Multi-Channel AR Model for Audio Signal Restoration
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional single-channel autoregressive models for audio signal restoration often result in audible distortion, especially when dealing with voiced speech or music, and are limited in interpolating long gaps, making them unsuitable for restoring corrupted audio segments in digital audio broadcasts and other applications.
Innovation Solution
A multi-channel autoregressive model is employed, which models the corrupted signal as a linear combination of scaled time-shifted portions of correlated audio signals from multiple channels, allowing for real-time restoration and interpolation of missing or corrupted audio packets.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If a Single-Channel AR Model is used for audio signal restoration, then the restoration process is simple and computationally efficient, but the restored signal contains audible distortion and the model cannot effectively interpolate long gaps
Solution Approach 1:
The patent transitions from single-channel to multi-channel AR modeling, adding the channel dimension to the restoration process. The multi-channel model uses correlations between multiple audio channels (e.g., stereo left and right channels) to restore corrupted segments, leveraging redundant information across channels to improve restoration quality while maintaining computational feasibility
Solution Approach 2:
The patent modifies the AR model parameters by incorporating inter-channel correlation coefficients and time-shifted versions of multiple channels into the restoration equation. This parameter expansion allows the model to capture more structural information about the audio signal, reducing distortion and improving gap interpolation performance
2Duration of action of moving object
If AR-based interpolation is used for long gaps, then the model can attempt to restore extended corrupted segments, but the performance deteriorates significantly toward the middle of the gap
Solution Approach 1:
By incorporating multiple channels with different temporal characteristics and correlations, the model gains additional dimensions for predicting missing samples in long gaps. The inter-channel relationships provide constraints that maintain accuracy even when interpolating extended durations of corrupted audio
3Ease of operation
If a Single-Channel AR Model is used, then parameter adjustment is straightforward, but the reconstruction becomes overly smooth and loses typical audio signal characteristics
Solution Approach 1:
The multi-channel AR model introduces additional parameters including inter-channel correlation coefficients and time-shift values, but these parameters are estimated automatically from the input multi-channel audio data. The model preserves audio signal characteristics by incorporating information from multiple correlated channels, preventing excessive smoothing while maintaining ease of operation through automatic parameter estimation
Data Source
AI summary
A method of restoring a corrupted audio signal includes the steps of inputting the corrupted audio signal in a first channel, inputting one or more further correlated audio signals in one or more further channels, and restoring the corrupted audio signal using a Multi-Channel Autoregressive (AR) Model that models the corrupted signal as a linear combination of scaled time shifted portions of the further signal(s) and the corrupted signal. Embodiments are described in which the method is used to improve received audio signals in DAB receivers and mobile telephones.