一种基于自组织神经网络的语音识别系统

建立了一种基于自组织神经网络的语音识别系统。对语音信号进行了预处理,提取了语音信号的线性预测系数、线性预测倒谱系数和Mel倒谱特征系数,建立了基于自组织神经网络的识别判决模型。深入分析和改进了自组织神经网络的分类聚类能力,通过加强训练和设定阈值函数的方法,有效地确定了边界神经元的归属,划分出了合理的输出模式类。验证了自组织神经网络适合于处理孤立词语音识别,并具有快速性和结构简单等特征。MATLAB仿真实验表明,语音识别率达到96%。

Research on the Speech Recognition System Based on Self-organizing Neural Networks

A speech recognition method, which is based on self-organizing neural networks is presented. After the pretreatment, the linear prediction coeffcients, linear prediction cepstrum coeffcients and Mel-frequency cepstrum coeffcients with which the speech can be recognized by the self-organizing neural networks are extracted. The classified and clustering ability of the self-organizing neural networks are analyzed and improved. Through reinforcing the training and setting the gate, the borderline neurons are confirmed. The self-organizing neural networks is fast, simple and fit for speech recognition. The experimental results show that the recognition accuracy reaches 96%.