Evaluation of MFCC for speaker verification on various windows

Speech identification has been an imperative part of human life acting as one of the five senses of individual body, due to which application developed on the basis of speech identification has high degree of approval. Different steps are carried out in order to analyze the speaker verification using small programs with the help of MATLAB 10.0 and for this paper; the system will recognize continuous words or sentences. This paper discusses a modified approach to implement a text-independent speaker verification system and to evaluate the recognition rate using different windows. The experiment has been performed on English language speech database for speaker recognition (ELSDSR) and Indian voice database (IVD) and has been found that as the number of training voice samples increases the recognition rate of the system increases and out all of the windows used hamming window gives the best result. Further result show that highest recognition rate can be achieved by other windows in case if test voice samples persists in one train voice samples.