Abstract
This paper presents a comparison and analysis of speech emotion recognition in the context of Arabic and English languages. Four emotions (neutral, sadness, happiness and anger) were considered from two speech corpora: the King Saud University Emotions (KSUEmotions) corpus for Arabic and the Emotional Prosody Speech and Transcripts (EPST) corpus for English. Six speakers (three men and three women) were selected from each corpus. Many acoustic features were extracted for use in the recognition and analysis stages. Additionally, an Analysis Of Variance (ANOVA) was used to determine which acoustic features should be used in our emotion recognition system. Results show that there is a benefit in terms of emotion recognition for Arabic words with the use of specific acoustic features. Results also show that certain speech features, such as the first three formants, help in the accuracy of emotion recognition.