Page 234 - Control Theory in Biomedical Engineering
P. 234
Wearable mechatronic devices for upper-limb amputees 215
Muscle Reinnervation consists of biological signal amplification by means of
innervation of electrical nerves into new groups of surface muscles so that
surface electrodes can acquire and record the user movement intention.
4.4.1 EMG control strategies
Because a myoelectric prosthesis is considered a wearable robotic device,
control techniques for acquiring and processing EMG signals should be
taken into account.
To control upper-limb prostheses by means of the acquisition of EMG
signals, there are seven known control schemes well summarized by
Geethanjali (2016):
• ON-OFF control: As its name suggests, this is a control mode where the
electric motor in the terminal device is either ON or OFF. This action is
achieved by setting a threshold value, for which the processed informa-
tion from the EMG signal is usually compared with a Mean Absolute
Value (MAV) or root-square mean value.
• Proportional control: This scheme focuses on controlling the velocity of
the actuator, that is, motor velocity, as a function of the amplitude or
mean value of the acquired EMG signal. In other words, the speed of
the terminal device becomes proportional to the levels of EMG signals
(Mazumdar, 2004; Bottomley et al., 1963).
• Direct control: This belongs to the proportional scheme where a direct
control (Hahne et al., 2014) and communication between the incom-
plete electrical nerve and muscle control the exact part of the limb that
was amputated, for example, each finger has its terminal nerve where the
EMG is extracted and processed to be used as input signal to control it.
• Finite-State Machine control: This is a mode where some states are pre-
defined and programmed for some finite positions (Dosen et al., 2010).
• Pattern Recognition-based control, Regression and Posture control
schemes: These are modern techniques where signal classification,
regression analysis with a pre-processing of feature extraction, and esti-
mation using adaptive approaches are used (i.e., machine learning).
(Fougner et al., 2012; Muceli and Farina, 2011).
The technology used to develop affordable upper-limb prostheses is based
on the approaches previously discussed, from ON-OFF to myoelectric con-
trol strategies.
The functionality requirements of the prosthesis increase with the level
of amputation, which leads to a paradox seen in myoelectric control. The
functionality and therefore the control site requirements (i.e., EMG sensors’