Emotion recognition based on the energy distribution of plosive syllables

<p><span lang="EN-US">We usually encounter two problems during speech emotion recognition (SER): expression and perception problems, which vary considerably between speakers, languages, sentence pronunciation. In fact, finding an optimal system that characterizes the emotions overcoming all these differences is a promising prospect. this perspective, we considered emotional databases: Moroccan Arabic dialect database (MADED), Ryerson audio-visual on song (RAVDESS) present notable in terms of type (natural/acted), language (Arabic/English). We proposed detection process based 27 acoustic features extracted from consonant-vowel (CV) syllabic units: </span><em><span lang="EN-US">\ba, \du, \ki, \ta</span></em><span lang="EN-US"> common to both databases. tested classification strategies: multiclass (all combined: joy, sadness, neutral, anger) binary (neutral vs. others, positive (joy) negative (sadness, anger), sadness anger). These strategies were three times: i) MADED, ii) RAVDESS, iii) MADED RAVDESS. The method gave better accuracy case classification. rates reach average 78% for multi-class classification, 100% neutral other cases, (i.e. anger sadness), 96% emotions.</span></p>