NATURE INSPIRED OPTIMIZATION ALGORITHMS IN ARTIFICIAL NEURAL NETWORK FOR SPEAKER RECOGNITION

Speaker recognition system has gained significant research interest, owing to security enforcement in many applications. Typically, the speaker recognition system is employed for authentication means identifying the person using their voices. Hence, a speaker recognition system must be capable of achieving greater recognition accuracy rates irrespective of text that speaker spokes. This work is concerned with ANNs that have turned out to be an intense pattern recognition tools effectively utilized for some real world applications in the course of the last few years. The novelty of these techniques relies on restructuring the conventional structure of ANN (layer and neurons) optimally. This work incorporates couple of evolutionary techniques and swarm intelligence techniques namely EA, GA and PSO. The results intent that the restructure network reveals optimal performance (classification accuracy) over conventional ANN structure.