Event-driven Based Design Of Gait Recognition Method

With the development of technologies such as ubiquitous computing and wireless communication,Body Sensor Network(BSN)has been widely used in sports training,medical diagnosis,social security and other fields.In the research field of BSN,human gait recognition has become a hot spot.By recognizing human behavior,computers can perceive and make sense of motivations,which can provide basic services for specific research topics such as game entertainment,medical monitoring and sports assisted training.Human gait recognition methods usually use accelerometers to collect acceleration data and use slide window technology to segment the data.On this basis,features are extracted from the segmented data to achieve gait recognition.These gait recognition methods have some shortcomings: the acceleration data is susceptible to gravity acceleration,which identifies fewer types of gait and does not recognize changes of human body turning.The slide window is difficult to divide data based on the action period of the asynchronous gait.And this technology needs to process duplicate data,resulting in large amount of calculation and low real-time performance.This paper designs a gait recognition method based on event-driven,which can make up for the above deficiencies: This method recognizes gait behavior by combining angular velocity with acceleration,thus increasing recognizable gait types,and recognizing the change of human body turning by quaternion.An event-driven data partitioning method is adopted to detect gait behavior by searching for key features corresponding to gait events,and dividing data according to gait cycle.It does not need to process duplicate data,and has less computation and high real-time performance.The gait recognition method can extract 19 kinds of waveform features and behavior features from the segmented data to identify gait behavior.This paper designs and implements a gait recognition system based on event-driven gait recognition method.The gait recognition system can recognize 15 kinds of common gait behaviors in daily life.After testing,the average accuracy of the gait recognition system for identifying the asynchronous behavior is 96.8%,and the accuracy of identifying the gait behavior of different objects ranges from 96.4% to 97.6%.The system has the advantages of identifying more types of gaits and higher recognition accuracy.