In recent years, an electroencephalogram (EEG) signal
which is one of the biological signals is used as an input signal
for the various robots. An EEG signal is an electric signal that
can be measured along a scalp non-invasively. One of the
advantages of EEG based control of the robots is the potential
to be applicable to almost all people. Therefore, an EEG signal
is one of the strong candidates of the additional input signals
for wearable robots. The Interface between a robot and EEG
signals is called as Brain Computer Interface (BCI) [7]. In the
past, frequency analysis, event-related potential, evoked
potential, and so on were carried out. The purpose of these
analyses was not for the control of the robots in accordance
with the user’s motion intention, and those analyses are not
suitable for real-time control of the robots basically because
many of those analyses require a certain length of time-series
data of EEG signals and it was carried out in offline. In the
brain, the sum of electric signals which are generated at
multiple locations and are conveyed to the scalp is recorded as
the EEG signal. The measured EEG signal does not show a
one-to-one relationship with the corresponding brain part. It is
more difficult to estimate the user’s motion intention based on
the EEG signals for the control of the robots compared with the
EMG signals. Therefore, in general, the reliability of the
estimated user’s motion based on the EEG signals is lower than