Please use this identifier to cite or link to this item:
標題: Recurrent type-2 fuzzy neural network using Haar wavelet energy and entropy features for speech detection in noisy environments
作者: Tu, C.C.
Juang, C.F.
關鍵字: Fuzzy neural networks;Type-2 fuzzy systems;Recurrent fuzzy neural;networks;Wavelet transform;Speech detection;word boundary detection;inference network;robust algorithm;recognition;systems
Project: Expert Systems with Applications
期刊/報告no:: Expert Systems with Applications, Volume 39, Issue 3, Page(s) 2479-2488.
This paper proposes a new method to detect the boundary of speech in noisy environments. This detection method uses Haar wavelet energy and entropy (HWEE) as detection features. The Haar wavelet energy (HWE) is derived by using the robust band that shows the most significant difference between speech and nonspeech segments at different noise levels. Similarly, the wavelet energy entropy (WEE) is computed by selecting the two wavelet energy bands whose entropy shows the most significant speech/nonspeech difference. The HWEE features are fed as inputs to a recurrent self-evolving interval type-2 fuzzy neural network (RSEIT2ENN) for classification. The RSEIT2ENN is used because it uses type-2 fuzzy sets, which are more robust to noise than type-1 fuzzy sets. The recurrent structure in the RSEIT2FNN helps to remember the context information of a test frame. The RSEIT2ENN outputs are compared with a parameter threshold to determine whether it is a speech or nonspeech period. The HWEE-based RSEIT2ENN detection was applied to speech detection in different noisy environments with different noise levels. Comparisons with different detection methods verified the advantage of the proposed method of using HWEE. (C) 2011 Elsevier Ltd. All rights reserved.
ISSN: 0957-4174
DOI: 10.1016/j.eswa.2011.08.100
Appears in Collections:電機工程學系所

Show full item record

Google ScholarTM




Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.