摘要
阐述MFCC声纹特征提取的原理、MFCC特征提取模式和基于MFCC声纹识别的实现.提取模式包括MFCC提取流程、短时傅立叶变换STFT、梅尔滤波器组的构造、离散余弦变换(DCT)与MFCC特征值的提取.针对融合特征提取方案可分性与鲁棒性的缺陷提出改进方案.
Abstract
This paper describes the principle of MFCC voiceprint feature extraction,MFCC feature extraction mode,and implementation based on MFCC voiceprint recognition.The extraction mode includes the MFCC extraction process,short-time Fourier transform(STFT),the construction of Mel filter banks,discrete cosine transform(DCT),and the extraction of MFCC feature values.Propose an improvement plan to address the shortcomings of separability and robustness in the fusion feature extraction scheme.
NETL
NSTL
万方数据