UltraPoser: Pushing the Limits of IMU-based Full-Body Pose Estimation with Ultrasound Sensing on ...
summary
UltraPoser fuses IMU data from a smartphone smartwatch and earbud with ultrasound sensing using built-in speakers and microphones to extend coverage and remove IMU drift. It extracts Doppler spectrum and channel impulse response via narrowband single-tone and odd-even frequency-division multiplexed signals across a 4 kHz band, encodes modalities separately, uses cross-attention and a graph convolutional kinematic model for fusion, and evaluates on a 10-person 300+ frame dataset showing 28.46% improvement overall and up to 67.28% limb error reduction.
AI-written summary · about
chapters (9)
- 0:00Introduction and motivation
- 1:34Limitations of IMU-only pose estimation
- 2:38UltraPose key idea and ultrasound sensing
- 4:03Design challenges and solutions
- 5:51Sensor setup and hardware details
- 6:54Graph-based multimodal fusion framework
- 7:38Evaluation results and demo
- 9:03Summary and code release
- 9:29Q&A session
related talks
10mProactive Sensing for Improving Hand Pose Estimation
06mIMUTube: Automatic extraction of virtual on-body accelerometry from video for human activity ...
06mRobust Inertial Motion Tracking through Deep Sensor Fusion across Smart Earbuds and Smartphone
06mApproaching the Real-World: Supporting Activity Recognition Training with Virtual IMU Data
20mAdvancing Hand Gesture Recognition with High Resolution Electrical Impedance Tomography
23mCounting exercise repetition using live camera video
14mBack-Hand-Pose: 3D Hand Pose Estimation for a Wrist-worn Camera via Dorsum Deformation Network
06mWR-Hand: Wearable Armband can Track User's Hand
