Gender identification of the speaker using DTW method

摘要

This paper aims at automatically recognition of the gender through the speech signal being independent from the speaker. In this study it is purposed to decide the gender of the speaker by evaluating the distance of MFCC feature vectors. DTW method has been used to warp the time series in the process of evaluating of the distance. As a test data, as well as the speech recordings consisting of one word, the sentences consisting more than one word have been used. The test data are consisted of the records in which speakers' languages are different such as Turkish, English and German. The results of the test experimented in the study show that the accuracy is % 100 in the state of one-word while the accuracy is % 98 in the state of more than one- word.