Neural Network Human Pose Recovery from Partial Joints
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional techniques for recovering a human body model in medical settings often rely on comprehensive joint location information, but are limited in accurately determining both pose and shape due to partial obstruction by equipment or clothing, necessitating a method to infer these parameters from limited data.
Innovation Solution
An artificial neural network is trained to predict pose and shape parameters from a subset of joint locations, using both 2D and 3D joint information, and adjusts its parameters based on differences between inferred and actual joint locations, enabling recovery of human models despite partial knowledge.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional techniques use comprehensive joint location information to recover human body models, then measurement precision is improved, but the system becomes vulnerable to obstruction by medical equipment and clothing that block joint locations
Solution Approach 1:
The system recovers complete pose and shape parameters from a partial subset of observable joint locations. Instead of requiring all joint locations to be visible, the neural network is trained to infer the complete human body model from the limited available joint data, effectively performing partial action (using only visible joints) to achieve the complete goal (full pose and shape recovery).
Solution Approach 2:
A neural network acts as an intermediary between the partial joint location observations and the complete human body model. The network learns to map from incomplete joint data to comprehensive pose and shape parameters, serving as a mediator that bridges the gap between limited measurements and full reconstruction.
2Adaptability or versatility
If the system recovers both pose and shape from limited joint locations, then adaptability to medical settings is improved, but measurement precision may deteriorate due to incomplete data
Solution Approach 1:
The neural network is pre-trained using comprehensive joint location data from training datasets before deployment. This preliminary training establishes the network's ability to recognize patterns and relationships between joint locations and body parameters. When deployed in medical settings with obstructions, the pre-trained network can generalize from the training data to accurately recover pose and shape even from limited observations.
Solution Approach 2:
The system uses a feedback mechanism where the recovered pose and shape parameters are used to infer expected joint locations, which are then compared with the actual observed joint locations. This feedback loop allows the system to iteratively refine its predictions and adjust its estimates to better match the observed partial data, improving measurement precision despite incomplete inputs.
Data Source
AI summary
The pose and shape of a human body may be recovered based on joint location information associated with the human body. The joint location information may be derived based on an image of the human body or from an output of a human motion capture system. The recovery of the pose and shape of the human body may be performed by a computer-implemented artificial neural network (ANN) trained to perform the recovery task using training datasets that include paired joint location information and human model parameters. The training of the ANN may be conducted in accordance with multiple constraints designed to improve the accuracy of the recovery and by artificially manipulating the training data so that the ANN can learn to recover the pose and shape of the human body even with partially observed joint locations.