摘要
一种基于定点数字信号处理器TMS320VC5402的实时语音识别系统。能够实现非特定人、小词汇表、孤立词的语音识别。系统以线性预测倒谱参数为特征参数,采用动态时间归整识别模型,在小词汇量特定人条件下,该系统的正识率可达到98%以上,在非特定人情况下正识率在93%以上。
This paper introduces a real speech recognition system (DSP) chip. It can realize the speech recognition of nonspecific people on fixed point digital signal processor ,litth word list and isolated word. It regards linear prediction cepstrum coefficient(LPCC) as the characteristic parameter, and adopts the dynamic time warping(DTW) model. Under little vocabulary specific people terms, the recognition accuracy of this system is over 98% ,and it can be still more than 95% in ease of nonspecific people.
出处
《东北电力大学学报》
2006年第2期60-64,共5页
Journal of Northeast Electric Power University
关键词
DSP
语音识别
线性预测倒谱参数
端点检测
DSP
speech recognition
linear prediction cepstrum coefficient
terminal detection
作者简介
杨占军(1979-),男,东北电力大学计算机应用技术专业硕士研究生,主要从事DSP技术及语音信号处理方面的研究.