基于PCNN和RBF的孤立词语音识别研究  

Speech recognition of isolated digit based on PCNN and RBF

作　　者：刘琨[1,2] 金文标[1] 

机构地区：[1]重庆邮电大学计算机科学与技术学院,重庆400065 [2]河南理工大学计算机科学与技术学院,河南焦作454000

出　　处：《计算机工程与设计》2008年第24期6298-6301,共4页Computer Engineering and Design

基　　金：国家863高技术研究发展计划基金项目(2007AA12Z238);重庆市自然科学基金项目(2006BB2374)

摘　　要：通过对孤立词语音识别现状的研究,提出了一种利用简化脉冲耦合神经网络PCNN和径向基函数RBF神经网络进行语音识别的新方法。利用语音信号的"可视"特性——语谱图,采用PCNN得到它的时间序列标识图作为语音信号的特征参数,辅以传统的RBF的语音识别方法,实现了孤立词语音识别。仿真结果表明,与其它方法比较,该方法能够达到较高的语音识别率。Through the present research of isolated word speech recognitions, a novel approach, based on simplified PCNN and RBF, is presented. By means of the visible spectorgram of speech, PCNN is employed to it to produce the time series icon, from which the feature parameters are extracted, and the conventional RBF is applied to recognize isolated digit. Simulation results are given to validate the efficiency of the proposed algorithm compared with other recognition methods.

关 键 词：PCNN 语谱图 特征参数 RBF 孤立词 语音识别 

分 类 号：TP391.42[自动化与计算机技术—计算机应用技术]