【摘要】 目的探索基于表面肌电信号(surface electromyography,SEMG)中国手语手势识别的可行性,尽量减少用户执行手势动作主观用力差异对SEMG分类的不利影响。方法以中国手语30个字母手势动作为研究对象,通过分析手势动作过程,提出一套基于SEMG的中国手语手势动作规范方案。在实验室搭建的识别系统上,15名受试者参加了特定用户和与用户无关的规范前后识别实验。结果经过规范学习后,30种手语手势的动作识别率得到了显著提高,平均提升值为37%。结论手势动作规范可使不同动作的SEMG具有更高的可分性,此研究结果为实现基于SEMG大词汇量的中国手语手势识别提供了一种可行的方案。更多还原

【Abstract】 Objective To explore the feasibility of Chinese sign language(CSL) recognition based on surface electromyogram(SEMG) and to eliminate the effects on SEMG classification caused by subjective force differences when users perform hand gestures.Methods Thirty CSL alphabet gestures were presented as exemplification in the study.A set of gesture definition improvement and action normalization schemes were proposed for SEMG-based CSL recognition.According to the analysis of sign language gesture process,the experiments were conducted for 15 subjects in user-specific and user-independent classification sessions,on SEMG-based CSL alphabet recognition system under laboratory condition.Results The average recognition accuracies of 30 CSL alphabets were improved by 37% after user learned the action normalization schemes.Conclusion The proposed action normalization schemes can enhance the classification ability significantly.And the study provides a feasible method for realizing large vocabulary SEMG-based CSL recognition.更多还原