基于Mel倒谱特征和RBF网络的孤立词语音识别方法

Mel谱失真测度是一种弯折频率谱失真测度,用Mel频率尺度可较充分地反映人耳对频率及幅度的非线性感知特性,以及人耳在听到复杂声音时所表现的频率分析和谱合成特性。本文针对孤立词语音识别,对常规LPC倒谱特征提取方法进行改进,即将LPC倒谱按符合人耳听觉特性的Mel尺度进行非线性变化,得到LPC Mel倒谱系数(LPCMCC)作为特征参数。识别网络使用RBF神经网络,进行了孤立词语音识别。实验结果表明此种方法抗噪性能好,识别效果高。

Speech recognition method of isolated words based on Mel cpestrum feature and RBF neural network

The measurement of Mel spectrum distortion is a kind of warped frequency spectrum distortion measure.Using Mel frequency scale can reflect sufficiently the nonlinear perceptive characteristic of humans hearings to frequency and amplitude,and frequency analysis and spectrum synthesis characteristics when hearing complex sounds.Aiming at speech recognition of isolated words,an improved algorithm for normal LPC cpestrum feature is put forward in this paper.That is to say,LPCC is made nonlinear changes by means of Mel scale according to auditory characteristic,and the LPC Mel cepstrum coefficient(LPCMCC) is used as feature parameter.Through using RBF neural network to recognize current network,speech recognition of isolated words is carried on.The experiment shows that this method is good for SNR and effective on recognition.