TinyML Unknown Gestures: Use an Anomaly Score Without Mistaking It for Class Confidence
How can a microcontroller gesture classifier assign a confident known label to an unfamiliar movement, and what does Edge Impulse K-means add to the accept/reject decision?