Online Discrete Event Detection for Sensory Feedback Embedded in a Robotic Hand Prosthesis

Sensory feedback is deemed as a key aspect for understanding sensory perception, user acceptability and embodiment in upper limb prosthetics. However, despite the several attempts to demonstrate the benefits of sensory feedback in upper limb prosthetics, only few studies presented corroborating results in lab and daily life conditions. Technical challenges in the integration of sensory components in the hand hinder the implementation of home studies with research prostheses, and thus the assessment of the actual efficacy of sensory feedback. To contribute bridging this gap, in this work an algorithm was developed and assessed for the robust detection of salient tactile events in object manipulation. Starting from sensors embedded in a commercial hand prosthesis, an inertial compensation algorithm was designed to suppress force disturbances and enable reliable measurements. Then, a state-machine logic was implemented for the detection of the discrete tactile events of Touch, Lift Off, Replacement and Release. Both algorithms were developed to work in real-time and then ported in the low-power microcontroller hosted in the hand. The embedded algorithms were real-time tested showing effective suppression of disturbances up to 84%, and accurate event detection with error rate lower than 2%, paving the way for extensive tests of discrete feedback in unstructured domestic scenarios.