Personalised Emotion Recognition Utilising Speech Signal and Linguistic Cues

This study proposes an emotion recognition system, an essential part in many affective computing application for natural language processing, utilizing open source platforms with combined Automatic Speech Recognition and Text analysis for a speaker personalized system. The suggested realization is accomplished by exploiting Pocketsphinx as an Automatic Speech Recognizer and linear support vector machine(Linear SVM) for multiclass Text analysis and classification. Further, training and testing of the suggested emotion recognition is carried out using standard annotated Emotion database of International Survey on Emotion Antecedents and Reactions (ISEAR). Ultimately, an aim lies in arriving at a speaker dependent system, which serve as a personalized assistant for classifying human emotions that finds novel application in Human Computer Interaction (HCI). The tests regarding the performance accuracy will be conducted using ISEAR data bases.