Continuous hand pose tracking with wrist-worn antenna impedance characteristic sensing

Inventors

KIM, Daehwa • Szini, Istvan J. • Harrison, Christopher

Abstract

Embodiments are disclosed for a continuous hand pose tracking system employing at least one wrist-worn antenna, from which real-time dielectric loading resulting from different hand poses). The sensor data is interpreted by a machine learning backend, which outputs a fully-posed three-dimensional (3D) hand that can be continuously tracked. In some embodiments, two degrees of freedom (2DOF) wrist angle and micro-gestures are tracked. The hand pose tracking system can be extended to include two or more and/or different types of antennas operating at different self-resonances. In an embodiment, a method comprises: determining, with at least one processor of a wrist-worn device, a complex impedance characteristic variation based on a dynamic finite electric ground plane of at least one antenna coupled to the device; and predicting, with the at least one processor, a hand pose of a user wearing the device on a their wrist based on the determined complex impedance characteristic variation.

Core Innovation

The invention is a method and apparatus for continuous hand pose tracking using wrist-worn antenna loading mode sensing based on a dynamic finite electric ground plane of at least one antenna coupled to a device. A user's hand changes the antenna finite electric ground plane, shifting a self-resonance frequency and thereby causing a complex impedance characteristic variation. The system determines the complex impedance characteristic variation associated with the dynamic finite electric ground plane.

The complex impedance characteristic variation is determined from frequency-dependent antenna measurements, including return loss magnitude and phase shift and/or complex impedance magnitude and phase shift. Measurements taken as a function of frequency are converted into impedance-variation features that capture variations in the determined impedance characteristics. Such impedance-variation features are derived using return loss magnitude and complex impedance magnitude/phase, including aggregation and statistical measures.

Based on the determined complex impedance characteristic variation, the system predicts a hand pose of a user wearing the device on their wrist. The predicted hand pose includes fully-posed 3D hand pose and a 2DOF wrist angle, including micro-gestures, and is described as robust to occlusion by fabric. Prediction is performed using a machine learning backend, such as neural networks or ExtraTrees, using an input feature vector derived from the impedance characteristics.

Claims Coverage

The independent claims cover three implementations of the same inventive concept: a method, an apparatus, and a non-transitory computer-readable storage medium. Across these independent claims, the core coverage includes determining a complex impedance characteristic variation from a dynamic finite electric ground plane and predicting a hand pose based on that determined variation.

Collectively, the independent claims focus on impedance-based sensing driven by a dynamic finite electric ground plane and subsequent hand pose prediction from the determined complex impedance characteristic variation, implemented as a method, an apparatus, and computer-readable instructions.

Stated Advantages

Documented Applications

No documented applications found