Regression-Based Frame Error Concealment for Audio Signal Reconstruction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing frame error concealment methods are inefficient in reconstructing audio signals, leading to degradation in sound quality due to the propagation of errors in decoded frames, as they do not account for the specific characteristics of the input signal.
Innovation Solution
A frame error concealment method that utilizes regression analysis, specifically linear and non-linear regression methods, to analyze and set the appropriate concealment strategy based on signal characteristics, such as voiced or unvoiced sounds, to accurately reconstruct error frames by predicting parameters from previous good frames.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If the same frame error concealment method is used for all types of input signals, then the device complexity is reduced, but the sound quality degradation increases due to inefficient error concealment
Solution Approach 1:
The patent applies different frame error concealment methods based on the specific characteristics of the input signal. For voiced signals, one method is used, while for unvoiced signals, a different method is applied. This localized approach ensures optimal sound quality for each signal type without unnecessarily increasing overall system complexity.
Solution Approach 2:
The patent changes the concealment method parameters based on signal characteristics. By detecting whether the signal is voiced or unvoiced, the system selects appropriate regression analysis methods (linear or non-linear) to reconstruct error frames, thereby improving sound quality while maintaining manageable complexity through parameter adaptation.
2Reliability
If regression analysis is used to reconstruct error frames, then the sound quality is improved, but the computational complexity increases
Solution Approach 1:
The patent selects between linear and non-linear regression analysis based on signal characteristics. For voiced signals, linear regression is used which is computationally simpler, while for unvoiced signals, non-linear regression is applied. This adaptive parameter selection improves sound quality while minimizing unnecessary computational complexity.
Solution Approach 2:
The patent applies complex non-linear regression analysis only when necessary (for unvoiced signals) rather than for all signals. This partial application of the more complex method reduces overall computational burden while still achieving improved sound quality where it is most needed.
3Reliability
If error frames are reconstructed using prediction from previous good frames, then the sound quality degradation is reduced, but the error propagation to next frame occurs
Solution Approach 1:
The patent extracts and reconstructs only the necessary parameters (such as pitch period, gain, and spectral characteristics) from previous good frames using regression analysis, rather than simply copying the entire frame. This selective extraction and reconstruction approach reduces error propagation while maintaining sound quality.
Solution Approach 2:
The patent performs preliminary analysis of signal characteristics before reconstruction to determine the appropriate regression method. This preliminary action ensures that the reconstruction is optimized for the specific signal type, reducing the likelihood of error propagation to subsequent frames while maintaining high sound quality.
Data Source
Figure 1A~1B
Figure 2
Figure 3A~3B
AI summary
A frame error concealment method and apparatus and a decoding method and apparatus using the same. The frame error concealment method includes setting a concealment method to conceal an error based on one or more signal characteristics of an error frame having the error and concealing the error using the set concealment method.