Slip detection using the power spectrum of sensory nerve recordings

Abstract

In cases of spinal cord injury primary sensory afferents below the injury remain fully functional and could be used as natural sensors to provide a feedback signal for FNS control. This paper presents a slip detection method using the power spectrum of spiral cuff ENG. The central toe pad of a cat is used as a model for human glabrous skin. Three modes of tactile stimuli (slip, stretch and pressure) are applied to the central toe pad in an anesthetized preparation while recording the electroneurogram (ENG) from the tibial nerve using a three band spiral cuff electrode. Discriminant analysis of the power spectra of 10 ms data segments was used to classify randomly selected segments as "slip" or "non-slip". The highest classification rate (81% correct) was achieved by normalizing the power spectra by the total power of the segment. Three window sizes (10 ms, 20 ms and 40 ms) were compared and found to have no effect on classification rate. The classification rate appeared to correlate well with the velocity of the slipping object, ranging from 55% to 90% for velocities of 2.5 mm/s to 15 mm/s. In conclusion, the power spectrum can be used to distinguish slip from other types of tactile stimulus and higher velocities are detected more reliably.