A Comparative Study of Gender and Age Classification in Speech Signals

Accurate gender classification is useful in speech and speaker recognition aswell as speech emotion classification, because a better performance has been reported whenseparate acoustic models are employed for males and females. gender classification is alsoapparent in face recognition, video summarization, human-robot interaction, etc. althoughgender classification is rather mature in applications dealing with images, it is still in itsinfancy in speech processing. age classification, on the other hand, is also concerned as auseful tool in different applications, like issuing different permission levels for differentaging groups. this paper concentrates on a comparative study of gender and ageclassification algorithms applied to speech signal. experimental results are reported for thedanish emotional speech database (des) and english language speech database forspeaker recognition (elsdsr). the bayes classifier using sequential floating forwardselection (sffs) for feature selection, probabilistic neural networks (pnns), supportvector machines (svms), the k nearest neighbor (k-nn) and gaussian mixture model(gmm), as different classifiers, are empirically compared in order to determine the bestclassifier for gender and age classification when speech signal is processed. it is proven thatgender classification can be performed with an accuracy of 95% approximately usingspeech signal either from both genders or male and female separately. the accuracy for ageclassification is about 88%.