Alternating Quantizer Ensemble for Audio Video Signal Reconstruction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional quantization methods introduce perceptible artefacts at low bit rates, leading to decreased signal power and annoying artefacts in reconstructed audio or video signals, which existing techniques struggle to mitigate effectively.
Innovation Solution
The method involves using an ensemble of quantizers that alternate in encoding and decoding, with each quantizer having its own partition and reconstruction points, selected randomly or following a predetermined sequence, to minimize statistical correlation between quantization errors and signal values, thereby reducing observable distortion.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Quantity of substance
If conventional quantization is applied to reduce bit rate, then bandwidth and storage cost are reduced, but perceptible quantization noise and artefacts appear in the reconstructed signal
Solution Approach 1:
The patent segments the single quantization process into multiple alternating quantizers with different partitions. Instead of using one quantizer that introduces consistent artefacts, the signal is quantized by multiple quantizers sequentially, each with its own partition structure. This segmentation distributes the quantization error across different patterns, preventing the formation of persistent artefacts while maintaining low bit rate operation.
2Quantity of substance
If the number of reference levels is reduced to lower bit rate, then transmission efficiency improves, but signal power decreases and quantization noise becomes more perceptible
Solution Approach 1:
The patent implements periodic action by alternating between multiple quantizers in a repeating sequence. Each quantizer is applied for a certain number of samples, then switched to the next quantizer in the ensemble. This periodic switching ensures that no single quantizer's artefacts dominate, while the alternating pattern maintains signal power distribution across different reference level sets, improving perceived quality at reduced bit rates.
3Object-affected harmful factors
If dithering is applied to mitigate quantization artefacts, then statistical correlation between quantization error and source signal is reduced, but the probability density of the reconstructed signal differs significantly from the original at low rates
Solution Approach 1:
The patent introduces an intermediary approach by using multiple quantizers as mediators between the original signal and the quantized output. Instead of applying dithering directly to a single quantizer (which distorts the probability density), the signal passes through multiple quantizers that collectively act as intermediaries. Each quantizer processes the signal with its own characteristic error pattern, and the combination of these intermediaries preserves the overall probability density while still reducing statistical correlation between error and signal.
Data Source
Figure 1
Figure 2
AI summary
The invention relates to improved signal quantization and inter alia addresses the problem of reducing artefacts in the reconstructed signal that arise from unintended statistical correlation between quantization errors and signal values. In one aspect, the invention provides an encoder characterized by the fact that several quantizers (determined by a partition of quantization cells and reconstruction points) are applied, one at a time, in an alternating fashion. In another aspect, the invention provides a decoder characterized by a corresponding alternation of quantizers, so that quantized samples are reconstructed by a corresponding quantizer on the decoder side. The invention further provides a method of manufacturing an encoder, decoder or quantizer according to the invention, which includes designing an ensemble of quantizers.