Signal Processing with Domain Conversion for Mismatched Learning Signals
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing signal processing technologies struggle to perform quality improvement processing effectively when the signal characteristics of an input signal differ from those of the student data used during learning, as they often lack suitable teacher data and cannot seamlessly switch between learning and inference.
Innovation Solution
A signal processing device that includes a quality improvement processing unit using a learning result from a first domain and a domain conversion unit to adapt input signals with different characteristics, utilizing deep neural networks for both units and a discriminator to determine appropriate processing paths.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If quality improvement processing is performed using a learning result obtained from student data with specific signal characteristics, then the processing is effective for signals with matching characteristics, but it cannot be applied appropriately to signals with different characteristics
Solution Approach 1:
A domain conversion unit is introduced as an intermediary component that converts input signals with different characteristics into the first domain matching the student data characteristics. This mediator enables the quality improvement processing unit to handle diverse input signals by transforming them into a compatible format before processing, thus resolving the contradiction between processing effectiveness and signal compatibility
Solution Approach 2:
The system changes the domain parameters of the input signal through the domain conversion unit. By transforming the signal characteristics (domain parameters) to match the student data, the system maintains effective quality improvement processing across different input signal types without requiring separate learning models for each signal characteristic
2Adaptability or versatility
If learning is performed directly on unknown video signals or as online/background processing, then the processing can adapt to new signal types, but teacher data with the same video content and high image quality cannot usually be obtained
Solution Approach 1:
Instead of requiring actual teacher data with the same video content, the system creates a copied or transformed version of the unknown signal by converting it to the first domain. This copied representation serves as a substitute for the unavailable teacher data, enabling the quality improvement processing to proceed without losing the ability to process unknown signal types
3Manufacturing precision
If a learning method is used without considering inference, then learning optimization can be achieved, but it is not possible to continuously output a signal and switching between learning and inference is necessary
Solution Approach 1:
The domain conversion unit serves multiple functions: it operates during the learning phase to prepare student data and during the inference phase to convert unknown signals. This multi-functional design eliminates the need to switch between separate learning and inference systems, enabling continuous signal output while maintaining optimized learning results
Data Source
AI summary
Quality improvement processing to which a learning result is applied can be performed appropriately even if the signal characteristics of an input signal are different from the signal characteristics of student data at the time of learning. A quality improvement processing unit applies quality improvement processing to an input signal to obtain an output signal using a learning result obtained using a signal of a first domain having first signal characteristics as student data. A domain conversion unit converts the input signal to a signal of the first domain and sends the same to a quality improvement processing unit when the input signal is a signal of a second domain having second signal characteristics different from the first signal characteristics.