A Review on Speech Emotion Recognition for General Cases and Indonesian Spoken Language Oscar Utomo Kumala, Amalia Zahra
Computer Science Department, BINUS Graduate Program
Master of Computer Science, Bina Nusantara University
Jakarta, Indonesia 11480
Abstract
Emotion recognition is one of many widely studied topics today. Emotions that come from speech can contain a lot information that can be used for many purposes. In this paper, a review is conducted on the studies and researches that have been done in Speech Emotion Recognition (SER) aspects. The important aspects are the speech features (acoustic, prosodic, lexical, linguistic, etc.), speech corpora (for training set and test set), and machine learning algorithms for classification. The review is conducted on the SER studies for general emotion recognition, and later focuses on the case of Indonesian spoken language. From the review, it can be seen that a feature selection method is mandatory in SER studies that involve a large number of speech features. Other findings are the importance of good speech corpus for better training set and test set, and ones of the best-performing machine learning classifier methods for most general cases are Support Vector Machine (SVM) and Extreme Learning Machine (ELM).
