核心概念
High-fidelity acoustic sensing system enables continuous 3D hand pose tracking on home assistant devices.
统计
Beyond-Voice can track joints with an average mean absolute error of 16.47mm without personalized training data.
User-adaptive evaluation reduces the mean absolute error to 10.36mm.
引用
"Our system leverages the existing acoustic sensors of commercial home assistant devices to enable continuous fine-grained hand tracking of a subject."
"The key idea is to transform the device into an active sonar system and analyze the reflections to reconstruct a 3D hand continuously."