SPEECH EMOTION RECOGNITION SYSTEM FOR INDONESIAN LANGUAGE USING LONG SHORT-TERM MEMORY

• Main Author: JASON LASIMAN (NIM : 13514021), JEREMIA
• Format: Final Project
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/28252
• Institution: Institut Teknologi Bandung

<p align="justify"> Emotion is an aspect that always involved in interaction between humans. However, computer now still can’t interact with human through emotion or even know user’s emotion. Moreover, in Indonesia there is not much research done about interaction between human and computer involving emotion. Therefore, conducting experiment to recognize emotion is needed for progress in interaction between human and computer. <br /> <br /> Emotion recognition system is a system that can classify emotion expressed by humans. Expression of emotion can be captured as signal about image, sound, motion, etc. This emotion recognition system built with capturing sound signal along with its transcript. This system then extract feature from sound signal and words to be able to recognize emotion. <br /> <br /> Some speech emotion recognition system has been built for Indonesian language. Yet, none of that can understand the context, as sound is a sequence, because the algorithm is not able to do that. In this experiment, an algorithm that can process that is tested. <br /> <br /> The model built is using neural network, especially LSTM give better result. Model experimented are acoustic model, lexical model, and joined model. This joined model is the performance benchmark for modeling algorithm. After experimenting, LSTM algorithm can produce model with f1 measure of 65.9%. <p align="justify">