Emotion recognition using auditory cues

In this study, the development of a system capable of classifying human emotions during conversations in Turkish language is aimed. Thus, an artificial intelligence or a robotic system can be produced, which is able to analyze the mood of the users and provide services specific to a person. To achieve this a Turkish emotion corpus is created, potential practical problems and their solutions are considered. Furthermore, the impact of auditory feature extraction algorithms on the recognition performance are evaluated for Turkihs language, and the training methods to be used in the classifier are tested. In the offline experiments, the performance of the recognizer was 87% using four classes of emotions and 77% using seven classes of emotions.