Continuous Human Posture Estimation with Joint Motion Tracking
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing deep neural network-based human posture estimation methods require high computational resources, are slow, and lack continuity between video frames, limiting real-time applications and accuracy.
Innovation Solution
Integrate a pretrained deep neural network posture estimator with a block matching motion estimation algorithm, utilizing a video encoding standard like H.264, to achieve continuous human posture estimation by combining the strengths of both methods.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a deep neural network is used for human posture estimation, then estimation accuracy is improved, but computing resource requirements increase and real-time performance deteriorates
Solution Approach 1:
The patent segments the posture estimation task into two independent parts: a deep neural network processor that handles accuracy-critical tasks and a motion estimation processor that handles real-time continuity tasks. This segmentation allows each component to optimize for its specific function without compromising the other, resolving the contradiction between accuracy and real-time performance.
Solution Approach 2:
The patent introduces an intermediary mechanism that uses motion estimation results from the previous frame as input to the current frame's deep neural network processor. This intermediary connection ensures temporal continuity and reduces the computational burden on the deep neural network, enabling real-time processing while maintaining high accuracy.
2Device complexity
If deep neural network processes each video frame independently, then calculation simplicity is maintained, but continuity and smoothness of posture estimation deteriorates
Solution Approach 1:
The patent applies preliminary action by using the motion estimation results from the previous frame as pre-computed input for the current frame's deep neural network processor. This preliminary preparation of motion information ensures that the deep neural network can focus on accuracy while the overall system maintains temporal continuity and smoothness.
3Productivity
If a block matching algorithm is used for motion estimation, then real-time performance is improved, but estimation accuracy for complex poses deteriorates
Solution Approach 1:
The patent segments the estimation task by using block matching only for motion continuity where it excels (real-time performance), while reserving deep neural network processing for accuracy-critical pose estimation. This selective segmentation allows the system to achieve both real-time performance and high accuracy without relying on the limitations of either algorithm alone.
Data Source
AI summary
The present invention discloses a key joint motion estimation based method for estimating continuous human postures. A motion estimation block matching algorithm is applied to human key joint tracking, so as to obtain continuous human posture results. Meanwhile, the results are continuously corrected by using a deep neural network based human posture estimator. The present invention may estimate the continuous human postures in a video stream, where the human postures are specifically embodied as coordinate positions of human joints in a video frame. Compared with a posture estimation method completely relying on a deep neural network, the posture estimation method provided by the present invention has the advantages of high frame rate, low hardware requirements, and sequential continuity of recognition results; and compared with a posture estimation method completely relying on a motion estimation algorithm, the present invention may correct a cumulative error, to improve the estimation accuracy.