emg gesture classifier.
classifying hand gestures from two-channel forearm EMG using classical ML.
surface EMG picks up the electrical activity of muscles through skin electrodes. with two channels on the forearm, different hand gestures produce recognizably different signal patterns, and the question this project asks is how far classical machine learning can go on that signal before anyone reaches for a neural network.
the pipeline: extract time-domain features from each channel, then classify with an SVM and a random forest and compare. classical models are small enough to run on a microcontroller and easy to inspect when they get a gesture wrong.
full write-up coming as the prototype matures.