Wrist Force Myography (FMG) Exploitation for Finger Signs Distinguishing

The Force Myography (FMG) is an non-invasive technique where force sensitive resistors (FSRs) are used on the surface of the skin to detect the volumetric variations in the underlying muscles and tendons complex. Recent works have proposed various FMG systems for gesture recognition with a big number of sensors or combined systems with other sensors such as electromyography to identify gestures with objects interaction or force level variation. This paper propose two FMG detection systems with minimal number of FSR sensors (four and eight) for American sign language recognition based on raw FMG with implementation of Extreme learning machine (ELM) for evaluating the accuracy of nine ALS alphabet recognition. The first feasibility test for ALS sign detection with FMG systems was tested with one subject with an ELM accuracy of 78% with four sensors and 97.90 % with eight and the minimal efficient sensor number was preliminary investigated in the second band. Other nine subjects tested the sight sensor band which resulted an ELM accuracy of 83,30% for identification of nine ALS signs from 10 subjects while the SVM resulted an accuracy of 64,9 with the same database.