The aim of this project is to perform phone classification on the Mel
Frequency Cepstral Coefficients (MFCCs) generated by a speech
recognition front-end. We have used two classification techniques,
namely Linear Discriminant Analysis and Decision Trees. The former is
a linear technique where as the latter is a nonlinear one. The
motivation for the application of the nonlinear techniques was to
observe some reduction in the misclassification error rate. Our
experiments have shown that the decision tree approach does well on
the training data but the performance degrades in case of the test
data due to over-training.