Published Paper Details:

Feature selection is the basic part in the area of research. The outcomes and goal achievement of the research are mainly affected by the features chosen for the research in any field. Till date it has been observed that mostly used acoustic features for speech recognition are that Mel-Frequency Cepstral Coefficient and its derivatives. Diminishing the number of features unveils to further vigorous estimates of typical constraints and runs up the classification process more swiftly. It is very conclusive for concurrent speaker recognition while running such classification task on small source devices. Feature extraction can be implied as a phase to lessen the dimensionality of the contributed data, a lessening which obviously primes to some information malfunction. Mostly, during the speech recognition process, the speech signals are separated by using frames and then from each frame feature mining is done. In this paper, we are going to focus on some commonly used feature selection algorithms and feature dimension reduction techniques by comparing the effectiveness on the model functioning.