Audio Encoding Using Neural Network Up-Sampling for Weak Networks
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing audio encoding and decoding methods suffer from data loss during compression, leading to poor media quality, especially in weak network environments where transmission efficiency is compromised.
Innovation Solution
The proposed method involves up-sampling and down-sampling techniques using convolutional neural networks to enhance encoding and decoding processes, where encoding vectors of historical audio frames are used to improve the quality of current frames by caching up-sampling or down-sampling feature values, and utilizing a shared codebook for efficient data transmission.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of energy
If traditional audio encoding and decoding methods are used, then network transmission costs are reduced through compression, but media data quality deteriorates due to data loss during encoding
Solution Approach 1:
The patent applies preliminary action by pre-training a neural network model with large amounts of audio data before actual encoding/decoding operations. This pre-training enables the model to learn optimal encoding strategies and maintain high audio quality even during compression in weak network conditions, resolving the contradiction between transmission cost reduction and quality preservation
Solution Approach 2:
The patent dynamically adjusts encoding parameters based on network conditions. In weak network environments, the system modifies encoding parameters to prioritize quality over compression ratio, while in good network conditions, it increases compression. This adaptive parameter adjustment resolves the contradiction by allowing the system to optimize for either transmission efficiency or quality depending on real-time conditions
2Productivity
If compression is increased to reduce data transmission size, then network transmission efficiency improves, but data loss increases resulting in poor media quality
Solution Approach 1:
The patent implements feedback mechanisms where the neural network continuously monitors decoding quality and adjusts encoding parameters accordingly. The model uses feedback from previous encoding/decoding cycles to optimize the balance between compression ratio and quality, reducing information loss while maintaining transmission efficiency
Solution Approach 2:
The system dynamically adjusts compression levels based on real-time network conditions and content characteristics. Rather than using fixed compression ratios, the neural network adapts encoding strength dynamically, applying stronger compression only when network conditions permit and content allows, thereby minimizing information loss while maximizing transmission efficiency
3Loss of energy
If encoding compression is applied to reduce bandwidth usage, then transmission costs decrease, but audio quality deteriorates especially in weak network environments
Solution Approach 1:
The neural network model is pre-trained on extensive audio data to learn robust encoding strategies that maintain quality in weak network conditions. This preliminary training enables the system to reliably encode audio at lower bitrates without significant quality degradation, resolving the contradiction between bandwidth reduction and quality maintenance
Solution Approach 2:
The patent applies different encoding strategies to different portions of audio content based on their importance and characteristics. Critical audio elements are encoded with higher quality while less important portions receive stronger compression. This local quality approach allows overall bandwidth reduction while preserving essential audio quality, especially important in weak network environments
Data Source
AI summary
An audio decoding method performed by a computer device includes obtaining encoding vectors of audio frames in an audio frame sequence, and performing, in response to a current audio frame in the audio frame sequence being to be decoded, up-sampling on an encoding vector of a historical audio frame to obtain an up-sampling feature value describing the historical audio frame. The historical audio frame includes one or more audio frames decoded before the current audio frame in the audio frame sequence. The method further includes performing, based on the up-sampling feature value, up-sampling on an encoding vector of the current audio frame to obtain decoded data of the current audio frame.