The speech recognition is the most important research area to recognize the speech signal by the computer. To develop the recognition rate of the continuous speech signal, we preferred frontend process such as speech segmentation, feature extraction (MFCC) and clustering techniques i.e., Fuzzy c means clustering is the formation of clusters from the extracted features based on similar sense and form the optimum number of clusters. In speech recognition the acoustic models are the major role to testing the trained data. Here the acoustic models for continuous speech recognition was discussed i.e., The Hidden morkov model (HMM),Gaussian mixture model(GMM) and GMM-UBM(Universal Background Model) are the most suitable acoustic models which are used for train the speech signal and recognize the corresponding text data.