Hand Pose Estimation Using Adaptive Recursive Least Squares
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional hand tracking systems in artificial reality systems face challenges in accurately estimating the pose of a user's hand due to sensor nonlinearity, drift, and noisy ground truth data, especially when the instrumented glove shifts on the hand.
Innovation Solution
A wearable tracking sensor integration system with a plurality of sensors, including elastomeric stretch sensors, locators, and an inertial measurement unit, is used to collect data and apply an adaptive weighted recursive least squares estimation algorithm to determine the estimated pose of the user's hand, which is then integrated with an artificial reality system to update the visual presentation of a virtual hand.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If optical hand tracking systems are used in artificial reality systems, then the system can capture hand motion and position, but the pose estimation accuracy deteriorates due to sensor nonlinearity, drift, and noisy ground truth data
Solution Approach 1:
The patent combines multiple sensors (optical tracking sensors, inertial measurement units, and stretch sensors) into an integrated hand tracking system. By merging data from these different sensor types, the system compensates for individual sensor limitations such as optical drift and nonlinearity, thereby improving overall pose estimation accuracy while maintaining reliable measurements
Solution Approach 2:
The system implements a calibration framework that uses feedback from the sensors to continuously adjust and correct pose estimates. The feedback mechanism detects drift and noise in real-time and applies corrections to maintain accurate hand pose estimation, resolving the contradiction between measurement precision and sensor reliability
2Quantity of substance
If the instrumented glove is used to attach multiple sensors, then comprehensive sensor data can be collected, but the glove shifts on the hand causing measurement errors
Solution Approach 1:
The patent employs dynamic calibration parameters that adapt in real-time as the glove moves or shifts on the hand. The system continuously updates calibration data to account for changes in glove position, ensuring that comprehensive sensor data collection does not compromise measurement precision even when the glove shifts during use
3Measurement precision
If calibration is performed before applying estimation algorithms, then accurate pose estimation can be achieved, but the process becomes complex due to sensor nonlinearity and drift
Solution Approach 1:
The system performs preliminary calibration actions by collecting calibration data from all sensors before normal operation begins. This preliminary calibration establishes baseline parameters that simplify subsequent pose estimation, reducing the complexity of real-time processing while maintaining high measurement precision through pre-computed calibration parameters
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
The system provides accurate and adaptive estimation of the user's hand pose, reducing errors caused by sensor drift and noise, and dynamically updates the estimation parameters to accommodate changes in hand position, thereby enhancing the accuracy of virtual hand tracking in artificial reality environments.
Implementation Method 1
a plurality of sensors including elastomeric stretch sensors
Implementation Method 2
an inertial measurement unit
Data Source
AI summary
A tracking sensor integration system presented herein collects sensor data obtained for each time frame by a plurality of sensors attached to a wearable garment placed on a user's hand. A controller coupled to the tracking sensor integration system calculates a measurement gain based at least in part on collected sensor data, and determines prediction for a pose of the user's hand for the current time frame using the collected sensor data and a plurality of estimation parameters for the current time frame. The controller then updates the estimation parameters for the current time frame, based in part on the measurement gain and the prediction for the pose of the user's hand. The controller determines an estimated pose for the user's hand, based in part on the updated estimation parameters and the collected sensor data.