基于VQ和HMM的双层声纹识别算法

摘要

为了给盲人设计一种可靠度高、功耗低、方便快捷的嵌入式说话人身份识别系统,提出一种基于VQ矢量量化和HMM隐马尔科夫模型的双层声纹识别算法.该算法采用复杂度相对低的VQ矢量量化算法,对模板库进行快速初次筛选,得出有效匹配结论或选出可能符合的模板, 缩小模板库范围,并采用离散HMM隐马尔科夫模型进行精确识别.仿真结果表明,该算法在保证识 别精度的同时,缩短了识别时间,提高了识别效率.%In order to design a reliable, low-power and efficient speaker recognition embedded system which is used to help the blind, the two-level speaker recognition model based on VQ and HMM is put forward.The VQ algorithm, which has low complexity, is adapted to filter the library quickly.The no matching result can be made or the possible templates can be selected after this.Then the HMM is employed to make an accurate conclusion.The simulation result shows that the algorithm can reduce the time of recognition with good recognition rate, which improves the recognition efficiently.