Simulated RF Capacitance for Facial Expression Training
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
It is challenging to train a neural network for facial expression recognition due to the diverse range of expressions a human face can form and the wide range of diversity between different individual faces.
Innovation Solution
The method involves recognizing a plurality of digital human face models, simulating various facial expressions for each model, finding simulated capacitance measurements for an array of simulated RF antennas, and using these measurements as input training data for a neural network configured to output facial expression parameters.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If real human face data is used for training, then the neural network can learn from actual facial expressions, but the cost and manual effort for data collection and processing increases significantly
Solution Approach 1:
The patent uses photorealistic 3D rendered images of digital human faces as copies of real human faces for training the neural network. These synthetic images replicate the visual characteristics and facial expression variations of real human faces without requiring actual human subjects, thereby maintaining training data quality while eliminating the need for extensive manual data collection and processing
Solution Approach 2:
The system pre-generates a comprehensive library of photorealistic 3D rendered facial expressions covering various emotions, angles, and lighting conditions before training begins. This preliminary creation of diverse training data enables the neural network to be trained on a wide range of facial variations without requiring real-time or on-demand data collection during the training process
2Measurement precision
If diverse facial expressions and individual face variations are covered, then the neural network becomes more accurate, but the training complexity and data requirements increase
Solution Approach 1:
The patent employs a parameterized 3D face model that can dynamically generate infinite variations of facial expressions and individual face characteristics through adjustable parameters such as emotion intensity, facial geometry, and skin texture. This dynamic generation approach allows the system to cover diverse facial variations without manually creating and managing large static datasets, thereby reducing training system complexity while maintaining recognition accuracy
Solution Approach 2:
The system varies multiple parameters including facial expression type, emotion intensity, head pose, lighting conditions, and individual face characteristics to generate diverse training samples. By systematically changing these parameters, the neural network learns robust facial expression recognition across various conditions without requiring complex manual data curation for each scenario
3Reliability
If real user data is collected for training, then the training data reflects actual usage scenarios, but user privacy concerns arise
Solution Approach 1:
The patent replaces real user facial data with photorealistic 3D rendered images of digital human faces that replicate the visual and electromagnetic characteristics of real faces. These synthetic copies maintain the representativeness needed for accurate training while completely eliminating the use of actual user biometric data, thereby preserving user privacy without sacrificing training data quality
Solution Approach 2:
The system introduces a layer of synthetic digital human face models as intermediaries between the training process and real human users. These intermediary models serve as proxies that capture the essential characteristics needed for training while acting as a privacy-protecting barrier that prevents direct collection or storage of real user facial data
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
This approach improves the accuracy of facial expression prediction by generating more representative simulated capacitance measurements, reducing the costs and manual effort associated with training the neural network, and preserving user privacy by using simulated data.
Implementation Method 1
a wearable device equipped with suitable radio frequency (RF) antennas may generate an e-field in proximity to the user's body
Implementation Method 2
the capacitance between the RF antenna and the user's skin may vary as the distance between them changes
Data Source
AI summary
A method for training a neural network for facial expression recognition includes recognizing a plurality of digital human face models. For each of the plurality of digital human face models, a plurality of simulated facial expressions are simulated. Simulated capacitance measurements for an array of simulated radio frequency (RF) antennas are found for each of the plurality of simulated facial expressions. The simulated capacitance measurements for each simulated facial expression are provided as input training data to a neural network configured to output facial expression parameters based on input capacitance measurements.