Emotion Recognition Using Short Time Speech Analysis - Emotional Engineering - page 90

Emotion Recognition Using Short Time
Speech Analysis

Hao Zhang, Shin ' ichi Warisawa and Ichiro Yamada

Abstract In recent years, most developed countries have been facing serious issues
with an increasing number of lifestyle-related diseases. Recognizing human emo-
tions and their strength has been an essential challenge to improving healthcare
services. In this research, a purely segment-level approach is proposed that entirely
abandons utterance-level features. We focus on better extracting emotional infor-
mation from a number of selected segments within an utterance and establishing a
method for recognizing the emotion of an utterance. Validation of the proposed
method was carried out on a 50-person emotional speech database that was spe-
ci ! cally designed for this research, and a signi ! cant improvement of more than
20 % was achieved in the average accuracy compared with the existing utterance-
level approaches. Moreover, testing results based on speech signals stimulated by
the International Affective Picture System (IAPS) database showed that

the

proposed method could be also used in emotion strength analysis.

1 Introduction

Recently, increasing attention has been drawn to identifying emotions by using
speech signals. There are many reasons for the popularity of using speech signals
for emotion recognition. One main reason is that speech is the most natural and

H. Zhang ( & )
Department of Mechanical Engineering, School of Engineering,
The University of Tokyo, 7-3-1 Hongo, 113-8656 Bunkyo-Ku, Tokyo, Japan
e-mail: [e-mail]
S. Warisawa I. Yamada
Department of Human and Engineered Environmental Studies, Graduate School
of Frontier Sciences, The University of Tokyo, 5-1-5 Kashiwanoha,
Kashiwa-Shi, Chiba 277-8563, Japan
e-mail: [e-mail]
I. Yamada
e-mail: [e-mail]