A new approach: suggesting a new technique for audio signal preparation to increase the classification rate (speed) applied in speaker’s gender recognition

Increasing the classification rate in speech processing domain, particularly in recognizing speaker’s gender is considered a key challenge. In the present research, a new technique has been suggested for speeding up the classification and recognition of speaker’s gender for ten times. In the suggested technique, the new signal developed out of the main speech signal is M times as small as the main speech signal. This new signal preserves the information of its source including speaker’s gender. Therefore, MFCC feature is extracted from the new signal rather than the main speech signal. Gender recognition is done via SVM classification with RBF kernel. Since the newly designed signal is M times as small as the main speech signal, the rate of gender recognition is also significantly increased. According to the findings, the accuracy of speaker’s gender recognition via the suggested technique is 90% successful and the rate of classification is 10 times as fast.

Keywords:

increased classification rate , speaker’s gender recognition , SVM classification , MFCC feature