Hand Motion Identification Using Independent Component Analysis of Data Glove and Multichannel...
1
Hand Motion Identification Using Independent Component Analysis of Data Glove and Multichannel Surface EMG Pei-Jarn Chen, Ming-Wen Chang, and and Yi-Chun Du * Department of Electrical Engineering, Southern Taiwan University of Science and Technology, Tainan, Taiwan * Corresponding Author:[email protected] The applications of hand motion identification (HMI) have been widely used for prosthetic control, remote control of home appliances, Human Computer interaction (HCI).. In this study, we used five bend sensors mounted on a data glove with bend sensors as signal sources for finger motions identification. In addition, Independent component analysis (ICA) has been demonstrated that could extract independent components from multi- channel EMG that are corresponding to specific muscle groups not only dimensionality reduction of required computation data for identification hand motion. Introduction Material and Methods Five channels of bend sensors in the data glove and three channels of SEMG sensors in the MES are used to detect finger and forearm motions. Independent component analysis (ICA) and grey relational analysis (GRA) is used to data reduction and as the core of identification. Six features are extracted from each SEMG channel, and three features are computed from five bend sensors in the data glove. After signal acquisition, motion intervals were detected from the SEMG signals. Motion intervals define the time intervals for subsequent analysis of both SEMG and data glove signals. Then, pre-processing and feature extraction were applied within the identified time intervals. The features from the SEMG and data glove signals were combined in a single feature pattern. Finally, the feature patterns were input to a trained ICA and GRA based classifier. The ICA and GRA based classifier is used to identify 20 hand motions and gestures recognition from 10 subjects. The results of this study indicate that the ICA is a reliable pre-process to reduce the amount of data and, in turn, reduce the processing time about 30% in the GRA method, but the performance of the system decreased significantly following ICA in BPNN method. That maybe caused by difficult convergence in BPNN. The performance of the proposed HMI system using bend sensors and multichannel SEMG was based on accuracy, number of hand motions and training times needed to meet the requirement for real- time processing. The architecture of the proposed system meets the requirement of portability, but with the flexible mechanism, the proposed method can be Results Conclusion Fig. 1. The Structure of the proposed HMI system with ICA and GRA EM G signalsfrom three electrodes Dataglove signals from Flex sensor Energy Distribution (1) (2) (3) (4) (5) Ten subjects (average age, 24±3 years) were recruited to evaluate the performance of the proposed method. Each subject performed a set of 20 different gestures while standing. Figure 2 shows the 20 hand motions classified by the proposed. Fig.2 Motion intervals of energy evaluation with acting five hand motions Fig.3. Twenty hand motions are assigned in number order Fig.4 Average training times (N=10) versus the numb training sets for GRA and BPNN with/without ICA M ES pre-processing D ata G love signal pre-processing SEM G feature extraction Bend Sensors feature extraction IEMG WL WAMP SSC ZC VAR Area LC MV H and m otion classification ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) 2 ( ) 2 ( ) 2 ( ) 2 ( ) 1 ( ) 1 ( ) 1 ( ) 1 ( 2 1 2 1 2 1 2 1 K K K K k k k k x n i n i n i n i Comparison array Real-time array Euclidean distance Comparison array 0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2 3 . 0 1 . 0 0 . 1 5 . 0 0 . 5 0 . 3 G rey Relational G rade ( r) Euclidean Distance(ED ) ICA ICA Φ SEM G Φ ICA Φ Glove Training Stage Threshold Updated IC num ber evaluation by PCA Database D ataG love M ES Training Sets 0 1 2 3 4 5 6 7 8 9 10 C P U tim e in lo g scale (sec.) 0.01 0.1 10 100 GRA G RA +ICA (Proposed m ethod) BPNN BPN N +ICA
-
Upload
scott-hunter -
Category
Documents
-
view
214 -
download
2
Transcript of Hand Motion Identification Using Independent Component Analysis of Data Glove and Multichannel...
