TY - JOUR
T1 - Wayang robot with gamelan music pattern recognition
AU - Tomo, Tito Pradhono
AU - Schmitz, Alexander
AU - Enriquez, Guillermo
AU - Hashimoto, Shuji
AU - Sugano, Shigeki
N1 - Funding Information:
This study was conducted as a part of the project (13L03) of the Research Institute for Science and Engineering, Waseda University, and as a part of researches at Humanoid Robotics Institute and Future Robotics Organization, Waseda University. This work was supported in part by Grants for Excellent Graduate Schools of JSPS and MEXT.
Publisher Copyright:
© 2017, Fuji Technology Press. All rights reserved.
PY - 2017/2
Y1 - 2017/2
N2 - This paper proposes a way to protect endangered wayang puppet theater, an intangible cultural heritage from Indonesia, by turning a robot into a puppeteer successor. We developed a seven degrees-offreedom (DOF) manipulator to actuate the sticks attached to the wayang puppet body and hands. The robot can imitate 8 distinct human puppeteer’smanipulations. Furthermore, we developed a gamelan music pattern recognition, towards a robot that can perform based on the gamelan music. In the offline experiment, we extracted energy (time domain), spectral rolloff, 13 Mel-frequency cepstral coefficients (MFCCs), and the harmonic ratio from 5 s long clips, every 0.025 s, with a window length of 1 s, for a total of 2576 features. Two classifiers (3 layers feed-forward neural network (FNN) and multi-class Support Vector Machine (SVM)) were compared. The SVMclassifier outperformed the FNN classifier with a recognition rate of 96.4%for identifying the three different gamelan music patterns.
AB - This paper proposes a way to protect endangered wayang puppet theater, an intangible cultural heritage from Indonesia, by turning a robot into a puppeteer successor. We developed a seven degrees-offreedom (DOF) manipulator to actuate the sticks attached to the wayang puppet body and hands. The robot can imitate 8 distinct human puppeteer’smanipulations. Furthermore, we developed a gamelan music pattern recognition, towards a robot that can perform based on the gamelan music. In the offline experiment, we extracted energy (time domain), spectral rolloff, 13 Mel-frequency cepstral coefficients (MFCCs), and the harmonic ratio from 5 s long clips, every 0.025 s, with a window length of 1 s, for a total of 2576 features. Two classifiers (3 layers feed-forward neural network (FNN) and multi-class Support Vector Machine (SVM)) were compared. The SVMclassifier outperformed the FNN classifier with a recognition rate of 96.4%for identifying the three different gamelan music patterns.
KW - Intangible culture
KW - Intelligent machine
KW - Machine learning
KW - Music information retrieval
KW - Wayang kulit
UR - http://www.scopus.com/inward/record.url?scp=85013880391&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=85013880391&partnerID=8YFLogxK
U2 - 10.20965/jrm.2017.p0137
DO - 10.20965/jrm.2017.p0137
M3 - Article
AN - SCOPUS:85013880391
SN - 0915-3942
VL - 29
SP - 137
EP - 145
JO - Journal of Robotics and Mechatronics
JF - Journal of Robotics and Mechatronics
IS - 1
ER -