Display Device Neural Network Compensation for Luminance Deviation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Display devices face issues with luminance deviation and afterimages due to variations in driving transistor characteristics and light emitting element deterioration, which existing technologies fail to adequately compensate for.
Innovation Solution
A display device and driving method that utilize a sensing unit and artificial neural network model to generate compensation image data by sensing current values, ensuring uniform current supply to pixels through a data driver, thereby compensating for transistor characteristics and improving image quality.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional sensing methods are used to compensate for pixel characteristics, then some level of compensation is achieved, but luminance deviation and afterimage effects persist due to insufficient compensation precision
Solution Approach 1:
The patent segments the sensing process into multiple stages: initial sensing during a sensing period, reference sensing for calibration, and compensation application during display periods. This multi-stage segmentation allows for more precise characterization of pixel characteristics and enables better compensation for luminance deviation and afterimage effects
Solution Approach 2:
The patent performs preliminary sensing actions during dedicated sensing periods before normal display operation. By sensing pixel characteristics in advance and storing reference values, the system prepares compensation data that can be applied during subsequent display periods, improving overall image quality without affecting real-time performance
2Measurement precision
If sensing operations are performed frequently to improve compensation accuracy, then sensing precision increases, but display time is reduced due to the time required for sensing operations
Solution Approach 1:
The patent implements periodic sensing operations where sensing periods are alternated with display periods. Reference sensing is performed periodically to update calibration data, while compensation sensing is performed at specific intervals. This periodic approach ensures adequate compensation accuracy while maximizing display time by not continuously sensing
Solution Approach 2:
Reference sensing and calibration are performed in advance during dedicated sensing periods, preparing compensation parameters that can be reused during multiple subsequent display periods. This preliminary action reduces the need for frequent full sensing operations, thereby increasing display time while maintaining compensation accuracy
3Measurement precision
If multiple reference sensing current values are used to improve neural network training accuracy, then compensation precision increases, but the complexity of the sensing unit and processing increases
Solution Approach 1:
The patent changes the parameter of initialization voltage to multiple reference levels (first reference initialization voltage, second reference initialization voltage, etc.). By sensing pixel characteristics at different voltage levels, the system obtains multiple reference sensing current values that provide more comprehensive training data for the neural network, improving compensation precision without requiring additional hardware complexity
4Reliability
If an artificial neural network model is implemented to improve compensation accuracy, then image quality increases, but device complexity and computational requirements increase
Solution Approach 1:
The neural network model is trained in advance using reference sensing data obtained during dedicated sensing periods. Once trained, the model's parameters are stored and can be applied during normal display operation without requiring real-time complex computations. This preliminary training action enables high-quality compensation while reducing real-time processing complexity
Solution Approach 2:
The patent uses reference sensing data to create a trained neural network model that copies the learned relationships between sensing current values and compensation requirements. This trained model can then be applied to generate compensation image data for multiple different input images without requiring the full computational complexity of the neural network for each individual image, thereby improving image quality while managing device complexity
Data Source
AI summary
A display device including pixels connected to a data line and a sensing line; a data driver supplying one of an image data signal and a sensing data signal to the data line; a sensing unit supplying an initialization voltage to at least one of the pixels through the sensing line and obtaining a sensing current value from at least one of the pixels through the sensing line; and a compensator generating compensation image data according to a sensing pixel current value among the sensing current value using an artificial neural network model learned to output reference image data corresponding to at least one reference sensing current value among the sensing current value.