Hand Motion Identification Using Independent Component Analysis of DataGlove and Multichannel Surface EMG
Pei-Jarn Chen, Ming-Wen Chang, and and Yi-Chun Du *
Department of Electrical Engineering, Southern Taiwan University of Science and Technology, Tainan, Taiwan
* Corresponding Author:[email protected]
The applications of hand motion identification (HMI) have been widely used for prosthetic control, remote control of home appliances, Human Computer interaction (HCI).. In this study, we used five bend sensors mounted on a data glove with bend sensors as signal sources for finger motions identification. In addition, Independent component analysis (ICA) has been demonstrated that could extract independent components from multi-channel EMG that are corresponding to specific muscle groups not only dimensionality reduction of required computation data for identification hand motion.
Introduction
Material and Methods
Five channels of bend sensors in the data glove and three channels of SEMG sensors in the MES are used to detect finger and forearm motions. Independent component analysis (ICA) and grey relational analysis (GRA) is used to data reduction and as the core of identification. Six features are extracted from each SEMG channel, and three features are computed from five bend sensors in the data glove. After signal acquisition, motion intervals were detected from the SEMG signals. Motion intervals define the time intervals for subsequent analysis of both SEMG and data glove signals. Then, pre-processing and feature extraction were applied within the identified time intervals. The features from the SEMG and data glove signals were combined in a single feature pattern. Finally, the feature patterns were input to a trained ICA and GRA based classifier. The ICA and GRA based classifier is used to identify 20 hand motions and gestures recognition from 10 subjects.
The results of this study indicate that the ICA is a reliable pre-process to reduce the amount of data and, in turn, reduce the processing time about 30% in the GRA method, but the performance of the system decreased significantly following ICA in BPNN method. That maybe caused by difficult convergence in BPNN. The performance of the proposed HMI system using bend sensors and multichannel SEMG was based on accuracy, number of hand motions and training times needed to meet the requirement for real-time processing. The architecture of the proposed system meets the requirement of portability, but with the flexible mechanism, the proposed method can be further developed for implementation in a field programmable gate array (FPGA) or digital signal processor (DSP) based processor. It has highly potential to integrate as an assistive tool for many HMI applications..
Results
Conclusion
Fig. 1. The Structure of the proposed HMI system with ICA and GRA
ME
S
pre
-pro
cess
ing
Dat
a G
love
sign
al
pre
-pro
cess
ing
SE
MG
fe
ature
extr
acti
on
Ben
d S
enso
rs f
eatu
re e
xtr
acti
on
IEMG
WL
WAMP
SSC
ZC
VAR
Area
LC
MV
Han
d m
oti
on
cla
ssif
icat
ion
)()()()(
)()()()(
)2()2()2()2(
)1()1()1()1(
21
21
21
21
KKKK
kkkk
x
ni
ni
ni
ni
Comparison array
Real-time array
Euclidean distance
Comparison array
0
0.2
0.4
0.6
0.8
1
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2
3.01.0
0.15.0
0.50.3
Gre
y R
elat
iona
l G
rad
e (r
)
Euclidean Distance (ED)
ICA
ICA
Φ SEMG
Φ ICA
Φ Glove
Training Stage
Threshold Updated
IC number evaluation by PCA
Database
Data GloveMES
EMG s
ignals
from
three
electro
desDa
ta glov
e signa
ls fro
m Flex
sensor
Energ
y Dis
tributio
n
(1) (2) (3) (4) (5)
Ten subjects (average age, 24±3 years) were recruited to evaluate the performance of the proposed method. Each subject performed a set of 20 different gestures while standing. Figure 2 shows the 20 hand motions classified by the proposed.
Fig.2 Motion intervals of energy evaluation with acting five hand motions
Fig.3. Twenty hand motions are assigned in number order
Training Sets
0 1 2 3 4 5 6 7 8 9 10
CP
U t
ime i
n l
og
scale
(sec.)
0.01
0.1
10
100GRA GRA+ICA (Proposed method) BPNN BPNN+ICA
Fig.4 Average training times (N=10) versus the number of training sets for GRA and BPNN with/without ICA
