises: Determining a tilt angle extremum and an imbalance duration based on the final anomaly segment, the attitude angular speed; Determining a gait symmetry variance coefficient based on the final anomaly segment and the gait symmetry; determining a shock peak and a waveform symmetry based on the final anomaly segments and the body surface acceleration; Determining the physiological fall index based on the inclination angle extremum, the imbalance duration, the impact peak, the waveform symmetry, and the gait symmetry variation coefficient; The environmental slip index is determined based on the ground friction coefficient and the impact peak.
  7. 7. The method of event prediction and alarm generation based on multimodal physiological time series data according to claim 6, wherein the determining that the event type corresponding to the suspected fall trend is a true prediction type or an interference prediction type based on the degree of difference between the physiological fall index and the environment slip index comprises: Calculating the mean value, variance, maximum value and occurrence frequency of the physiological fall indexes in the preset observation period to obtain the mean value, variance, maximum value and occurrence frequency of the physiological fall indexes; Calculating the mean value, variance, maximum value and occurrence frequency of the environment slip index in the preset observation period to obtain the mean value, variance, maximum value and occurrence frequency of the environment slip index; Calculating the correlation coefficient of the physiological fall index and the environment slip index in the preset observation period to obtain an event correlation coefficient; Determining a degree of variability based on the physiological fall index mean, the environmental slip index mean, the physiological fall index variance, the environmental slip index variance, and the event correlation coefficient; The event type is determined based on the preset observation period, the physiological integrated index, the balance index, the gait symmetry, the ground friction coefficient, the impact peak value, and the illumination intensity.
  8. 8. The method of event prediction and alarm generation based on multimodal physiological time series data according to claim 7, wherein determining the event type based on the preset observation period, the physiological integrated index, the balance index, the gait symmetry, the ground friction coefficient, the impact peak value, and the illumination intensity comprises: Based on the predicted observation period, determining the physiological composite index, the balance index and the maximum value of the impact peak value respectively to obtain a physiological composite index maximum value, a balance index maximum value and an impact peak value maximum value; Determining an illumination intensity dip amplitude maximum based on the predicted observation period and the illumination intensity; determining a ground friction coefficient mean value based on the predicted observation period and the ground friction coefficient; Determining a gait symmetry maximum coefficient of variation based on the predicted observation period and the gait symmetry; When the difference degree is larger than a preset difference degree high threshold value and the gait symmetry maximum variation coefficient is larger than a preset gait symmetry threshold value, judging that the event type is a real prediction type; and judging that the event type is an environment slipping type when the difference is smaller than a preset low difference threshold, the ground friction coefficient average value is smaller than a preset friction threshold, the impact peak value maximum value is larger than a preset impact threshold, and the illumination intensity abrupt change amplitude maximum value is smaller than a preset illumination intensity threshold.
  9. 9. The method of event prediction and alarm generation based on multimodal physiological time series data according to claim 1, wherein the process of adjusting the preset impact amplitude threshold based on the time series distribution characteristics of the disturbance prediction type and the gait symmetry within the next preset observation period comprises: Counting the total occurrence times of the interference prediction types in a preset unit time window, and determining the total number of unit time of the interference prediction event; Determining an instantaneous density based on the preset unit time window and the total number of unit time of the interference prediction event; Determining an average density based on the instantaneous density, the preset unit time window, and the preset observation period; Determining a distribution fluctuation standard deviation based on the instantaneous density, the average density, the preset unit time window and the preset observation period; And adjusting the preset impact amplitude threshold value based on the average density and the distribution fluctuation standard deviation.
  10. 10. The method of event prediction and alarm generation based on multimodal physiological time series data according to claim 1, wherein the process of adjusting the preset impact amplitude threshold based on the average density and the distributed fluctuation standard deviation comprises: When the average density is greater than a preset density upper threshold, the preset impact amplitude threshold is adjusted up; and when the average density is smaller than a preset density lower limit threshold value, reducing the preset impact amplitude threshold value.

Description

Event prediction and alarm generation method based on multi-mode physiological time sequence data Technical Field The invention relates to the technical field of electric digital data processing, in particular to an event prediction and alarm generation method based on multi-mode physiological time sequence data. Background With the increasing aging of the global population, falls have become one of the major health risks faced by elderly and mobility-impaired people. The falling down can not only cause body injuries such as fracture, dislocation of joints and the like, but also increase medical burden and influence life quality. According to world health organization statistics, falling is one of the main reasons for accidental injury of the elderly, so developing effective fall prevention technologies has urgent social demands. The traditional fall prevention measures mainly depend on manual nursing or simple sensors, but the methods have the problems of insufficient real-time performance, high false alarm rate, low individuation degree and the like. In recent years, with the development of wearable devices and artificial intelligence technology, intelligent prediction systems based on physiological data are becoming research hotspots. The system realizes early prediction and alarm of falling risk by collecting physiological signals of a user such as heart rate, blood oxygen saturation and the like and environmental data and combining a machine learning model, so that the initiative and accuracy of prevention are improved. However, the prior art still has limitations in terms of data fusion, model adaptability and alarm mechanism, and further optimization is needed to improve reliability. The patent document with the publication number CN119970009A discloses an intelligent prediction alarm system for preventing falling, which comprises a data acquisition and recording module, a falling prediction module, a first alarm module and a second alarm module, wherein the data acquisition and recording module is used for acquiring a first characteristic dataset of a user and a real-time environment dataset of the user, the falling prediction module is used for constructing a prediction model, inputting the recorded first characteristic dataset and the real-time environment dataset into the prediction model for analysis, outputting the falling probability of the user, the first alarm module is used for carrying out first alarm reminding when the falling probability of the user reaches a preset condition, the second alarm module is used for carrying out second alarm reminding after the falling of the user is monitored, meanwhile, acquiring the second characteristic dataset after the falling of the user, carrying out secondary analysis on the second characteristic dataset based on the prediction model, and updating the falling probability of the user in real time based on an analysis result. Therefore, the prior art has the following problems that firstly, the data mode is single, the movement data such as acceleration and angular velocity and the basic data such as age and falling history are mainly relied on, but key physiological indexes and environmental factors are not integrated, so that the model cannot comprehensively capture multidimensional dynamic characteristics of falling risks, the model is insufficient in adaptability, the model is judged by relying on a fixed threshold value, a dynamic adjustment mechanism based on historical data or environmental change is lacking, the false alarm rate and the false alarm rate are increased due to frequent data drift or interference events in long-term use, the event classification is rough, the alarm is output only through a single probability value, different types such as self unbalanced falling and environmental slipping are not distinguished, and the risk cause is accurately identified due to the lack of pertinence of an intervention strategy and the influence on nursing efficiency. Disclosure of Invention Therefore, the invention provides an event prediction and alarm generation method based on multi-mode physiological time sequence data, which is used for solving the problems of low prediction accuracy, high false alarm rate and poor self-adaptive capacity of a system caused by single data mode, neglected individual difference and excessive dependence on static parameters in the prior art through multi-mode time sequence data fusion, a real-time monitoring technology and a dynamic threshold adjustment mechanism. In order to achieve the above object, the present invention provides a method for event prediction and alarm generation based on multi-modal physiological time series data, comprising: acquiring body surface acceleration in the daily activity process of a target individual in the endowment nursing park in real time, and judging whether a suspected falling trend occurs or not according to the body surface acceleration and a preset impact amplitude thres