Concepts de base
Developing a high-fidelity acoustic sensing system for continuous 3D hand pose tracking on home assistant devices.
Résumé
The content discusses the development of Beyond-Voice, a system that enables continuous hand pose tracking using acoustic sensors on home assistant devices. It outlines the challenges faced in traditional voice user interfaces and proposes a novel method to improve accessibility and usability. The system utilizes deep learning models to analyze acoustic signals and predict the 3D positions of finger joints. Various experiments and tests are conducted to evaluate the system's performance across different users and environments.
- Introduction to the challenges with current voice user interfaces on home assistant devices.
- Proposal of Beyond-Voice as a solution for continuous hand pose tracking.
- Explanation of the system's operation using acoustic sensors and deep learning models.
- Details on data preprocessing techniques and model training strategies.
- Results from user studies evaluating system performance across different scenarios.
Stats
"A user study with 11 participants in 3 different environments shows that Beyond-Voice can track joints with an average mean absolute error of 16.47mm without any training data provided by the testing subject."
"The MAE can further decrease to 10.36mm in a user-adaptive evaluation."
"In a user-dependent test, the MAE is around 12.49mm."