Research Of Gesture Recognition Method Design And Performance Evalution Based On DSP/BIOS

With the continuous development of human-computer interaction and informationtechnology, as a natural, simple and efficient way to interact, gestures have been attractingmore and more attention in the field of human-computer interaction and researches relatedto gesture recognition are developing rapidly.Therefore, a gesture recognition method andperformance evaluation program based on hardware is significant to preferablyoptimization and product process of gesture recognition.This paper builds a DSP/BIOS real-time system for gesture recognition and algorithmperformance evaluation, in the background where experimental validation is relativelyinadequate, for gesture recognition method in the embedded system platform. An I-DTWgesture recognition algorithm is proposed based on in-depth study of the currentmainstream gesture recognition method. Compared with other algorithms, the algorithmhas low complexity, a relatively small amount of computation and high recognitionaccuracy and other advantages. In this paper, we port the I-DTW algorithm to theDSP/BIOS real-time system processing platform, and realize the real-time gesturerecognition and algorithm performance evaluation.Firstly, this paper compares and analyzes several currently popular gesturerecognition algorithms for algorithm complexity and accuracy of identification, and thendesigns the I-DTW gesture recognition algorithm for DSP/BIOS system.Secondly, gesture data signals are collected by the acceleration sensor AMI602, andthen these data will be sent to the DSP/BIOS system processing platform through theUART serial port for real-time data processing.Thirdly, the I-DTW gesture recognition algorithm has been rewritten in C language,and realizes the algorithm software migration on the DSP/BIOS system.Finally, the paper makes an experimental verification and algorithm performanceevaluation experiment of the gesture recognition method on the DSP/BIOS systemevaluation platform.Experiments show that this design of gesture recognition method is feasible in the DSP/BIOS embedded systems, and can obtain a higher recognition rate. The system canmake a performance evaluation from the required number of instruction execution, CPUload and other aspects of recognition algorithm, and verify the effectiveness of theproposed gesture recognition method and DSP/BIOS system performance evaluation.