Research On Human Behavior Recognition Based On Mobile Phone Gyroscope Data And Its Application In Medical Internet Of Things

Human behavior recognition refers to the classification and recognition of specific human actions or behavior states.This research has high scientific research value in various fields,and it is also a way for humans to deepen their self-cognition.With the popularization of Io T-related technologies,the medical industry is also accelerating research on data acquisition and analysis,and has constructed a diversified medical Io T system.Among them,behavioral data has considerable research value and potential in chronic disease control and disease prevention.Behavioral data collected through the medical Internet of Things can provide data support in intelligent medical services such as home diagnosis and treatment and remote consultation,thereby improving in medical quality and service level.Current research on human behavior recognition still faces many practical problems and challenges,such as high requirements for collection equipment when collecting behavioral data;lack of types of behavioral data used in experiments and fixed data acquisition scenarios;noise and deviations in data collection.Based on the above situation,this paper constructs a medical Internet of Things system,and on this basis,proposes a Kalman-SVM dynamic-static behavior recognition model based on dynamic window and mean optimization to classify and recognize the data collected and uploaded by mobile phones.Based on the above situation,this paper constructs a medical Io T big data system,and on this basis,proposes and uses a KalmanSVM dynamic-static behavior recognition model based on dynamic window and mean optimization to classify the human behavior data collected and uploaded by mobile phones Identification,the main research content is as follows:1.Aiming at the problems of high requirements on equipment for behavioral data collection and few types of data,this article uses common mobile phone devices as the source of human behavioral data.This paper studies the data acquisition methods of mobile phone gyroscopes,and investigates and analyzes the dynamic and static state classification of human behavior.Finally,a total of 28 specific behavior data are collected to participate in the recognition research.2.Aiming at the problem of noise and deviation generated during data collection,this paper proposes a Kalman filter algorithm based on mean optimization to filter behavioral data.This method improves the filtering effect of filtering on data noise,and at the same time reduces the dependence of Kalman filtering on a priori estimation during filtering.The innovation of the Kalman filter algorithm optimization scheme proposed in this paper is that when filtering processing for each type of data,it will uniformly calculate the overall mean of this type of data and participate in the calculation of the filtering result.This method not only retains the basic characteristics of the original data,but also reduces the value range of similar data improves the performance indicators such as recognition rate and accuracy rate for subsequent classification and recognition tasks.3.Researched and proposed a dynamic window state extraction method,which abstracted the behavior data in a period of time into a set of corresponding state data.Based on the traditional fixed window state extraction method,the idea of dynamic programming is introduced to optimize the state extraction window.When the optimized method performs state extraction,the actual window size and moving step length will be dynamically determined according to the stability of the data.This method avoids the problem of large data differences in the fixed window.The subsequent experimental results verify that the optimized state extraction window can effectively improve the recognition accuracy of the model.4.Based on the above research,construct dynamic and static identification analysis models under different optimization parameters,and identify the collected mobile phone gyroscope data.The results show that the model is under the 20% average and dynamic extraction window.The classification effect obtained is the best.Compared with the model recognition accuracy before optimization,the recognition accuracy of the model is improved,and the effect of dynamic and static behavior recognition is better.5.Aiming at the problem of behavior recognition in medical applications,this paper designs a corresponding medical Io T structure.Use 4 servers to build a Hadoop cluster under the big data processing mode,form a medical Internet of Things through cluster servers,medical equipment and smart phones,and develop mobile applications that integrate data collection and data transmission functions to achieve behavioral data acquisition and construction Complete medical Io T data analysis system and realize the application of the model.