Designing a Personalised Sensor Glove Using Deep-Learning
summary
A sensor glove with IMUs, flex sensors, pressure, gyros and magnetometers records 31 gestures from 22 participants and trains a neural network augmented with an importance-estimate layer that reweights input channels. Training is two-phase (freeze model then freeze weights) and a cost-aware variant adds a second network to bias selection. Removing 20% or 40% of sensors shows little drop until roughly 40% removal; performance falls thereafter. The paper compares personalized vs generic selection and validates retained sensors by true positive rates per gesture.
AI-written summary · about
chapters (6)
- 0:00Start
- 0:47Problem: Gesture Recognition
- 3:58Importance Estimate
- 6:00Validation & Verification
- 6:53Conclusion
- 7:33References
related talks
19mBackHand: Sensing Hand Gestures via Back of the Hand
17mEMPress: Practical Hand Gesture Classification with Wrist-Mounted EMG and Pressure Sensing
05mA Feature Adaptive Learning Method for High-Density sEMG-Based Gesture Recognition
26mBiosensor Machine Learning with Julia
14mLeveraging Dual-Observable Input for Fine-Grained Thumb Interaction Using Forearm EMG
17mPredicTouch: A System to Reduce Touchscreen Latency using Neural Networks and Inertial ...
28mStrike a Pose - Gesture recognition in JavaScript w/ ML & Arduino
20mAdvancing Hand Gesture Recognition with High Resolution Electrical Impedance Tomography