Estimation of Accuracy and Recall Values for the Different Age Groups Based on Voice Signals using Different Classifiers

Abstract

Voice signals are the most common and essential medium to exchange information. The generation of the voice signal can be possible due to the synchronized movement of the vocal tract. The voice signals contain extensive information about the speakers, such as age, gender, emotions, etc. The gender identification of the speaker can be done by the analysis of the voice signals. The voice samples are classified into the teenage, middle, and old age groups. This paper presents the recall values of the genders for the three age groups using different classification algorithms. This paper also concentrates on the accuracy of the gender identification system for the different types of classifiers. Mel Frequency Cepstral Coefficients (MFCCs) are used as an extracted feature. Linear Discriminant Analysis (LDA), Recurrent Neural Network Bidirectional Long Short-Term Memory (RNN-BiLSTM), and Support Vector Machine (SVM) are used as the classification algorithms to identify the gender of speakers. The male recall values in the teenage group have the highest value in the RNN classification algorithm. Similarly, the female recall values in the old age group have the highest in the RNN classification algorithm. The LDA shows the lowest recall values for each gender. It also shows the lowest gender identification accuracy. Further, the recall values of the genders and the accuracy of the proposed system of the RNN are higher than SVM and LDA.