In this paper, we presents a scheme of implementation of automatic speech recognition system using Odia language. The proposed system is realized using Kaldi toolkit for automatic recognition of district models of the state of Odisha. We investigate the monophone and triphone model for Odia speech recognition. Acoustic modeling of Odia speech recognition was performed using HMM and GMM. A significant reduction in word error rate (WER) was observed using the triphone model. We report the preliminary speech recognition results using Kaldi speech toolkit.