Gender Identification from Speech Signals Using Mel Frequency Cepstral Coefficients Based Feature Extraction

Gender identification of a speaker, which is the everyday distinguishing speech's characteristic. It can effortlessly be identified by an individual who hears it. It is substantially vital for many applications to identify gender information driven from signals of speech. With the help of gender recognition, the systems which are dependent on gender are defined. Proper gender identification can increase the efficiency and robustness of any gender-dependent system. In this research, Identification of gender is developed using MFCC coefficients and other acoustic properties taken from signals of speech with GMM. Testing was conducted using Surftech's Free American English Dataset (SLR45) with speech from ten speakers (five females and five males). So here we are determining the gender of a speaker using MFCC and other acoustic features and GMM and five other types of machine learning algorithms (Neural Network, Decision tree, Random Forest SVM and Gradient boosting) for classification of gender. The results achieved show that GMM and gradient boosting perform better using MFFCC and other acoustic features.