Arabic speaker emotion classification using rhythm metrics and neural networks

In this paper, rhythm metrics are calculated and used to classify five Arabic speech emotions; namely, neutral, sad, happy, surprised, and angry. Eight speakers (four male and four female) simulated the five emotions in their speech by speaking three selected sentences two times each. A human perception test was conducted using nine listeners (male and female). The results of a neural network-based automatic emotion recognition system using rhythm metrics were similar to the human perception test results, although less accurate. Anger was the most recognized speaker emotion and happiness was the least. One of our findings is that the emotions of male speakers are easier to recognize than those of female speakers. In addition, we found that the neural networks and rhythm metrics can be used for speaker emotion recognition using speech signals, but only when the dataset size is large enough.