Abstract: In this paper human unique voice characteristics used to increase the human to computer interaction (HCI) and automation in the machine uses. This system is a combination of training phase and testing phase, training phase includes the training of the system and testing phase work on the to identify the new or existed speaker information based on training. Noise elimination algorithm applied on the new audio file to eliminate the noise from voice signal and help to extract the features easily. The Mel Frequency Cepstral Coefficient (MFCC) features extraction technique used to extract unique features from voice, on the extracted feature Gaussian Mixture Model (GMM) and Dimension reduction technique applied to increase the efficiency of performance, A GMM is a parametric probability density function. GMM parameters are estimated from training data using the iterative Expectation-Maximization (EM) algorithm or Maximum a Posteriori (MAP) estimation from a well-trained prior model. GMM creates the super vectors as features for a Support Vector Machine (SVM) model for classification of a speaker voice according classified age group like child, young, adult, senior, age group classification help identify the age group and precise age of speaker, reduce the complexity of pattern matching using the SVM classification. Proposed techniques increase the performance and accuracy of system.
Keywords: Mel Frequency Cepstral Coefficient (MFCC), Gaussian Mixture Model (GMM), support vector machine (SVM), Expectation-Maximization (EM), Maximum a Posteriori (MAP), Hidden Markov Models (HMMs), Suprasegmental Hidden Markov Models (SPHMMs).