Receiving Device Codec-Matched Upscaling for Low-Bitrate Audio
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing wireless communication systems face challenges in maintaining high-quality audio streaming due to unstable network conditions, leading to significant data loss and degradation of output signals when using low bitrates and narrow frequency bands.
Innovation Solution
A receiving device employs a neural network model to upscale decoded frames based on codec information, utilizing a super resolution (SR) neural network to generate high-resolution output signals by compensating for data loss during encoding.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If a low bitrate and narrow frequency band are used for encoding in unstable network conditions, then data transmission reliability is improved, but audio signal quality deteriorates due to significant data loss
Solution Approach 1:
Multiple SR neural network models are pre-trained corresponding to different codec information (bitrate, frequency band, sampling rate) before actual use. When decoding, the appropriate pre-trained model is selected based on the codec parameters, enabling rapid restoration of audio quality without real-time training overhead.
Solution Approach 2:
Different SR neural network models are trained with different codec parameters (bitrate, frequency band, sampling rate) as input conditions. The receiving device selects and applies the model that matches the specific codec parameters used during encoding, allowing adaptive restoration of audio quality based on the encoding conditions.
2Loss of information
If multiple SR neural network models are trained for each codec information, then audio signal restoration quality is improved, but device complexity increases
Solution Approach 1:
Multiple SR neural network models are pre-trained offline corresponding to different codec information (bitrate, frequency band, sampling rate) before actual use. When decoding, the appropriate pre-trained model is selected based on the codec parameters, enabling rapid restoration of audio quality without real-time training overhead.
Solution Approach 2:
Instead of training multiple complex models from scratch, the system uses copies of a base SR neural network model that have been pre-trained on different codec conditions. These model copies share the same architectural structure but have different learned parameters optimized for specific bitrate, frequency band, or sampling rate conditions.
Data Source
AI summary
A receiving device is provided. The receiving device includes: a memory configured to store a plurality of encoded frames received from a transmission device; and a processor configured to receive, from the transmission device, the plurality of encoded frames from the memory, receive first codec information corresponding to parameters used to encode the plurality of encoded frames, decode the plurality of encoded frames based on the first codec information, generate a plurality of decoded frames, up-scale the plurality of decoded frames using a first super resolution (SR) neural network model corresponding to the first codec information, and generate a plurality of up-scaled frames.