Paper Abstract and Keywords |
Presentation |
2013-01-26 09:55
Evaluation of Outliers Based on a Petal Structure of Surface EMG Signals in Motion Type Recognition for a Myoelectric Forearm Prosthesis Takuya Abiru, Hiroaki Sato, Masashi Oda, Hidetoshi Nagai, Toshiaki Ejima (Kyushu Inst. of Tech.), Nobuyuki Kimura (Active State), Yoshimitsu Kihara (Robo Future) MBE2012-83 |
Abstract |
(in Japanese) |
(See Japanese page) |
(in English) |
We are developing a Myoelectric Forearm Prosthesis with five-fingered Robot Hand.
If we can visualize how a controller in a prosthesis recognize surface EMG signals of a motion sample,that is useful to users and their trainers for improving efficiency and effectivity in training for handling a prothetic hand.
For visualization,we compressed high-dimensional feature vectors acquired from wavelet coefficients to three-dimensional vectors by LFDA in our previous work.
The 3D vectors of motion samples construct a structure in which each sample distribution of a motion radiate from a distribution of "no-motion" samples.
We call the structure "Petal structure".
A Petal structure can make distributions of motions visualize.
However,it is difficult for a user of a prosthetic hand to understand the position of the motion and its relationship with each distributions,
and to estimate how accurately he activate his muscles for a target motion.To control a prosthetic hand smoothly,we must properly discriminate outliers from samples based on accurate motions.
But in the past,we didn't define clear standards for it.
Therefore,in this paper,we propose a method to demonstrate degree of separation about characteristics of motions based on radial features of petals on Petal structure,and by the same standards to evaluate and visualize outliers. |
Keyword |
(in Japanese) |
(See Japanese page) |
(in English) |
Myoelectric Forearm Prosthesis / visualization / Petal structure / outliers / / / / |
Reference Info. |
IEICE Tech. Rep., vol. 112, no. 417, MBE2012-83, pp. 55-59, Jan. 2013. |
Paper # |
MBE2012-83 |
Date of Issue |
2013-01-18 (MBE) |
ISSN |
Print edition: ISSN 0913-5685 Online edition: ISSN 2432-6380 |
Copyright and reproduction |
All rights are reserved and no part of this publication may be reproduced or transmitted in any form or by any means, electronic or mechanical, including photocopy, recording, or any information storage and retrieval system, without permission in writing from the publisher. Notwithstanding, instructors are permitted to photocopy isolated articles for noncommercial classroom use without fee. (License No.: 10GA0019/12GB0052/13GB0056/17GB0034/18GB0034) |
Download PDF |
MBE2012-83 |
|