The HMM-based recognition involved grouping the MFCC features of isolated-digit utterances using K-means clustering. The HMM models obtained for the isolated-digits using training data was concatenated, to predict the continous speech utterances. The report and code can be looked into for further details.