Multiple Description Coding with Noiseless Correlating Transforms
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing multiple-description coding techniques introduce quantization noise and are computationally intensive, degrading the quality of reconstructed signals and requiring significant resources.
Innovation Solution
Applying a correlating transform with exact recovery capabilities using quantized signal elements and decorrelating transforms, along with heterogeneous quantizing resolutions and Hadamard transforms, to minimize noise and optimize computational efficiency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If known correlating transform techniques are applied to encode information into multiple descriptions, then the information can be distributed into separate bitstreams for transmission, but quantization noise is injected into the encoded information which degrades the perceived quality of the reconstructed signal
Solution Approach 1:
The patent applies quantization before the correlating transform rather than after, so that the transform operates on coarsely quantized values. This preliminary quantization action allows the transform to distribute already-quantized information without introducing additional quantization noise during the transform process itself, thereby maintaining signal quality while achieving multiple description capability
Solution Approach 2:
The patent inverts the conventional order of operations by applying quantization before the correlating transform instead of after. This reversal of the typical processing sequence allows the transform to work with quantized inputs and produce quantized outputs that can be exactly recovered, eliminating the quantization noise problem that plagues conventional approaches
2Reliability
If known correlating transform techniques are used to divide encoded information into parts, then multiple descriptions can be generated for robust transmission, but the computational resources required to perform the transforms are considerable
Solution Approach 1:
By applying quantization before the correlating transform, the patent reduces the precision requirements for intermediate computational steps. The transform operates on coarsely quantized values rather than high-precision floating-point numbers, which significantly reduces computational complexity while maintaining the reliability benefits of multiple description coding
Solution Approach 2:
The patent changes the parameter precision from high-precision floating-point to low-precision quantized values at the input stage. This parameter change propagates through the transform process, allowing the use of simpler arithmetic operations and reducing the computational burden while preserving the essential functionality of the correlating transform for reliable transmission
3Productivity
If quantized signal elements are used as input to correlating transforms, then computational efficiency can be improved, but exact recovery of the original signal elements may not be possible
Solution Approach 1:
The patent inverts the conventional approach by applying quantization before the transform rather than after. This reversal ensures that the quantized values are the input to the transform, and the transform is designed to exactly recover these same quantized values at the decoder, not the original high-precision signal elements. This inversion resolves the contradiction by making exact recovery of the quantized elements the goal rather than recovery of the original elements
Solution Approach 2:
The patent creates an exact copy of the quantized signal elements through the invertible transform process. The correlating transform and its inverse are designed so that the quantized elements passed through the transform can be exactly reconstructed, creating a perfect copy of the quantized data that can be used for multiple description coding without loss
Data Source
AI summary
Transmitters and receivers in multiple description coding systems use correlating and decorrelating transforms to generate and process multiple descriptions of elements of an input signal. The multiple descriptions include groups of correlating transform coefficients that permit recovery of an inexact facsimile of the signal if some of the correlating transform coefficients are lost or corrupted during transmission. Noiseless implementations of the correlating and decorrelating transforms are described that allow the signal elements to be quantized with different quantizing resolutions. Implementations using the Fast Hadamard Transform are described that reduce the resources needed to perform the transforms.