Neural Network for Bullying Emotion Recognition

Bullying is a kind of aggressive behaviour of unjustified actions and speeches which usually occur among students. Bullying behaviours are often followed by verbal disputes such as abuse, crying and other negative voices. Bullying emotion recognition can assist the detection of bullying behaviours. This paper uses neural networks for speech emotion recognition to recognize bullying behaviours. Spectrograms and MFCC features are extracted from speeches, and CNN and RBF neural networks are used for classification. According to experimental results, CNN outperforms RBF and gets an average recognition accuracy of 82.5% for bullying emotions.