The paper presents a new variant of parameter estimation methods for discrete hidden Markov models(HMM) in speech recognition. This method makes use of a codeword dependent distribution normalization(CDDN) and a distance weighting by fuzzy contribution in dealing with the problems of robust state modeling in a FVQ based modeling. The proposed method is compared with the existing techniques using speaker-independent phonetically balanced isolated words recognition. The results have shown that the recognition rate of the proposed method is improved 4.5% over the conventional FVQ based method and the distance weighting to the smoothing of output probability is more efficient than the distance based codeword weighting.