A novel emotion recognition technique from voiced-speech

In the framework of the beginning of the investigation due to a work of an undergraduate student, the authors at Mathematical Modeling Research Group (GRIMMAT) propose the use of emotion recognition algorithms previously developed by them adapting it to the FAU Aibo emotion corpus which was the database used in the INTERSPEECH 2009 Emotion Challenge. Firstly, by resampling the audio signal and windowing process, the audio signal is segmented. Next, each segment is decomposed through the discrete wavelet transform, then the descriptive characteristics of the decomposed signal are extracted. Finally, a supervised classification scheme is used. This paper presents the main results and conclusions obtained.