Neural Network Residual Signal Generation for Audio Coding
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current audio coding methods fail to effectively reduce the amount of information in residual signals, which are the largest in audio data, leading to inefficiencies in information transfer between encoders and decoders.
Innovation Solution
A neural network with a convolutional layer is trained to restore and generate residual signals, allowing for direct generation in the audio decoding process by combining reference and residual signals, thereby improving coding efficiency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of information
If typical methods are used for audio coding, then the audio coding process is simple, but the amount of information of residual signal cannot be reduced
Solution Approach 1:
The neural network is trained in advance using a training dataset containing reference signals and corresponding residual signals. During training, the network learns to predict residual signals by processing reference signals through convolutional layers and activation functions. This preliminary training enables the network to efficiently generate accurate residual signals during actual audio coding without requiring complex real-time processing
Solution Approach 2:
The patent replaces traditional mechanical signal processing methods with a neural network-based system. Instead of using conventional filtering and transformation techniques to process residual signals, the system employs a trained neural network that uses convolutional operations and non-linear activation functions to predict and reconstruct residual signals, achieving better compression while maintaining audio quality
2Reliability
If residual signal is used in audio coding, then audio information is preserved, but information transfer efficiency deteriorates
Solution Approach 1:
The neural network creates an accurate copy of the residual signal by predicting it from the reference signal. Instead of transmitting the full residual signal, the system transmits only the necessary parameters to reconstruct the residual signal at the decoder using the pre-trained neural network model, significantly reducing information transfer requirements while maintaining signal accuracy
Solution Approach 2:
The patent transforms the residual signal representation by changing from direct signal transmission to parameter-based reconstruction. The neural network processes the reference signal to generate parameters that can be efficiently transmitted, and these parameters are then used to reconstruct the residual signal at the decoder, improving information transfer efficiency while preserving signal fidelity
Data Source
AI summary
Disclosed is a method of processing a residual signal for audio coding and an audio coding apparatus. The method learns a feature map of a reference signal through a residual signal learning engine including a convolutional layer and a neural network and performs learning based on a result obtained by mapping a node of an output layer of the neural network and a quantization level of index of the residual signal.