Low Dimension Pose Space for Head Rotation Estimation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for estimating head rotation angles in human-computer interactions are limited by coarse and discrete pose estimations, computational inefficiency, and difficulty in handling varying lighting conditions.
Innovation Solution
A method is developed to generate a low-dimensional pose space using local binary patterns (LBP) and principal component analysis (PCA) followed by linear discriminant analysis (LDA), allowing for efficient estimation of head rotation angles by projecting user image frames into a trained pose space.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If a classifier is trained to distinguish between a small number of discrete poses, then the pose estimation is simplified and computationally efficient, but the pose estimation accuracy and granularity are limited to pre-defined discrete poses
Solution Approach 1:
The patent transforms the pose estimation problem from discrete classification to continuous angle estimation by introducing a new dimensional approach. Instead of classifying poses into predefined discrete categories, the system estimates continuous rotation angles (azimuth and elevation) by detecting facial landmarks and computing their spatial relationships, thereby achieving both computational efficiency and continuous precision.
Solution Approach 2:
The patent changes the parameter representation from discrete pose categories to continuous angular parameters. By detecting facial landmarks (eyes, nose, mouth corners) and computing angles based on their coordinates, the system transitions from categorical classification to continuous parameter estimation, improving measurement precision while maintaining computational efficiency through geometric calculations.
2Measurement precision
If certain facial points are detected and used to estimate head pose, then the pose estimation accuracy is improved, but the computational demand and initialization requirements increase significantly
Solution Approach 1:
The patent extracts only the essential facial landmarks needed for pose estimation (eyes, nose, mouth corners) rather than processing the entire face image or all facial features. This selective extraction of critical points reduces computational complexity while maintaining sufficient accuracy for head pose estimation, avoiding the need to compute information beyond what is relevant for the specific task.
3Adaptability or versatility
If a large number of training image frames are used, then the model robustness under varying conditions is improved, but the training time and computational resources required increase
Solution Approach 1:
The patent performs preliminary action by pre-detecting and storing facial landmark coordinates from training images before the actual pose estimation task. By extracting and caching the essential facial point information in advance, the system reduces the computational burden during real-time operation and enables faster processing while maintaining robustness through diverse training data coverage.
Data Source
AI summary
Methods are provided for generating a low dimension pose space and using the pose space to estimate one or more head rotation angles of a user head. In one example, training image frames including a test subject head are captured under a plurality of conditions. For each frame an actual head rotation angle about a rotation axis is recorded. In each frame a face image is detected and converted to an LBP feature vector. Using principal component analysis a PCA feature vector is generated. Pose classes related to rotation angles about a rotation axis are defined. The PCA feature vectors are grouped into a pose class that corresponds to the actual rotation angle associated with the PCA feature vector. Linear discriminant analysis is applied to the pose classes to generate the low dimension pose space.