Skeleton Key Point Compensation for Human Rotation Pose Estimation

Solution Overview

Problem

Existing AI engines for estimating skeleton key points in human motion analysis are inaccurate for rotation motions due to insufficient training data, leading to low pose state recognition accuracy.

Innovation Solution

A processing apparatus and method that utilize three-dimensional or two-dimensional skeleton key point coordinate estimation values and a gravity direction vector to calculate compensated coordinate values based on silhouette widths and anatomical knowledge, correcting the estimated key points for improved accuracy.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If the AI engine is trained with more data or the learning model is reconstructed to improve skeleton key point estimation accuracy for rotation motions, then the estimation accuracy is improved, but the device complexity and training time increase

Engineering Contradiction:

Improveskeleton key point estimation accuracyVSAvoidlearning model complexity

Core Design Contradiction:

Measurement precisionVSDevice complexity

Solution Approach 1:

The patent introduces an intermediary compensation processing unit that receives the estimated skeleton key points and corrects them using pre-calculated compensation amounts based on rotation angle. This mediator resolves the contradiction by providing accurate correction without requiring the AI engine itself to be more complex or heavily trained on rotation motions.

Inventive Principle:

Principle #24Intermediary (Mediator)

Solution Approach 2:

The patent performs preliminary action by pre-calculating compensation amounts for various rotation angles before actual use. The compensation table is prepared in advance, storing the relationship between rotation angles and compensation amounts, which eliminates the need for real-time complex calculations or retraining the AI model during operation.

Inventive Principle:

Principle #10Preliminary action

2Measurement precision

If the AI engine is trained with more data or the learning model is reconstructed to improve skeleton key point estimation accuracy for rotation motions, then the estimation accuracy is improved, but the training time and processing time increase

Engineering Contradiction:

Improveskeleton key point estimation accuracyVSAvoidtraining time

Core Design Contradiction:

Measurement precisionVSLoss of time

Solution Approach 1:

The patent performs preliminary action by pre-calculating compensation amounts for various rotation angles before actual use. The compensation table is prepared in advance, storing the relationship between rotation angles and compensation amounts, which eliminates the need for real-time complex calculations or retraining the AI model during operation.

Inventive Principle:

Principle #10Preliminary action

Solution Approach 2:

The patent extracts the rotation correction function from the main AI estimation engine and handles it separately through a dedicated compensation processing unit. This separation allows the AI engine to focus on general skeleton estimation while the compensation unit handles rotation-specific corrections, reducing overall processing time.

Inventive Principle:

Principle #2Taking out (Extraction)

3Measurement precision

If the AI engine is trained with more data or the learning model is reconstructed to improve skeleton key point estimation accuracy for rotation motions, then the estimation accuracy is improved, but the amount of training data required increases

Engineering Contradiction:

Improveskeleton key point estimation accuracyVSAvoidtraining data amount

Core Design Contradiction:

Measurement precisionVSQuantity of substance

Solution Approach 1:

The patent introduces an intermediary compensation processing unit that receives the estimated skeleton key points and corrects them using pre-calculated compensation amounts based on rotation angle. This mediator resolves the contradiction by providing accurate correction without requiring the AI engine itself to be more complex or heavily trained on rotation motions.

Inventive Principle:

Principle #24Intermediary (Mediator)

Solution Approach 2:

The patent changes the approach from training the AI model with diverse rotation data to using a parameter-based compensation method. By calculating compensation amounts based on rotation angle parameters and storing them in a lookup table, the system achieves accurate rotation handling without requiring extensive training data covering all possible rotation scenarios.

Inventive Principle:

Principle #35Parameter changes

Data Source

PatentUS20250371730A1Processing apparatus, processing method, and non-transitory computer readable medium

Publication Date: 2025.12.04 NEC CORP

AI summary

A processing apparatus according to the present disclosure includes at least one memory configured to store instructions, and at least one processor configured to execute the instructions to: receive inputs of either three-dimensional or two-dimensional skeleton key point coordinate estimation values and a gravity direction vector as input information for two images obtained by image capturing a frontal plane of a person at a time of standing and after rotation; calculate a width of a specific part from each of two silhouette images respectively indicating silhouettes of the two images, collate the width of the specific part that has been calculated and the gravity direction vector with anatomical knowledge, and calculate compensated coordinate values for the skeleton key point coordinate estimation values of the specific part in an image after the rotation; and output the compensated coordinate values that have been calculated.