A Low-Power Low-Complexity Circuit for Event-Based Feature Extraction from sEMG

This paper presents an analog circuit for calibration-free event-driven myoelectric control of sEMG-based applications. The proposed solution is to be installed downstream of the conditioning chain of an sEMG sensor and consists of a Sallen-Key filter, acting as a differentiator in the main sEMG frequency band, and a voltage comparator. The output of the circuit is a quasi-digital signal, in which the muscle activity is mapped onto the time distribution of digital events. The design phase focused on noise robustness, and a prototype was tested during in-vivo experiments on both upper and lower limbs. Among the obtained results, besides a current consumption of only 12.92 mu A, a median increase in the number of events of more than 25% was achieved by varying the exerted muscle force in steps of 20% MVC.