Wavelet packet analysis for speaker-independent emotion recognition

Authors

Kunxia Wang
Guoxin Su, University of WollongongFollow
Li Liu
Shu Wang

RIS ID

142085

Publication Details

Wang, K., Su, G., Liu, L. & Wang, S. (2020). Wavelet packet analysis for speaker-independent emotion recognition. Neurocomputing,

Abstract

© 2020 Elsevier B.V. Extracting effective features from speech signals is essential to recognize different emotions. Recent studies have demonstrated that wavelet analysis is a useful technique in signal processing. In this study, we extract emotion features using wavelet packet analysis from speech signals for speaker-independent emotion recognition. We explore and evaluate these features from two databases, i.e., EMODB and EESDB. It is found that the extracted features are effective for recognizing various speech emotions. Furthermore, compared with common features such as Mel-Frequency Cepstral Coefficients (MFCC), these features can improve the recognition rates by 14.9 and 4.3 percentages on EMODB and EESDB, respectively.