Speech segmentation is a difficult and prominent issue in the basic research on Hmong speech. Its essence is the boundary ambiguity between Hmong speech syllables and silent segments (silence and noise). At present, there are few related research results. In order to solve the ambiguity problem of Hmong language syllable segmentation boundary, the time domain feature segmentation was performed on the obtained initial speech boundary. By constructing the optimization evaluation function model for speech syllable boundaries, the boundary ambiguity problem between syllables and silent segments was transformed into the error minimization between the real syllable boundary and the algorithmic predicted boundary. The elite strategy was used to preserve the best optimized boundary population and extract the precise boundary between the syllable and the silent segment of Hmong speech. Experimental results demonstrate that the proposed algorithm significantly improved the adaptive boundary search capability of Hmong speech syllables. Additionally, the experimental results show that the proposed adaptive segmentation algorithm out performs typical speech segmentation algorithms in handling the challenge of ambiguous boundaries in speech segmentation.