This project was part of an assignment in our elective course, "Speech Processing". We have used MATLAB as our platform since signal processing is very intuitive in it. For any recognition system, features are the most important part. In our project, we have used the Mel-Frequency Cepstrum Coefficients(MFCC) of speech signal as our feature. For the recognition part we have used the Gaussian Mixture Model(GMM). First we train the model using training data. 10-20 seconds of speech is sufficient. The training of the model results into templates for the speakers. Later, the unknown signal is compared with the templates to find out which speaker is speaking.