Psychophysical principles of discrete event-driven vibrotactile feedback for prostheses

Abstract

Purpose/aim of the study: We aimed to establish psychophysical principles for non-invasive vibrotactile feedback signalling discrete transition events (e.g., extension to flexion) during use of prostheses, especially for the upper limbs.

Materials and methods: Two vibrotactile actuators were used on both upper arms of 10 able-bodied human participants. Absolute thresholds, psychometric functions, and magnitude estimates were measured to equalize the sensation magnitudes for the tested vibrotactile frequencies and skin sites. Then, same-different and pattern recognition tasks were run to evaluate, respectfully, the discrimination and closed-set identification of stimuli with varying parameters (2 frequencies, 2 magnitudes, 2 sites). Finally, parameters of the left/right stimuli were mapped to hypothetical prosthesis events representing object/force and movement type. The stimuli were applied sequentially in accordance with the discrete event-driven feedback paradigm.

Results: Reliable psychophysical models could be established for individual participants as verified by repetitive threshold measurements and relative adjustment of stimulus levels based on sensation magnitudes. Discrimination accuracy was higher for magnitude versus frequency comparisons; and magnitude discrimination accuracy was correlated with magnitude estimate differences. Pattern recognition recall/precision rates decreased from ～0.7 to ～0.5 for sequential delivery of two stimulus patterns to one arm versus to two arms. Using the patterns as two and three consecutive prosthesis events yielded statistically similar performance rates not correlated with magnitude estimate differences.

Conclusions: By careful calibration of stimuli based on psychophysical principles, discrete event-driven vibrotactile feedback can be used to signal manipulated object and movement information with moderate identification rates as shown by confusion matrices.