Marathi speech emotion detection: A retrospective analysis

Computer and human interaction should be effective and natural. The interface will be more real and useful if it is sensitive to human emotions. Compared to image processing based facial emotion expressions, speech emotion detection is research of interest due to wider application areas and a fact that, person mostly coveys emotions via speech. This field is challenging as there may be mixture of emotions and to classify them, no well defined standards are available. Due to present life style ‘Stress’ is experienced by everybody, everyday either mild or intense. Stress is a silent killer in medical terms. Stress can exacerbate mood disorders like anxiety, fear, and depression. Overload of work, targets, nature of work, night shifts are the parameters mostly affecting Information technology (IT) professionals. It develops hypertension and diabetes resulting in depression and alcohol consumption. Stress or anxiety may result in shaky voice, stuttering voice and linguistic problem. Early detection of symptoms will reduce these cases. In India this problem is seen in 68% of IT professionals. Maharashtra being major player in IT sector, this paper focuses on stress emotion detection in local language i.e. Marathi. The paper surveys work done by researchers on speech emotion in different languages and will try to conclude about the approach for analyzing emotions in Marathi language.