面向手语识别的肌音信号无线采集系统设计  

Design of MMG signal wireless acquisition system for sign language recognition

作　　者：封万俊 夏春明[1] 章悦[1] 蒋文都 刘爽 余静 FENG Wanjun;XIA Chunming;ZHANG Yue;JIANG Wendu;LIU Shuang;YU Jing(School of Mechanical and Power Engineering,East China University of Science and Technology,Shanghai 200237,China)

机构地区：[1]华东理工大学机械与动力工程学院

出　　处：《现代电子技术》2019年第18期36-39,44,共5页Modern Electronics Technique

基　　金：国家自然科学基金资助项目(51405236);上海市浦江人才计划(16PJ1402300)~~

摘　　要：为了实现对手语动作的模式识别,设计一套基于多通道肌音信号的无线采集系统.系统硬件部分采用STM32单片机以SPI通信方式收集ADXL355采集到的肌音信号,通过WiFi模块将数据传输到电脑;系统软件部分设计STM32内部程序实现对数据的采集及Matlab图形用户界面上位机采用TCP/IP通信方式收集数据实现对数据的存储.设计实验采集18组常用手语动作,提取动作小波包能量特征,运用支持向量机分类的识别率达92.3%.结果表明,该系统具有可穿戴、便携式等优点,可应用于手语识别相关领域.A wireless acquisition system based on MMG(multi-channel mechanomyographic signal)is designed to realize the pattern recognition of sign language actions. In the hardware part of the system,STM32 microcontroller is used to get MMG collected by the ADXL355 under the condition of SPI communication mode,and then the data is transmitted to computer by WIFI module. In the software part of the system,the internal program of STM32 is designed to collect MMG data,and the upper computer of Matlab GUI(graphical user interface)collects and saves the data by means of TCP/IP communication mode. Experi- ment is designed to collect 18 groups of commonly-used sign language actions and extract the energy characteristics of action wavelet packet. The recognition rate of the system with support vector machine for classification can reach 92.3%. The system is wearable and portable,and can be used in the related field of sign language recognition.

关 键 词：手语识别 肌音信号 无线采集系统 数据采集 图形用户界面 实验验证 

分 类 号：TN911.23.34[电子电信—通信与信息系统] TP301.6[电子电信—信息与通信工程]