Automatic boundary detection based on entropy measures for text-independent syllable segmentation

Abstract

In this paper, we study the boundary detection in syllable segmentation field. We describe an algorithm proposed for text-independent syllable segmentation. This algorithm provides a performance comparison between the entropies of Shannon, Tsallis and Renyi in an effective detection of beginning-ending points of syllable in a speech signal. The Shannon generalizations (Tsallis and Renyi) quantify the degree of signal organization and offer the relevant information such as the voicing degree on the first syllable segment that we obtained from the temporal dynamics of singularity exponents. The method we propose is focused on an aggregation measure based on entropies to enhance the syllable boundaries detection. It has been also demonstrated in this paper that the best suited entropy for efficient boundary detection is Renyi entropy. Once evaluated, our algorithm produced better performance with efficient results on two languages, i.e., the Fongbe (an African tonal language spoken especially in Benin, Togo, and Nigeria) and an American English. The overall accuracy of syllable boundaries was obtained on Fongbe dataset and validated subsequently on TIMIT dataset with a margin of error < 5ms.