In-Vehicle Negative Emotion Estimation with Per-Occupant CG Models
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing technologies face challenges in accurately determining a passenger's negative emotions due to individual variability and difficulty in capturing facial expressions, making it hard to assess whether a passenger is experiencing discomfort.
Innovation Solution
A negative emotion determination device for vehicles that uses a learned emotion estimation model combining vehicle state data and computer graphics models to determine negative emotions, either with or without facial expression data, through machine learning.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional emotion estimation techniques are used, then the system can determine driver emotions, but it cannot accurately estimate passenger emotions due to individual variability and capture limitations
Solution Approach 1:
The patent segments the emotion estimation approach by creating individualized emotion estimation models for each occupant. The system divides the general emotion estimation problem into person-specific sub-problems, training separate models that adapt to each individual's emotional patterns and characteristics, thereby achieving accurate estimation despite individual variability.
Solution Approach 2:
The system performs preliminary action by collecting and analyzing facial expression data during a learning period before actual emotion estimation is needed. During this phase, the system accumulates facial images and corresponding emotion information to train personalized models in advance, so that when emotion estimation is required, the pre-trained models can immediately provide accurate results without needing to adapt on the fly.
2Measurement precision
If facial expression capture is required for emotion determination, then emotion accuracy improves, but the system fails when facial images cannot be captured
Solution Approach 1:
The patent creates a virtual copy of the facial expression through a CG (computer graphics) model. When real facial images are unavailable or insufficient, the system generates a synthetic facial representation that replicates the essential expression characteristics. This CG model serves as a substitute copy that can be processed by the emotion estimation model, ensuring continuous operation even when direct facial capture fails.
Solution Approach 2:
The emotion estimation system is designed with multi-functionality to handle multiple input scenarios. It can process both real captured facial images and synthesized CG facial models, making the system universal in its ability to estimate emotions regardless of whether direct facial capture is available. This dual-input capability ensures the system remains reliable across different operating conditions.
3Measurement precision
If individualized emotion models are created for each occupant, then emotion estimation accuracy improves, but system complexity increases
Solution Approach 1:
The system implements self-service by automatically managing the creation, training, and maintenance of individualized emotion models without requiring manual intervention. The learning control unit autonomously collects data, trains models for each occupant, updates them as new data becomes available, and manages the model lifecycle. This automated self-service approach handles the complexity internally while presenting a simple interface to users.
Solution Approach 2:
The system incorporates feedback mechanisms where emotion estimation results and new facial expression data continuously feed back into the model training process. As more data is collected from each occupant, the models are automatically refined and updated, creating a closed-loop system that improves accuracy over time. This feedback-driven approach manages complexity by using the system's own operational data to maintain and enhance the models.
Data Source
AI summary
A negative emotion determination device determines whether an occupant has a negative emotion using a learned emotion estimation model for each occupant. The learned emotion estimation model is configured to output whether the occupant has the negative emotion regarding the state of a vehicle and a facial expression of the occupant by machine learning using both the state of the vehicle and a CG model as input data. The CG model is a model created based on the facial expression of the occupant captured by an in-vehicle camera.