Non-invasive EMG interfaces could be able to resolve many issues that affect people suffering from disabilities or require a more efficient method to complete an activity. This is why one manufacturer known under the name “Hut,” to create their own open-source device known as PsyLink. It operates by analyzing the small electrical impulses that trigger muscles to contract and sending them to be processed and inferring using an algorithm for machine learning.
The first prototype from PsyLink was designed on its Nano 33 BLE Sense due to its numerous ADC pins, and the possibility of Bluetooth connectivity. It has an aluminum foil pair of pads connected to wires. However, the pads were later substituted for studs encased in an insulated sleeve that was more secure. Signals are detected by the electrodes, and then sent through a set of filters based on the op-amps before being sent into an analog multixer. The signal is digitalized through the internal ADC and then sent over Bluetooth Low Energy where it can be displayed in an application for desktops that is customized.
Hut utilized TensorFlow to collect a variety of data and build the neural network to identify the moment a particular type of signal corresponds to a particular keypress. After training was completed and the model was able to be employed to accomplish anything from typing faster to doing shortcuts, to playing video games.