The hand prosthesis that learns from the user
summary
A machine‑learning myoelectric prosthesis from Otto Bach records eight EMG channels around the forearm, extracts simple time‑domain features, and trains on a few seconds of labeled gestures so the device adapts to the user rather than the user learning a state machine. The embedded controller runs feature extraction and a linear discriminant classifier in under 2 ms latency, draws about 17–18 mA, and uses LDA on mean/zero‑crossing/wavelength features to deliver simultaneous multi‑degree movements.
AI-written summary · about
chapters (12)
- 0:00Introduction
- 3:08Current prosthesis
- 5:41The problem
- 12:00Demonstration
- 14:48How it works
- 23:09QA
- 25:20Stress
- 27:26Arm amputation
- 28:27Leg amputation
- 29:27Cost
- 30:23User experience
- 32:01proprioceptive feedback
related talks
03mPowered by TensorFlow: Creating a custom, machine learning-powered drumming arm
26mBiosensor Machine Learning with Julia
09mBody Integrated Programmable Joints Interface
10mAugmenting Human Capabilities to New Dimensions
41mCreating Intelligent Music Instruments with Machine Learning
19mBackHand: Sensing Hand Gestures via Back of the Hand
20mAdvancing Hand Gesture Recognition with High Resolution Electrical Impedance Tomography
17mOpen Source Robotics: Advances in Teleoperation for Dexterous Manipulation