Interacting with Soli: Exploring Fine-Grained Dynamic Gesture Recognition in the Radio-Frequency ...
summary
Prototype uses 60 GHz Soli radar to produce range-Doppler frames and trains an end-to-end CNN+LSTM to classify 11 dynamic gestures from a collected dataset of 10 users. The model runs on commodity CPU at 140 Hz and achieves 87% avg accuracy across gestures, resolving fine finger motions within roughly 20 cm. It needs gesture segmentation for lowest-latency output and struggles with unconstrained background motion, motivating larger datasets and continuous tracking.
AI-written summary · about
chapters (7)
- 0:00Introduction and motivation
- 2:45Soli sensor fundamentals
- 4:20Signal processing methodology
- 7:08Neural network architecture
- 9:53System pipeline and evaluation
- 12:20Temporal analysis results
- 14:08Conclusion and Q&A
related talks
06mPantomime: Mid-Air Gesture Recognition with Sparse Millimeter-Wave Radar Point Clouds
06mReal-time Arm Gesture Recognition in Smart Home Scenarios via Millimeter Wave Sensing
06mCrossGR: Accurate and Low-cost Cross-target Gesture Recognition Using Wi-Fi
28mStrike a Pose - Gesture recognition in JavaScript w/ ML & Arduino
06mTowards Position-Independent Sensing for Gesture Recognition with Wi-Fi
32mIntegrating CNN Models with Flutter Using TensorFlow Lite
08mDesigning a Personalised Sensor Glove Using Deep-Learning
22mStrike a Pose: Gesture Recognition in JS with Machine Learning