Speech recognition using the extraction of particular feature by the discrete wavelet transform

Issue title: Proceedings of the tenth International Symposium on Applied Electromagnetics and Mechanics ISEM-Tokyo

Article type: Research Article

Authors: Midorikawa, Yoichi^{a; *} | Akita, Masanori^a

Affiliations: [a] Department of Electrical and Electronic Engineering, Faculty of Engineering, Oita University, 700 Dannoharu, Oita 870-1192, Japan

Correspondence: [*] Corresponding author: Yoichi Midorikawa, Tel./Fax: +81 97 554 7817; E-mail: ymido@cc.oita-u.ac.jp

Abstract: We have proposed the method for recognition of the signals polluted with noise using modification of spectral envelopes shape. However, these spectral envelopes have wrong valleys. For that reason recognition rate is not improved. There is plenty of scope for improvement. In this paper, we propose that method for extracting particular feature by the discrete wavelet transform. (Wavelet transform is widely used for the extraction of particular feature from data.) Pattern recognition rate is improved by the new method. It was shown that the wavelet analysis was one of the promising methodologies for the pattern recognition noisy signal.

DOI: 10.3233/JAE-2002-301

Journal: International Journal of Applied Electromagnetics and Mechanics, vol. 13, no. 1-4, pp. 13-18, 2002