IMU-based Algorthms Research For Human Movement Patterns And Person Identification

Human movement pattern recognition and person identification are technologies that are used to accurately classify individuals by exploring the differences in characteristic features in different movements.In view of the high equipment cost,low recognition resolution,more complex algorithm and other problems that exist in the current recognition method,this paper is based on the IMU motion sensing unit to study the use of human gait characteristics for pattern recognition in different movement states and for the same movement mode.Recognizing the identification of the different characters,the main conclusions of the subject are as follows:1.For the collection of human motion data,select an intelligent portable device based on the Android system and write an APP for data collection;the best 50Hz sampling frequency was then determined and the wrist was used as the device wearing position;afterwards,the design is based on The data collection experiment with 50 testers;finally a sliding window method with 50%overlap was used for the Pre-processing and the data characterization analysis of the collected data.The results show that human body movement has a more obvious periodicity,which provides reliable data support for subsequent identification research.2.Feature analysis and research on the four movement patterns of the human body in standing,jogging,running and going up and down stairs;then,according to the movement patterns,the feature value of the wrist gait will be extracted;then the SVM classification algorithm was used to achieve the recognition of movement patterns;and finally the grid search optimization algorithm is used to optimize the parameters of the SVM classifier.A total of 20 feature values are selected based on the feature differences of the four types of movement patterns,and the optimization yields the best parameter pairs C and g(0.25,0.0625),with 100%recognition of the four movement patterns.3.Analyze and study the differences in the gait characteristics of the wrist in different human beings under the same exercise mode;and then obtain the following results according to the study.Four categories of wrist gait differences,angular,positional,inertial,and coherence differences,are selected to obtain 25 eigenvalues;and subsequently Calculation of recognition weights based on the dispersion of differences between different feature values in the human body and construction of a new mathematical model of 25-dimensional features.Finally,the BP neural network recognition method was used to achieve person identity,and for the network implicit layer nodes,training iterations number of Contrast analysis is performed to determine the network parameters.The experimental results show that the algorithm has an accuracy of 96.747%for 20 samples and is capable of realizing the accurate recognition of the identity of people in a small group.The research content of this paper can provide some reliable reference for human movement pattern recognition and person identification in specific environments,and provide the theoretical basis for recognition algorithms for security monitoring,security authentication,and behavior recognition in production life.