Wearable robots such as a power-assist robot or an artificial
limb are expected to assist activities of a user especially an
elderly or a person who has a physical handicap who have
difficulty in daily living independently. To assist the daily
living motion of the user, those robots try to estimate a motion
intention of the user. An electromyogram (EMG) signal is one
of the widely used biological signals to control those robots
based on the user’s motion intention. Many kinds of EMGbased
control methods have been proposed up to the present [1-
6]. An EMG signal is an electric signal which is generated
when a muscle is activated, and especially, surface EMG
signals are able to be measured non-invasively from electrodes
which are attached on the skin. The measured EMG signal
shows almost one-to-one relationship with the corresponding
muscle as long as the electrode is located correctly. Therefore,
each joint motion can be estimated relatively easily based on
the EMG signals, and the robots assist a user’s motion in realtime
based on the estimated motion.