The Extraction of Differential MFCC based on EMD
Feature extraction is the key to the object recognition.How to obtain effective,reliable characteristic parameters from the limited measured data is a question of great importance in feature extraction.This paper presents a method based on Empirical Mode Decomposition (EMD) for the extraction of Mel Frequency Cepstrum Coefficients (MFCCs) and its first order difference firom original speech signals that contain four kinds of emotions such as anger,happiness,surprise and natural for emotion recognition.And the experiments compare the recognition rate of MFCC,differential MFCC (Both of them are extracted based on EMD) or their combination through using Support Vector Machine (SVM) to recognize speakers emotional speech identity.It proves that the combination of MFCC and its first order difference has a highest recognition rate.
Feature extraction EMD Mel Frequency Cepstrum Coefficients (MFCCs) first order difference
Yunyun Chu Weihua Xiong Weiwei Shi Yu Liu
College of Mechanical and Automation Zhejiang SCI-TECH University,Hangzhou,Zhejiang Province,China
国际会议
济南
英文
1167-1170
2012-12-29（万方平台首次上网日期，不代表论文的发表时间）