Audio Frame Loss Concealment Using Selective Spectral Interpolation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing audio frame loss concealment methods, particularly for clean sinus tones, suffer from audible discontinuities and high computational complexity, especially when dealing with longer bursts of lost frames.
Innovation Solution
A controller is employed to analyze spectral representations of audio signals, determining the nature of bin values as complex or real to selectively apply interpolation methods, reducing complexity and improving frequency estimation accuracy by using magnitude approximation and fractional bin offsets.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional interpolation methods are used for frame loss concealment, then frequency estimation can be performed, but audible discontinuities occur at frame boundaries and computational complexity increases
Solution Approach 1:
The patent applies different interpolation methods based on the local characteristics of spectral bins. For bins with non-zero imaginary parts (indicating tonal components), complex-valued interpolation is used to maintain phase information and avoid discontinuities. For bins with zero imaginary parts (indicating noise-like components), simple magnitude-based interpolation is used to reduce computational complexity. This local differentiation resolves the contradiction by applying the most appropriate method to each local region.
Solution Approach 2:
The patent changes the interpolation approach based on the parameter state of spectral bins (whether imaginary parts are zero or non-zero). By detecting the nature of bin values and switching between complex-valued and magnitude-only interpolation methods, the system adapts its behavior to minimize audible discontinuities while controlling computational complexity dynamically.
2Measurement precision
If complex-valued interpolation is applied to all bins, then frequency estimation accuracy improves, but computational complexity increases significantly
Solution Approach 1:
Instead of applying complex-valued interpolation uniformly across all bins, the patent identifies local regions (tons) where complex interpolation is necessary based on non-zero imaginary parts, and applies simple magnitude interpolation to noise-like regions where imaginary parts are zero. This selective application maintains accuracy where needed while reducing overall computational complexity.
Solution Approach 2:
The patent applies complex-valued interpolation only partially - specifically only to bins that have non-zero imaginary parts - rather than excessively applying it to all bins. This partial action approach ensures sufficient accuracy for tonal components while avoiding unnecessary computational overhead for noise-like components.
3Device complexity
If magnitude approximation is used for interpolation, then computational complexity is reduced, but frequency estimation accuracy may deteriorate
Solution Approach 1:
The patent applies magnitude approximation locally only to bins with zero imaginary parts (noise-like components), while using full complex-valued interpolation for bins with non-zero imaginary parts (tonal components). This localized magnitude approximation reduces computational complexity where it is most beneficial (for noise regions) while preserving accuracy where it matters most (for tonal regions).
Data Source
Figure 1
Figure 2
Figure 3
AI summary
Controlling a concealment method for a lost audio frame associated with a received audio signal is provided. At least one bin vector of a spectral representation for at least one tone is obtained, wherein the at least one bin vector includes three consecutive bin values for the at least one tone. Whether each of the three consecutive bin values has a complex value or a real value is determined. Responsive to the determination, the three consecutive bin values are processed to estimate a frequency of the at least one tone based on whether each bin value has a complex value or a real value.