Emotion Recognition from Speech Signals using Excitation Source and Spectral Features

The task of recognition of emotions from speech signals is one that has been going on for a long time. In the previous works, the dominance of prosodic and spectral features have been observed when it comes to recognition of emotions. But a speech signal also consists of Source level information which gets lost during this process. In this work, we have combined several spectral features with several excitation source features to see how well the model can perform the emotion recognition task. For the task in hand we have taken 3 databases namely, Berlin Emotional Database (Berlin Emo-DB), Surrey Audio-Visual Expressed Emotion (SAVEE) Database and Toronto emotional speech set (TESS) Database. The reason behind taking these databases is that the variation they offer is effective to judge the robustness of the recognition model. We chose Sequential Minimal Optimization (SMO)and Random Forest to perform classification.