Abstract：This paper presents a HMM Local Optimal state Path-based Data Imputation (LOPDI) algorithm. Speech feature vector sequences are presumed to be the outputs of an L state HMM. The HMM state sequence is estimated by local optimal path procedure. Then, "missing" data is recovered by MAP procedure. Experimental result shows that LOPDI algorithm can greatly increase automatic speech recognition system’s robustness against additive noise.
罗宇; 杜利民. 基于隐马尔可夫模型局部最优状态路径的数据重建算法[J]. 电子与信息学报, 2004, 26(5): 722-726 .
Luo Yu; Du Li-min. HMM Local Optimal State Path-Based Data Imputation. , 2004, 26(5): 722-726 .
2004, Vol. 26 
