Bi-Spectral Acoustic Features for Robust Speech Recognition

Kazuo ONOE  Shoei SATO  Shinichi HOMMA  Akio KOBAYASHI  Toru IMAI  Tohru TAKAGI  

IEICE TRANSACTIONS on Information and Systems   Vol.E91-D   No.3   pp.631-634
Publication Date: 2008/03/01
Online ISSN: 1745-1361
DOI: 10.1093/ietisy/e91-d.3.631
Print ISSN: 0916-8532
Type of Manuscript: Special Section LETTER (Special Section on Robust Speech Processing in Realistic Environments)
bi-spectrum ,  non-Gaussianity,  phase information,  speech recognition,  

Full Text: PDF>>
Buy this Article

The extraction of acoustic features for robust speech recognition is very important for improving its performance in realistic environments. The bi-spectrum based on the Fourier transformation of the third-order cumulants expresses the non-Gaussianity and the phase information of the speech signal, showing the dependency between frequency components. In this letter, we propose a method of extracting short-time bi-spectral acoustic features with averaging features in a single frame. Merged with the conventional Mel frequency cepstral coefficients (MFCC) based on the power spectrum by the principal component analysis (PCA), the proposed features gave a 6.9% relative lower a word error rate in Japanese broadcast news transcription experiments.