Authors:
Z.-T. Liu
1
;
Z. Mu
2
;
L.-F. Chen
1
;
J.-J. Lu
2
;
T.-Y. Li
2
;
M. Yamaguchi
2
;
F. Yan
2
;
Y.-K. Tang
2
;
K. Ohnishi
2
;
M. L. Tangel
2
;
P. Q. Le
2
;
C. Fatichah
2
;
Y. Adachi
2
;
Y. Yamazaki
3
;
F.-Y. Dong
2
and
K. Hirota
2
Affiliations:
1
Tokyo Institute of Technology and Central South University, Japan
;
2
Tokyo Institute of Technology, Japan
;
3
Kanto Gakuin University, Japan
Keyword(s):
Emotion Recognition, Violin, Music, Support Vector Regression, Fuzzy Logic.
Related
Ontology
Subjects/Areas/Topics:
Affective Computing
;
Human-Robots Interfaces
;
Informatics in Control, Automation and Robotics
;
Intelligent Control Systems and Optimization
;
Robotics and Automation
Abstract:
Emotion recognition of violin music is proposed based on strings music theory, where the emotional state of violin music is expressed by Affinity-Pleasure-Arousal emotion space. Besides the music features from audio processing, three features (i.e., left-hand feature, right-hand feature, and dynamics) with regard to both composition and performance of violin music, are extracted to improve the emotion recognition of violin music. To demonstrate the validity of this proposal, a dataset composing of 120 pieces of author-performed violin music with six primary emotion categories is established, by which the experimental results of emotion recognition using Support Vector Regression report overall recognition accuracy of 86.67%. The proposal could be an integral part for analyzing the communication atmosphere with background music, or be used by a music recommendation system for various occasions.