Audio Signal Resampling Using Predicted Future Samples
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing audio frequency signal resampling methods introduce significant delays and artifacts, particularly in techniques like the AMR-WB codec, which complicate coding and decoding processes and result in audible discontinuities due to the use of FIR filters and delay compensation methods.
Innovation Solution
An adaptive linear prediction method is employed to determine the number of future signal samples based on the chosen resampling delay, constructing a resampling support vector from current and predicted samples, and applying a resampling filter to reduce delays while maintaining performance, allowing for flexible switching between different resampling configurations.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If FIR filter resampling is used to ensure linear phase and preserve waveform, then filtering characteristics are improved, but temporal spreading and pre-echo artifacts are introduced
Solution Approach 1:
The patent applies pre-emphasis filtering before resampling to pre-compensate for the temporal spreading and pre-echo artifacts that will be introduced by the subsequent FIR filter. This preliminary action modifies the signal characteristics in advance to counteract the harmful effects of the resampling operation.
Solution Approach 2:
The patent changes the parameter of the resampling filter by applying pre-emphasis with a specific time constant (typically 1-5 ms) that compensates for the expected temporal spreading. This parameter adjustment allows the system to maintain linear phase properties while reducing pre-echo artifacts through controlled modification of the signal's frequency spectrum.
2Device complexity
If IIR filter resampling is used to reduce filter order and complexity, then implementation complexity is reduced, but non-linear phase distortion is introduced
Solution Approach 1:
The patent extracts the phase distortion problem from the IIR filter design by separating the magnitude response design from the phase response compensation. The IIR filter is designed for optimal magnitude characteristics with reduced complexity, while phase linearity is addressed through additional all-pass filtering stages that compensate for non-linear phase without significantly increasing overall system complexity.
Solution Approach 2:
The patent introduces asymmetry in the filter structure by combining IIR filters (which have inherent non-linear phase) with all-pass compensation stages. This asymmetric approach allows the system to leverage the computational efficiency of IIR filters while correcting their phase deficiencies through targeted phase compensation networks.
3Manufacturing precision
If delay compensation by all-pass filtering is applied to IIR filters, then phase linearity is improved, but filter state dynamics reach high values and complexity increases
Solution Approach 1:
The patent applies partial phase compensation using all-pass filtering, recognizing that complete phase linearization is not necessary for acceptable performance. By applying only the essential phase compensation needed to meet perceptual quality standards, the system achieves adequate phase linearity while avoiding the excessive complexity and high state dynamics that would result from complete phase correction.
4Device complexity
If resampling delay is increased to reduce filtering requirements, then filter complexity is reduced, but coding delay increases
Solution Approach 1:
The patent changes the parameter of resampling delay to an optimized value that balances filter complexity reduction with acceptable coding delay. By carefully selecting the delay parameter and corresponding filter length, the system achieves significant complexity reduction while maintaining coding delay within perceptually acceptable limits, typically optimizing for delays below 10-20 ms depending on the application.
Data Source
AI summary
A method and device for resampling an audio frequency signal in an audio frequency signal coding or decoding. The method includes the following acts for each signal block to be resampled: determining, by adaptive linear prediction, a number of future signal samples, this number being defined as a function of a chosen resampling delay; constructing a resampling support vector from at least samples of the current block and determined future signal samples; applying a resampling filter to the samples of the resampling support vector.