Recognizing Speech Emotions in Iraqi Dialect Using Machine Learning Techniques

As is well known, all Arab countries speak their own dialect, which differs from classical Arabic in many ways. The Iraqi dialect is considered one of the most unique Arabic dialects due to its unique vocabulary and the fact that it pronounces some words differently than the rest of the Arabic dialects. With this peculiarity in the linguistic vocabulary, it is difficult to accurately identify the different feelings of the speaking person. In this paper, an artificial neural network-based speech emotion recognition (SER) system is proposed to detect three emotions for speakers speaking in Iraqi dialect, employing Mel-frequency cepstral coefficients (MFCC) and pitch as being effective features. Because there are no benchmark datasets for Iraqi SER, the evaluation dataset is created from scratch, and a life speech of a common sentence uttered by some Iraqi people of both genders is recorded. The experimental results show that the proposed method has significant classification accuracy, indicating that it might be used in practice.