语音分割是苗语语音基础研究的难点和热点问题，其本质是苗语语音音节与沉默段（静音、噪音）之间边界模糊问题，目前相关研究成果较少。针对苗语语音音节分割边界模糊问题，以时域特征分割获得初始的语音边界，通过构建语音音节边界优化评价函数模型，将音节与沉默段之间边界模糊问题转化为音节真实边界和算法预测边界的误差极小化问题，以精英策略保留算法最佳优化分割边界种群个体，提取苗语语音音节与沉默段间的精准边界。实验结果表明，所提方法能显著提高苗语语音音节自适应边界搜索能力。此外，所提自适应分割算法在语音分割模糊边界分割方面显著优于典型的语音分割算法。

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.