Emotion recognition from Moroccan dialect speech and energy band distribution

The aim of this study is to propose an approach to the automatic recognition of emotions from the signal speech [18], [13]: it relies on the extraction of spectral and prosodies characteristics, more particularly on the energy and the distribution of this energy in six bands. These characteristics are obtained from syllabic units (CV) where C denotes a consonant and V a vowel. To this end, we constructed a Moroccan emotional database (MEDB) from broadcasts on the YouTube channel representing different real emotional states. The features set is obtained from Matlab codes and Praat [23] software. The emotions used in this study are anger, joy, neutral state and sadness. The emotion recognition systems are developed using neural networks, Support Vector Machine and decision trees algorithms. The aim of this work is firstly to explore syllabic units to identify the emotions embedded in the speech signal and secondly to demonstrate the relevance of the energy distribution in determining emotional states.