whereRU: Aresty Poster 105 - A Method for the Elimination of Redundant Sensors by Aresty Posters 2009

A Method for the Elimination of Redundant Sensors from an Array

Reliable prosthetic control requires that neuromuscular volition can be accurately reconstructed from an array of interfacial sensors. However, it has not yet been determined whether a redundant sensor design is necessary for biomimetic device control in isometric grasping tasks. In the present study, we characterize the relationship between sensor array size and signal fidelity in surface muscular pressure (SMP), a non-invasive imaging modality detecting the radial expansion of skeletal muscles during voluntary concentric contractions via a sleeve array of pressure sensors. While the SMP has demonstrated highly accurate reconstruction of forearm signatures using 6 sensors, the minimally sufficient sensor set has not yet been established. To determine the lower limit of sensor subset accuracy, all combinations of SMP sensors were compared against the reference signal of all six sensors. A single male subject 34 years old, and with no known neurological deficit, performed a repetitive grasping task, targeting gradated grip force according to a computer prompt. To quantify signal fidelity to this standard, a comparison was made via Pearson?s correlation coefficient ?, yielding the signal reconstruction error E=1?|?| for each subset. 95.2% of sensor subsets yielded E<0.3, including five of the six single-sensor subsets. We conclude that a single sensor can represent grasp action to within an acceptable tolerance. Future work will assess whether more complex grasping tasks require multiple sensors.

Presenters: Laith Qumei (Lqumei@eden.rutgers.edu ), Shubhagata Deb Roy (sdroy@eden.rutgers.edu ), Nasir Uddin (nasirudd@eden.rutgers.edu)

Advisor: William Craelius (craelius@rci.rutgers.edu)

RU Rehab Lab (rurehablab.org )