Affective computing using speech processing for call centre applications

The paper deals with affective computing to improve the performance of Human-Machine interaction. The focus of this work is to detect affective state of a human using speech processing techniques primarily intended for call centre applications. Limited work is reported till date on affect detection using phase derived features. A unique combination of Group delay (GD), Phase delay (PD), One Sided Autocorrelation Linear Predictive Coefficients (OSALPC) and Residual Phase (RP) are used in this work. A comparative analysis is performed using Neural Networks, Support Vector Machine (SVM) and K Nearest Neighbor (KNN) to detect seven affective states. The proposed algorithm outperforms the earlier reported work using phase derived features work by a factor of around 10% for affect detection.