Improving MFCC Method for Speaker Recognition in Spying Operations

In this article, a new method is introduced for extracting features from vocal signals and discriminating speakers, which has a vast application in spy operations and passive defense sciences (for example; not only by using this method, it is possible to prevent the imitation of particular people’s voice such as commanders but also the detection of every speakers’ voice when eavesdropping the place where a number of people are discussing an important matter can be made possible). The introduced method is the improved version of MFCC method. Experimental studies say that the most useful information of vocal signals is in their low frequencies and their high frequencies are not useful in speaker recognition procedures. The method introduced in this article, the Mel frequency filter which is used in MFCC, is changed and improved. The results of MFCC and the proposed method are compared for 20 speakers and the speaker recognition percentage has improved approximately 4.5% for the linear and 9% for the exponential proposed methods.