摘要
嵌入式语音识别的应用还是比较少,主要还是通过DSP实现,而且准确率还不是太高。提出一种基于FPGA和NiosII软核处理器的嵌入式语音识别系统的设计方案。系统以EP2C35 CycloneII芯片和NiosII处理器为基础,采用软硬件结合的设计方式,共同完成语音识别的设计。系统结合改进的端点检测方法,提取线性预测倒谱系数(LPCC)的音频信号特征,采用IP核硬件实现动态时间规整(DTW)的识别算法,能达到较高的识别准确率。
There are a few applications of speech recognition in embeded system,mainly emplemented by DSP,and the accuracy is low.An embedded speech recognition system based on FPGA and NiosII is recommended.The design uses EP2C35 Cyclone II chip,NiosII processor and combines software with hardware.Combined with advanced endpoint detection method,the audio signal feature in this system,is the Linear Prediction Cepstrum Coefficient（LPCC） and recognition algorithm is Dynamic Time Warping（DTW） by IP core.The system reaches high speech recognition accuracy.
出处
《计算机工程与应用》
CSCD
北大核心
2010年第2期60-62,166,共4页
Computer Engineering and Applications
关键词
语音识别
现场可编程门阵列
线性预测倒谱系数
动态时间规整
IP核
speech recognition
Field Programmable Gate Array（FPGA）
Linear Prediction Cepstrum Coefficien（tLPCC）
Dynamic Time Warping（DTW）
IP core
