Methods to improve noise robustness of speech recognition systems often result in degradation of recognition performance for clean speech. Recently proposed Phase AutoCorrelation (PAC) \cite{ikbal03,ikbal03a} based features, showing noticeable improvement in noise robustness, also suffer from this draw back. In this paper, we try to alleviate this problem by using the PAC based features along with regular speech features in a multi-stream framework. The multi-stream system uses entropy of the posterior probability distribution, computed during recognition, as a confidence measure to adaptively combine evidences from different feature streams \cite{misra03}. Experimental results obtained on OGI Numbers95 database and Noisex92 noise database show that such a system yields best possible recognition performance in all conditions. Actually, the combination always performs better than the best performing stream for all the conditions.