Automatic segmentation of Chinese Mandarin speech into syllable-like

In this paper, we propose a novel approach to automatically segment a continuous Mandarin speech into syllable-like units. The main idea of our algorithm is to merge two round distinct selected boundaries using feature information generated from time and frequency domain. The first round segmentation is mainly based on aggregating the characteristics of different frequency regions and the second round employees zero crossing rate to improve the accuracy of segment results. The experimental results indicate that our hybrid method has high accuracy and coverage with respect to the reference boundaries even in low error tolerance.