摘要
系统以16位数字信号处理器TMS320VC5509为核心,采用TI音频编码解码芯片AIC23B对语音信号进行采集和编码转换。在对语音信号经过预处理、双门限端点检测、MFCC特征参数的提取后建立语音模板,利用CWRTs(Cross-Words Reference Templates)算法融合模板库,最终在DSP系统运行实现多人语音识别并测试。测试结果表明,系统能满足实时性能要求,识别效果良好。
Based on the16-bit digital signal processor TMS320VC5509,the system acquires and transcodes voice signal using TI audio codec chip AIC23B.We built voice templates after pretreating voice signal and accomplishing the detection of double threshold endpoint and extraction of MFCC characteristic parameters.Then we used CWRTs(Cross-Words Reference Templates)algorithm to fuse template library and finally run the DSP system to implement multi person voice recognition and testing.The test results show that the system can meet the real-time performance requirements,and the recognition result is good.
作者
陈业慧
陈磊
吴房胜
Chen Yehui;Chen Lei;Wu Fangsheng(School of Electronic Communications Engineering, Anhui Xinhua University, Hefei 230088;Anhui Business Vocation College, Hefei 230088,China)
出处
《北京印刷学院学报》
2017年第7期178-181,共4页
Journal of Beijing Institute of Graphic Communication
基金
2017年安徽新华学院校级质量工程项目"单片机教学团队"(编号:2017jxtdx01)
2016年安徽省质量工程项目"大学生"科技圆梦"创客实验室"(编号:2016ckjh095)
