Speaker recognition based on MFCC and BP neural networks

Speaker recognition has been developed over many years and it comes with many different methods. MFCC is one of more the successful methods due to it being generally modeled on the human auditory system. It represents high success rate of recognition and strong robustness against noise in the lower frequency regions. However, in the higher frequency regions, it captures speaker characteristics information less effectively. In recent years, Artificial Neural Networks have become popular. This paper presents a speaker recognition method based on MFCC and Back-Propagation Neural Networks. Experimental studies have proven that the recognition rate is successful when the number of questionable speakers is not very larger. When the number of speakers increases, the rate of recognition decreases. The potential problems and solutions are discussed, the number of training samples must be larger than the number of network model weights, 2–10 times. When the number of speakers increases, the number of training samples required also increases significantly.

I. Introduction

In modern times, our life is surrounded by AIT (artificial intelligence technology). For instance, people can get access to their own smart devices by finger print instead of typing a password. The images of car register number plates can be captured and recognized. A person's voice can be used for identification/verification purposes. The current state-of-the-art is methods of applying Deep Neural Network's to Speaker Recognition, has been proved to be successful and effective. The method uses a DNN trained as a classifier for the intended recognition task directly to discriminate between speakers for Speaker Recognition [9] [10]. It still leaves much room for improvement. and recent developments in the field of Neural Networks indicate interesting possibilities for better ways to recognize speakers [15].