摘要
介绍了2种改进型的语音识别方法:单边自相关LPC系数法和线性预测误差法。这2种方法与传统的线性预测编码LPC法相比,其抗噪能力增强,即在强噪声环境下仍能达到较高的识别率。把这3种方法分别应用于端点检测和语音识别,用实验数据说明了2种改进型方法显著的抗噪性特点。
Two improved methods of speech recognition are introduced at first： One-Sided Autocorrelation LPC and LPC Prediction Error. They have better resistance to noise than traditional LPC method,and can reach higher recognition rate in the environment of strong noise. Three kinds of methods are applied to endpoint detection and speech recognition separately. The experimental data shows that the two improved methods have strong resistance to noise.
出处
《电声技术》
2005年第10期51-53,共3页
Audio Engineering
关键词
语音识别
自相关
LPC
speech recognition
autocorrelation
LPC
