DYNAMIC THRESHOLDING ON SPEECH SEGMENTATION

Md. Mijanur Rahman,Md. Al-Amin Bhuiyan

Abstract: Word is the preferred and natural unit of speech, because word units have well defined acoustic representation. This paper presents several dynamic thresholding approaches for segmenting continuous Bangla speech sentences into words/sub-words. We have proposed three efficient methods for speech segmentation: two of them are usually used in pattern classification (i.e., k-means and FCM clustering) and one of them is used in image segmentation (i.e., Otsu’s thresholding method). We also used new approaches blocking black area and boundary detection techniques to properly detect word boundaries in continuous speech and label the entire speech sentence into a sequence of words/sub-words. K-Means and FCM clustering methods produce better segmentation results than that of Otsu’s Method. All the algorithms and methods used in this research are implemented in MATLAB and the proposed system achieved the average segmentation accuracy of 94% approximately.

Keywords: Blocking Black Area, Clustering, Dynamic Thresholding, Fuzzy Logic and Speech Segmentation.

DOI: https://doi.org/10.15623/ijret.2013.0209061