用于说话人识别的MFCC的改进算法

Improved MFCC algorithm used to speaker recognition

在说话人识别系统中,MFCC参数是使用最多的特征参数之一。MFCC参数主要描述了表征声道特性的谱包络特征,而忽略了基音频率对它的影响。基音频率会影响MFCC参数对声道特性的准确描述,进而影响说话人识别系统的性能。本文提出了一种基于平滑幅度谱包络的MFCC的改进参数,该参数不直接对语音短时幅度谱进行提取,而是先对幅度谱进行平滑,在谱包络的基础上计算MFCC参数,以降低基音频率对其的影响。

In the speaker recognition system, MFCC parameters is one of the most characteristic parameter. MFCC parameters is main describes the spectrum envelope features, which is used to state the vacal track characterizatics, while ignoring the impact of pitch frequency. The pitch frequency will affect the MFCC parameters to accurately describe the vacal track characteristics, and then impact the performance of the Speaker Recognition System. This article proposes a improved MFCC parameters which is based on smoothing amplitude spectrum envelope. In order to reduce the impact of its pitch frequency, the parameter is not directly to extract speech short-time amplitude spectrum, but first smoothing the amplitude spectrum, and based on the spectrum envelope to calculated the MFCC parameters.