Syllable based speech analysis fo affective robotics

This work presents a speech intonation processing method based on phonetic syllables and optimal pitch stylization and applied to dimensional and continuous emotion tracking. After presenting the automatic segmentation and stylization algorithms, experiments performed on emotional speech corpora show that results are competitive with the state of the art on dimensional emotion recognition and that continuous tracking is stable for the Activation axis. A real time porting of the analysis method is presented in a robotics setup where the behavior of the robot mimics the detected emotional state of a human user.