Design and Development of Emotion Recognition System Using Artificial Neural Network

Abstract

The study of emotion recognition makes the human-computer newlineinteraction more natural. Automatic Emotion Recognition (AER) can be achieved by newlineusing different approaches like by using facial expressions, tone of voice (speech), and newlinebody gestures of human being. newlineIn facial expression approach, the different features like eyes, eyebrow, nose newlineand mouth are very important portions of facial region. These features are responsible newlineto decide the exact human emotion by reading the muscles statutes like distance, newlinestretchiness, or normal position of these portions. Various feature extraction newlinetechniques are used by researcher in the field of emotion recognition. newlineIn the proposed work emotion recognition using Eigen values gives 100% newlinerecognition accuracy rate for all 7 basic emotions for standard JAFFE database. The newlinerecognition rate is reduced i.e. 93.5483% for local facial expression database. newlineThe proposed work contains some Marathi emotional words. The Marathi newlineemotional words Are Bapre (and#2309;and#2352;and#2375; and#2348;and#2366;and#2346;and#2352;and#2375; !), Kiti Wilakshan (and#1048584;and#2325;and#2340;and#2368; and#1048588;and#2357;and#2354;and#1048591;and#2339; ! ), how is newlineboring (and#2367;and#2325;and#2340;and#2368; and#2325;and#2306; and#2335;and#2366;and#2355;and#2357;and#2366;and#2344;and#2375;!) Oh my God! (and#2309;and#2352;and#2375; and#2342;and#2375;and#2357;and#2366; !) etc. newlineThe Fast Fourier Transform feature extraction method for speech database newlinegives 100% recognition rate for surprise and disgust emotion and 90% for sad. The newlineemotion recognition for fear and angry is 87.50% and the overall average recognition newlinefor five emotion is achieved 93%. newline