ADPCM Prediction Filter for Uniform Spectral Compression
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing ADPCM technologies face challenges in maintaining high prediction gain while ensuring stability and convergence of predictor settings, especially in the presence of transmission errors, and struggle to uniformly compress spectral dynamic ranges across varying audio signal densities.
Innovation Solution
The method involves using adaptive filtering with time-varying coefficients to produce a partially whitened signal, which is then further processed by an adaptive whitening filter to control the spectral relationship, ensuring that the output signal has an increased spectral dynamic range. This approach allows for controlled spectral compression and expansion, maintaining prediction gain while stabilizing adaptations.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Stability of the object's composition
If leakage is applied to improve stability of adaptation, then stability improves, but prediction gain is significantly reduced
Solution Approach 1:
The patent applies different leakage values dynamically based on the spectral density of the input signal. When spectral density is high, a first (larger) leakage value is used to maintain stability. When spectral density is low, a second (smaller) leakage value is used to preserve prediction gain. This dynamic adaptation resolves the contradiction by making the leakage parameter context-dependent rather than fixed.
Solution Approach 2:
The patent changes the leakage parameter based on the spectral density of the input signal. By monitoring spectral density and adjusting the leakage value accordingly, the system optimizes both stability and prediction gain for different signal conditions, resolving the trade-off between these two parameters.
2Productivity
If spectral dynamic range is compressed uniformly, then compression efficiency improves, but variations in spectral density cause inconsistent compression ratios
Solution Approach 1:
The patent applies different leakage values to different spectral regions based on their local spectral density characteristics. High spectral density regions receive one leakage value while low spectral density regions receive another, ensuring that compression is optimized locally for each region's characteristics rather than applying a uniform approach.
Solution Approach 2:
The system changes the leakage parameter based on the local spectral density of different frequency regions, allowing the compression ratio to be consistent across varying spectral densities while maintaining high compression efficiency in each local region.
3Speed
If predictor settings are adapted quickly, then adaptation speed improves, but convergence to correct settings is compromised
Solution Approach 1:
The patent dynamically adjusts the leakage value based on spectral density conditions. During transient conditions or when spectral density is high, larger leakage provides faster adaptation. When spectral density is low or during steady-state, smaller leakage ensures accurate convergence. This dynamic adjustment resolves the speed-accuracy trade-off.
Solution Approach 2:
The system changes the leakage parameter according to spectral density measurements, enabling fast adaptation when needed while ensuring accurate convergence when conditions permit, thus resolving the contradiction between adaptation speed and convergence accuracy.
Data Source
AI summary
In lossy data compression of a signal using ADPCM, an adaptive decorrelation or “prediction” filter is used to reduce the amplitude of the signal, the spectral dynamic range of the signal also being reduced. This latter reduction is effected in a nonuniform manner, if known techniques are used, with regions of high spectral density being compressed more than regions of low spectral density. The present invention recognises that using a uniform compression ratio results in a better tradeoff between compression and robustness to transmission channel errors. A method is described for obtaining a uniform compression ratio by adjusting coefficients of the decorrelation filter in dependence on coefficients of an adaptive training filter that is fed from the output of the decorrelation filter. A reverse method is also provided along with encoder, decoder and codec implementing the techniques.