8 A Swarm-Based Feature Extraction and Weight Optimization in Neural Network for Classification on Speaker Recognition

Speaker recognition is being evolving for the past decade in a noticeable way. Every person who is using mobile phones is exercising this technique. Mel-frequency cepstral coefficient (MFCC) and Linear prediction cepstral coefficient (LPCC) are the two most popular feature sets used for speech signal analysis. Nowadays, deep learning has become a part of every material which is developed using AI. By using this technique, we can be more successful in finding out the speakers in variable environments. Our chapter represents a speaker recognition method inspired upon MFCC and deep neural networking. Usually, when there is more number of speakers, rate of recognition is kept at the rock bottom. The problems and its solutions regarding participation of more number of speakers are discussed here. By giving more training samples to system, the accuracy of finding the speakers will be bettered through our proposed system.