基于MFCC与分形维数混合参数的语音识别

Method for Speech Recognition Based on Mixed Parameter of MFCC and Fractal Dimension

针对语音气流中具有混沌特征,而分形可以定量地分析混沌现象,故分形可用来分析语音信号。语音波形具有分形特征,将分形用于改善语音识别技术更好地表现语音的特征,避免传统的分段线性处理所产生的局限性。将传统特征参数MFCC与分形特征结合起来,组成混合参数用于语音识别。实验结论显示,基于MFCC与分形维数混合参数的语音识别方法要好于使用单一MFCC参数的语音识别方法。

Because of chaos characters in speech air flow, fractal can be used to quantify the chaotic phenomenon in speech signals. Speech wave appears in fractal feature. Fractal used to improve the technique of speech recognition will be more important. The traditional linear feature parameter, such as MFCC, can not represent nonlinear feature of speech. So in order to represent the feature of speech better and avoid the localization of using subsection linear method, MFCC is combined with fractal feature for speech recognition. The experiment result shows better effect is achieved by mixed parameter of MFCC and fractal dimension than only with MFCC.