A VOICE-BASED GENDER IDENTIFICATION METHOD USING A SUPPORT VECTOR MACHINESVM

摘要

A voice-based gender identification method using a support vector machine is provided to significantly enhance a performance of gender recognition in comparison with a conventional gender identification method using GMM(Gaussian Mixture Model) by applying the support vector machine to voice-based gender identification. A voice-based gender identification method using a support vector machine includes: detecting a voice from a voice database storing male's voice and female's voice(310); extracting a MFCC(Mel Frequency Cepstral Coefficient) feature vector for training from the detected voice(320); producing an SVM model obtaining an optimal weight vector and an optimal bias based on the extracted MFCC(330); detecting a voice from a voice signal to be tested(340); extracting a test MFCC feature vector from the detected voice(350); and applying the extracted test MFCC feature vector to the SVM model produced in a step of producing the SVM model to recognize a gender(360).