Driver Emotion Estimation with Thresholded Facial Features
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing emotion estimation systems for vehicle drivers are complicated and prone to inaccuracies, especially when the driver has a blank facial expression, leading to inappropriate emotion detection.
Innovation Solution
An apparatus that sets a face feature threshold based on deviations from a reference state in face images, specifies features exceeding this threshold, and normalizes them to estimate driver emotions using a trained neural network.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If characteristic points are detected from face images to determine facial expression, then emotion estimation can be performed, but the system becomes complicated and inaccurate when the driver has a blank facial expression
Solution Approach 1:
The patent extracts only the necessary facial feature quantities (distance between eyes, distance between eyebrows, mouth opening degree) rather than detecting all characteristic points. This extraction approach simplifies the system while maintaining emotion estimation accuracy by focusing on key features that reliably indicate emotional states.
Solution Approach 2:
The patent changes the parameters used for emotion detection from complex characteristic point coordinates to simplified facial feature quantities (distances and opening degrees). This parameter transformation reduces computational complexity while improving reliability, especially for blank expressions where characteristic point detection fails.
2Reliability
If characteristic points are used for emotion detection, then facial expressions can be identified, but blank facial expressions cause confusion in emotion estimation
Solution Approach 1:
The patent converts the harmful effect of blank facial expressions (which cause confusion in characteristic point detection) into a benefit by using feature quantities that remain stable and measurable even when expressions are blank. The distance-based measurements and mouth opening degrees provide reliable data regardless of whether the driver is expressing emotion.
3Measurement precision
If face features are normalized without thresholding, then all facial variations are considered, but significant features are lost among minor variations
Solution Approach 1:
The patent performs preliminary thresholding of facial feature quantities before normalization. By setting thresholds to identify significant deviations from baseline values, the system pre-processes the data to highlight emotionally relevant features while filtering out minor variations, thereby preserving significant information through the normalization process.
Data Source
AI summary
An apparatus for emotion estimation sets a face feature threshold based on a plurality of face features indicating a deviation from a reference state of a predetermined part of a face detected from each of a plurality of face images representing a face of a driver of a vehicle generated by a camera mounted on the vehicle in a predetermined time range, specifies face features larger than the face feature threshold among the plurality of face features detected from each of the face images generated by the camera at a time not included in the predetermined time range, and estimates an emotion of the driver by using normalized features obtained by normalizing the specified face features downward.