Frame Error Concealment Using Transform-Coefficient Sign Tracking
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing frame error concealment methods in audio transmission, particularly in real-time conversational scenarios, suffer from degradation in audio quality due to frame loss or corruption, especially for tonal components, and introduce additional delays or require extra bandwidth and computational complexity.
Innovation Solution
A decoder-based method that tracks sign changes between transform coefficients of consecutive good frames, accumulates sign changes, and reconstructs erroneous frames using the latest good frame with reversed signs for sub-vectors exceeding a threshold, without additional side information or delay.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If frame repeat-algorithm is used for error concealment, then computational complexity is low and no extra bits are required, but audio quality degrades due to energy discontinuities and poor perceived quality
Solution Approach 1:
The patent modifies the repeated frame parameters by applying sign reversal to transform coefficients based on tracked sign changes from consecutive good frames. This parameter modification maintains the low complexity of repetition while improving audio quality by reducing energy discontinuities and perceptual distortion.
Solution Approach 2:
The patent performs preliminary tracking of sign changes in transform coefficients during good frames before the error occurs. This advance preparation allows the concealment algorithm to apply appropriate sign corrections when frame loss occurs, improving quality without increasing computational complexity during the concealment process.
2Reliability
If interpolation-based FEC is used to reconstruct lost frames, then audio quality improves, but additional delay is introduced which is not acceptable in real-time applications
Solution Approach 1:
The patent uses the decoder's own previously received good frames to generate the concealment data through sign change tracking. This self-service approach eliminates the need for additional side information transmission or complex interpolation algorithms that would introduce delay, while still improving quality over simple repetition.
3Reliability
If noise injection is applied to missing spectral coefficients, then error concealment is provided, but negative perceptual impact occurs especially in regions with prominent tonal components
Solution Approach 1:
Instead of injecting random noise as in conventional approaches, the patent inverts the approach by using sign reversal based on tracked sign changes. This inversion transforms a potentially harmful random process into a structured correction method that preserves tonal components while providing error concealment.
4Ease of manufacture
If muting-based FEC is used where missing spectral coefficients are set to zero, then implementation is simple, but large energy discontinuities are created and perceived quality is poor
Solution Approach 1:
The patent modifies the simple muting approach by applying sign reversal to the repeated frame coefficients based on tracked sign changes. This maintains the simplicity of the repetition framework while significantly improving perceived quality by reducing energy discontinuities through parameter modification.
Data Source
AI summary
A frame loss concealment method. The method includes receiving a first frame comprising B bands of n transform coefficients, where B>1 and n>1, and receiving a second frame comprising B bands of n transform coefficients. The method also includes for the first band of the first frame, storing a first counter value (C[1](1)) and for the first band of the second frame, storing a second counter value (C[2](1)). The method also includes setting a value of a state variable (denoted ΔC[2](1)) equal to C[2](1)+C[1](1) and using ΔC[2](1) in a process for producing a constructed frame. The constructed frame comprises an ordered set of B vectors of n transform coefficients, and the process for producing the constructed frame comprises, for each one of the n transform coefficients of the first vector of the constructed frame, determining a sign for the transform coefficient based on ΔC[2](1).