Page 150 - Control Theory in Biomedical Engineering

Modeling simple and complex handwriting based on EMG signals 135

shapes. On average, the coefficient of determination was 0.88 0.09 and
0.69 0.24 for the x and y coordinates, respectively.
Results showed that the model predicted the simple curve more accu-
rately. The prediction of complex curves such as multipeak has certain lim-
itations and it is easy to lose part of the information, which may be related to
the limited data of the training. In addition, it is almost impossible to accu-
rately predict the coordinates of specific x and y. The prediction of the
model can only reflect the shape of the trajectory. A follow-up investigation
is being conducted to improve the accuracy of the model.

5 Modeling of cursive writing from two EMG signals
In this section, we present two handwriting approaches enabling the model-
ing of some cursive Arabic letters and different simple and complex geomet-
ric forms. The first model calculates handwriting velocities from two EMG
signals. The second is considered a robust approach based on the parametric
model and interval observer allowing to overcome the problem of paramet-
ric variation. For this, we start by describing the experimental data acquisi-
tion used to develop the first and the second handwriting approaches.

5.1 Experimental approach and system presentation
The generation of handwriting movement, on a 2D plane, is mainly based on
two active forearm muscles activities—EMG1 and EMG2—relative to the
muscles abductor pollicis longus (APL) for vertical displacement and the exten-
sor capri ulnaris (ECU) for horizontal movements, respectively (Yasuhara,
1983; Manabu et al., 2003). In order to characterize handwriting movements,
an experiment was carried out in Hiroshima City University to measure simul-
taneously some Arabic letters and geometric forms and two forearm EMG sig-
nals relative to APL and ECU muscles (Manabu et al., 2003). Participants
wrote horizontal, vertical complexes (a combination of horizontal and vertical
movements) and also rapid and slow movements (Table 1). The synchroniza-
tion of the data recording was realized by sending a specific signal, a step, from
the parallel interface port on the computer to the data recorder.
The following equipment was used in this experimental approach:
• Digitalized table: WACOM, KT-0405-RN
• Preamplifiers: TEAC, AR-C2EMG1
• Data recorder: TEAC, DR-C2
• Bipolar surface electrodes: MEDICOTEST, Blue Sensor N-00-S
• Computer
Fig. 3 depicts the positions of the electrodes on the forearm of a writer.