Summary: | 博士 === 國立臺灣大學 === 電機工程學研究所 === 83 === Characteristics of the surface electromyograpgy(EMG) were
studied with different electrode arrangements, features and
classifiers. Electrode pairs were located separately on dominant
muscles(S-type arrangement) and closely in the region between
muscles(C-type arrangement). Identification of the EMG obtained
from the sternocleidomastoid and the upper trapezius muscles
during ten motions of the head and shoulder showed that S-type
arrangement together with the cepstral method achieved the best
classification.
A simplified model was proposed to elucidate the performances
of the AR and cepstral coefficients with the two types of
electrode arrangement. As bandwidth decreased, the performance
of the cepstral method degraded more seriously than the AR
method. As high-frequency noise increased, the performance of
of the AR method improved faster than the cepstral method. The
application of the cepstral method to the C-type signals was
less effective than to the S-type signals.
The Euclidean distance measure(EDM), weighted distance measure
(WDM), and maximum likelihood method(MLM) were used to classify
the AR and cepstral parameters with the S-type arrangement. The
cepstral method achieved significant improvements over the AR
method for all the three classifiers. The cepstral coefficients
owned better cluster separability in feature space and they
emphasized the more informative part in the frequency domain.
The discrimination rate of the MLM was the highest among the
three classifiers. Proper choice of five of ten motions could
further raise the recognition rate to more than 95%.
In conclusion, the assembly of the S-type electrode arrange-
ment, the cepstral coefficients, and the MLM was recommended for
EMG pattern recognition.
|