Prosthetics and Myoelectric Control

When a muscle contracts, it produces a tiny electrical signal that sensors on the skin can pick up, called EMG. A myoelectric prosthesis uses it to control a hand. The raw signal is noisy and spiky, so it must be rectified and smoothed into an envelope. More smoothing means a steadier hand, but it also means the hand reacts later.

Smoothing Threshold Sensor noise Control

From EMG to a grip command

The raw signal $x(t)$ swings positive and negative around zero. Taking the absolute value rectifies it, and a low-pass filter smooths it into an envelope $e(t)$ that tracks the effort:

$$e_t = \alpha\,e_{t-1} + (1 - \alpha)\,\lvert x_t\rvert, \qquad u = \begin{cases} \text{grip} \propto (e - e_{\text{th}}) & e > e_{\text{th}} \\ 0 & \text{otherwise} \end{cases}$$

What to notice

With light smoothing, the envelope jitters and the hand twitches. With heavy smoothing, it is steady but lags, so the hand closes noticeably after the user decides to grip, which feels sluggish. Raise the noise and the threshold must go up too, or the hand twitches on its own. Modern prostheses use several sensors and pattern recognition to choose among many grips, and some use targeted muscle reinnervation or implanted electrodes for clearer signals.