基于频谱的语音识别研究——互相关卷积部分  

Based on Frequency Spectrum Speech Recognition Research——Mutual Correlation Convolution Part

作　　者：唐艺明[1] 

机构地区：[1]漳州职业技术学院,福建漳州363000

出　　处：《电子质量》2013年第10期43-46,共4页Electronics Quality

摘　　要：语音识别(Speech Recognition)是让机器通过识别和理解过程把语音信号转变为相应的文本或命令的技术。该文通过采用互相关卷积算法,实现对韵母(a,o,e,i,u,ü)的识别进行了初步的研究和探讨,实现了在MATLAB环境下对韵母(a,o,e,i,u,ü)的语音识别,并针对语音识别系统的频谱与互相关卷积的主要特点做出了总结与分析。The speech recognition （Speech Recognition） is lets the machine transforms through the recognition and the understanding process the pronunciation signal into the corresponding text or the order technology.This paper through uses the mutual correlation convolution algorithm,to realizes to the final （a,o,e,i,u,ǖ） the recognition has conducted the preliminary research and the discussion, has realized under the MATLAB environment to the final （a,o,e,i,u,ǖ ） the speech recognition,and aimed at the speech recognition system the frequency spectrum and the mutual correlation convolution main characteristic has made the summary and the analysis.

关 键 词：语音识别 快速傅立叶变换（FFT 语音频谱 互相关卷积 

分 类 号：TN912.34[电子电信—通信与信息系统]