Articulated Keypoint Jitter Filtering Using Physical Constraints
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
XR systems face power consumption challenges due to high processing demands for tracking six degrees of freedom, leading to non-portable devices that require external power sources, and jitter in keypoint detection causes errors in object tracking.
Innovation Solution
Implement jitter estimation using physical constraints by dynamically detecting and filtering jitter based on expected lengths between keypoints of an articulated object, applying a filter to correct keypoint locations.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If powerful processors are used to perform feature analysis and tracking functions quickly, then object tracking accuracy and speed are improved, but power consumption increases
Solution Approach 1:
The patent applies partial smoothing only to keypoints that exhibit jitter based on physical constraints, rather than smoothing all keypoints uniformly. The system calculates expected lengths between keypoints and only applies filtering when deviations exceed thresholds, thereby reducing unnecessary processing and power consumption while maintaining tracking accuracy for affected keypoints.
Solution Approach 2:
The system dynamically adjusts smoothing parameters based on physical constraints of the tracked object. By changing the smoothing intensity parameter adaptively - applying stronger smoothing only when physical constraints are violated - the system optimizes the balance between tracking accuracy and processing efficiency, reducing power consumption compared to constant high-intensity smoothing.
2Stability of the object's composition
If smoothing is applied to reduce jitter in keypoint detection, then tracking stability is improved, but actual object motion may be distorted
Solution Approach 1:
The patent implements selective smoothing by evaluating physical constraints for each keypoint pair. Smoothing is applied only when the distance between keypoints violates expected physical constraints, leaving other keypoints unchanged. This partial application of smoothing maintains tracking stability for jittery keypoints while preserving accurate motion information for keypoints that conform to physical constraints.
Solution Approach 2:
The system uses feedback from physical constraint validation to control smoothing application. By continuously monitoring whether keypoint distances satisfy expected physical relationships and adjusting smoothing intensity accordingly, the system maintains tracking stability only when necessary, preventing distortion of actual object motion while reducing jitter where it occurs.
3Reliability
If extensive filtering is applied to correct keypoint locations, then jitter is reduced, but processing time increases
Solution Approach 1:
The patent applies filtering selectively only to keypoint pairs that violate physical constraints, rather than processing all keypoints uniformly. By calculating expected lengths and comparing against actual measurements, the system identifies only the subset of keypoints requiring correction, significantly reducing processing time while maintaining detection accuracy for problematic keypoints.
Solution Approach 2:
The system dynamically changes processing parameters based on constraint violation detection. When physical constraints are satisfied, minimal or no filtering is applied. When constraints are violated, filtering intensity increases only for those specific keypoints. This adaptive parameter adjustment reduces overall processing time while maintaining reliability where needed.
Data Source
AI summary
Techniques and systems are provided for pose prediction. For instance, a process can include a process for jitter correction. The process includes receiving a set of keypoints for an articulated object, wherein keypoints of the set of keypoints correspond to joints of the articulated object; determining a first keypoint of the set of keypoints is associated with a second keypoint of the set of keypoints; determining there is jitter associated with the first keypoint based on a comparison of a length between the first keypoint and the second keypoint to an expected length; and applying a filter to a location of the first keypoint based on the determination that there is jitter associated with the first keypoint.