Based on traditional MFCC feature, this paper suggests a new kind of speech signal feature: CMFCC by introducing the method of nonlinear properties. Simulation results indicate that the method has a strong robust to noise and is able to enhance the recognition rate under low SNR.
Paper
Full text
The Improved MFCC Speech Feature Extraction Method and its Application
Semantic Scholar · Materials Science · 2013
Abstract
Based on traditional MFCC feature, this paper suggests a new kind of speech signal feature: CMFCC by introducing the method of nonlinear properties. Simulation results indicate that the method has a strong robust to noise and is able to enhance the recognition rate under low SNR.
References (8)
04Speech endpoint detection algorithm based on C 0 complexity and energy
05Gang.Improved Extraction Algorithm for MFCC Feature[J2005 · Data Acquisition and Processing
06Chinese speech recognition based on HMM and ANN
07Chinese speech recognition based on HMM and ANN[J2005 · Harbin: Harbin Engineering
08Speech endpoint detection algorithm based on C0 complexity and energy[J2009 · Computer Engineering and Applications