Research On Gesture Recognition Based On AMI602 Motion Sensor

The gesture of human body parts is a kind of language which is used for expressionand communication, and it is not affected by differences of language, cultural andGeography. The resulting gesture recognition technology supported by computer is moreand more attention. Gesture recognition is available for gesture recognition and imitationby using pattern recognition technology in the computer platform. It is a new model ofcommunication between computer and humans. And it is more direct and visible which isdifferent from the keyboard input. In this paper, a new design of gesture recognition iscarried out which is based on the depth-analysis of progress at home and abroad.In this article, the dynamic data of gestures are obtained by AMI602 sensor，and thecollected data are corrected through zero error , orthogonal error and so on, in order toremove the impact of process factors and environmental factors. Then the collected dataare regulared by outlier elimination, median filter and smoothing filter in order to makethe data more accurate and stable.Second, gesture recognition method based on AMI602 motion sensor is proposedthrough the research of the traditional gesture recognition method as well as thecharacteristics of the AMI602 motion sensor. The attitude angles of the vector includingpitch angle, roll angle and yaw angle are obtained through the acceleration signals andmagnetic signals which are collected by the sensor. Then to design a dynamic gesturerecognition method in three dimensional space combined with the dynamic time warpinggesture recognition method.Finally, the result of gesture recognition is analysised deeply, and to compare withgesture recognition method in this article. Then it is proved that gesture recognitionalgorithm in this article is feasible and effective through verification of gesture recognitionalgorithm in this article.The algorithm in this article is based on the the software platform of Matlab2007aand it is proved that the algorithm is feasible through the processing and simulation of thedata. Finally, the recognition result is obtained and the recognition rate can reach 93.46%.