Speech Emotion Recognition Overview and Experimental Results

Abstract

Nowadays, speech emotion recognition is a promising area of research mainly for human-computer interaction. Emotions play significant role in educational process. E-learning such as online classes or student-computer interaction setting may require monitoring of emotional state of students, due to the maintaining of quality of provided education. Thus, automatic speech emotion recognition represents a powerful tool for this purpose. The following paper provides an overview of speech emotion recognition and related works. A comparison of various forms of recurrent networks (LSTM, LSTM with peephole connections, GRU) and recognition accuracy on IEMOCAP database is also presented.