听觉计算模型在鲁棒性语音识别中的应用

利用听觉感知机理,建立一个基于听觉感知机理的语音信号特征提取模型。本文由两部分组成,一部分是在传统听觉计算模型基础上提出听觉倒谱特征AFCC(AnditoryFrequencyCepstralCoefficient)的提取方法,这样既压缩了特征维数,减小计算量,又使各个特征维之间相互独立,满足HMM模型的要求。并且根据听觉神经中枢的长时整合特性,文中提出了用低通滤波模型来模拟这种功能。结合该低通模型,提取的语音信号的听觉倒谱特征在HMM框架下取得较好的鲁律性。另一部分在研究听觉侧抑制机理的基础上,提出一个简单有效的听觉侧抑制处理模型。美尔倒谱特征MFCC谱特征经过该侧抑制模型处理,得到侧抑制美倒谱特征MFCCI,实验表明,该新特征MFCCI鲁棒性能比MFCC有大大提高。听觉倒谱特征AFCC经过该侧抑制处理得到侧抑制听觉倒谱特征AFCCI,实验表明,该新特征AFCCI鲁律性能比AFCC有大大提高。

Computational auditory model and its application in robust speech signal recognition

A computational auditory model is designed based on hearing perception mechanisms,and speech feattire can be extracted by this basic model.This paper is made up of two parts,in the first part,based on the model,an auditory representation as AFCC (Auditory Frequency Cepstral Coefficient) is proposed,thus the vector dimensions can be compressed by this method,also decorrelate each dimensions of the vector.Also a low pass filter is used to simulate the long temporal integration effect of auditory central system.With this low pass filter,new auditory spectrum is extracted,experiments show,the new feature AFCC have high robustness than traditional feature MFCC under HMM recognition system.In the second part,A very simple but efficient model of lateral inhibition is proposed based on lateral inhibition mechanism of auditory perception.MFCCI can be gotten if MFCC is processed by this model,the new representation shows high robustness in speech recognition experiment.AFCCI can be gotten if AFCC is processed by this model,the new represelltation also shows high robustness in speech recognition experiment.