Single Gauss Model set based Data Imputation (SGMDI) method is developed to recover Mel-frequency-filter-bank vectors of Chinese speech.Experiments are carried out to study how SGMDI method improves Automatic Speech Recognition (ASR) system's robustness against additive noise.Experimental results show that SGMDI method can improve phoneme correction of all kind of phonemes.Especially for unvoiced phonemes

which are easily distorted by additive noise

phoneme correction will be significantly improved.Thus

ASR system's robustness against additive noise can be greatly improved by SGMDI method.