Noise-Feedback Spectral Envelope Quantization for Low Bit Rate Audio
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In low bit rate audio/speech coding, precise spectral envelope quantization requires many bits, making it challenging to maintain quality while keeping complexity and memory usage low, especially in BandWidth Extension (BWE) and High Band Extension (HBE) applications.
Innovation Solution
The Noise-Feedback method for spectral envelope quantization, where the quantization error of previous spectral magnitudes is fed back to adaptively modify the quantization criterion for current magnitudes, using scalar quantization techniques such as direct or differential coding in the Log or Linear domain, to improve shape matching and energy compensation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If precise spectral envelope coding is used, then spectral envelope quantization precision is improved, but bit rate increases
Solution Approach 1:
The patent applies feedback by using the quantization error from previously coded spectral magnitudes to adaptively modify the quantization criterion for current magnitudes. Specifically, the quantization error is fed back to adjust the step size or threshold in subsequent quantization operations, allowing the system to maintain high precision while adapting to the actual error patterns and reducing overall bit requirements.
Solution Approach 2:
The quantization criterion is made dynamic rather than static. The patent adapts the quantization parameters (such as step size or precision level) based on the quantization error observed in previous spectral magnitudes. This dynamic adaptation allows the system to allocate bits more efficiently, using higher precision only when necessary and lower precision when the signal is more predictable, thereby reducing overall bit rate while maintaining precision.
2Measurement precision
If precise spectral envelope coding is used, then spectral envelope quantization precision is improved, but device complexity increases
Solution Approach 1:
The feedback mechanism uses simple error propagation where the quantization error from previous magnitudes is fed back to adjust current quantization. This approach achieves high precision without requiring complex algorithms, as it leverages the natural correlation between adjacent spectral magnitudes and uses straightforward error feedback rather than complex optimization procedures.
Solution Approach 2:
The patent performs preliminary quantization of spectral magnitudes in a systematic sequence, using the results from previous quantization steps to inform subsequent steps. This preliminary action approach allows the system to build up precision incrementally through simple, manageable operations rather than requiring complex simultaneous optimization, thereby reducing overall computational complexity.
3Measurement precision
If precise spectral envelope coding is used, then spectral envelope quantization precision is improved, but memory usage increases
Solution Approach 1:
The feedback mechanism requires storing only the quantization error from the immediately previous spectral magnitude, rather than storing complete spectral envelope data or large lookup tables. This minimal memory requirement for error feedback allows the system to achieve high precision quantization while using very little memory, as it only needs to retain the small error value from the prior step for adaptive adjustment.
Data Source
AI summary
A method of transmitting an input audio signal is disclosed. A current spectral magnitude of the input audio signal is quantized. A quantization error of a previous spectral magnitude is fed back to influence quantization of the current spectral magnitude. The feeding back includes adaptively modifying a quantization criterion to form a modified quantization criterion. A current quantization error is minimized by using the modified quantization criterion. A quantized spectral envelope is formed based on the minimizing and the quantized spectral envelope is transmitted.