Skip to main content
. 2023 Apr 6;10:1100411. doi: 10.3389/frobt.2023.1100411

FIGURE 2.

FIGURE 2

Diagram of implemented control methods, adapted from Lobo-Prat et al. (Lobo-Prat et al., 2017a). Bold font style symbols indicate vectors and regular font style symbols indicate scalars. The upper section represents the physiological system (participant), while the lower section represents the experimental system. To perform a movement participant first see the target on the screen (P tarx,y). The target is generated by a python script running in the host computer. This generates a neural command (Cnrl) with their central nervous system, which results in muscle activation at forearm muscles where sEMG signals (E sen) are measured. Intention of the user is decoded from these sEMG signals. In direct control the sEMG signals (E sendc) are measured from agonist/antagonist muscle pair from forearm (Flexor Carpi Ulnaris/Extensor Carpi Ulnaris) and the resting sEMG (E res) is subtracted to acquire the voluntary sEMG (E voldc). The signal is normalized to the maximum voluntary contraction (MVC) and control signals are generated from each muscle (Ue,Uf). A voluntary control signal (Uvoldc) is obtained by subtracting the control signal of the flexor muscle from that of the extensor muscle (reverse for left handed participants). A co-contraction switch, was used to alternate DOF. In case of the DMD participant, an electrode in the gastrocnemius was used to switch (blue line). In pattern recognition control sEMG signals (E senpr) are measured from six electrodes placed on forearm (hexagonal grid). Time domain features (FE) were extracted from measured sEMG signals and these features were then used by ANN classifier to identify the movement class (Acls). This class is then used to select the final control signal (U volx,ypr). This control signal (Uvolpr) is the normalized mean envelope of the six electrodes. In both control methods the estimated voluntary forces (Festx,y) are used as input to a first order admittance model (Hadm) that resembles the dynamics of a mass-damper system. The resulting velocity of the cursor ( P˙ curx,y) is send to an integrator (P curx,y) and then to the windows PC to control the position of the cursor on the screen. This motion was sensed by the participants proprioception and by visual feedback and was used to generate new neural commands to reach new target positions (P tarx,y).