AI Bitstream Decoding With Model Identifier Synchronization
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional audio and video compression technologies fail to adapt to long-term evolution and optimization of coding standards due to the lack of updates in AI models after standardization, affecting performance.
Innovation Solution
Incorporating a model identifier in the bitstream to facilitate synchronization of encoding and decoding models, allowing for model updates or switches between devices, and utilizing methods like index information or storage locations to reduce bandwidth overheads.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If AI models are not updated after coding standard standardization, then coding standard stability is maintained, but coding performance optimization is affected
Solution Approach 1:
The patent introduces a dynamic model update mechanism where the model identifier in the bitstream can be changed to indicate updated AI models. This allows the coding standard to evolve over time by enabling decoders to retrieve and apply new models while maintaining backward compatibility with existing standards, thus resolving the contradiction between standard stability and performance optimization.
Solution Approach 2:
The patent uses the model identifier as an intermediary element in the bitstream that mediates between the fixed coding standard and the evolving AI models. The identifier acts as a pointer or reference that connects the standardized bitstream structure to externally stored model information, allowing performance improvement without disrupting standard compliance.
2Reliability
If model information is transmitted in the bitstream, then model synchronization is achieved, but bit rate overhead increases
Solution Approach 1:
The patent extracts the actual model data from the bitstream and stores it externally in a model database or storage system. Only a compact model identifier (pointer/reference) is retained in the bitstream, dramatically reducing bit rate overhead while maintaining the ability to synchronize models between encoder and decoder through the external storage system.
Solution Approach 2:
The patent moves model information from the traditional bitstream dimension to an external storage dimension. The model identifier in the bitstream serves as a key that accesses model data in a separate dimensional space (external database), thereby decoupling the bitstream size from model information size and reducing overhead.
Data Source
AI summary
Embodiments of this application provide a decoding method and an electronic device. The method includes: first receiving a first bitstream, where the first bitstream includes a model identifier and a second bitstream, and the second bitstream is generated based on an encoded signal; obtaining model information when a value of the model identifier is a preset identifier value; then performing model reconstruction based on the model information, to obtain a decoding model; and decoding the second bitstream based on the decoding model, to obtain a reconstructed signal. In this way, a decoding model at a decoder side and an encoding model at an encoder side are synchronously updated, or a decoding model used for decoding at a decoder side and an encoding model used for encoding at an encoder side are synchronously switched.