Gender based Affection Recognition of Speech Signals using Spectral & Prosodic Feature Extraction

Abstract — Speech is one of the most fundamental and natural means of communication between human beings. Human beings use emotions extensively for expressing their intentions through speech. Affection Recognition through speech signals is a current research topic in the field of human-computer interaction with wide range of applications. The proposed system recognizes the emotional state of a person based on gender. The system compose of two subsystems: Gender Recognition and Emotion Recognition.Here, a speech emotion recognition system using both the spectral & prosodic features is proposed. Since both the spectral & prosodic features contain emotion information, the combination of these features improves the performance of the system. The gender recognized speech and the features extracted are given to the emotion recognition subsystem, where the emotions are recognized based on two classifiers (i.e., two support vector machines): the one trained on the basis of signals recorded by male speakers and the other one trained by that of female speakers