Lattice Vector Quantization with Successive Refinement at Low Bitrates

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing scalable audio and video codecs face performance limitations at low bitrates, particularly in mobile communication systems, where they offer worse quality compared to non-scalable codecs, and suffer from high complexity, storage requirements, and inefficiencies in dealing with outliers and redundancy in multiple descriptions.

Innovation Solution

The technology employs lattice vector quantization with voronoi codebooks, decomposing lattice vectors into quotient and remainder vectors using lattice division, and encoding these in successive voronoi codebooks, allowing for flexible and efficient successively refinable encoding and decoding, reducing complexity and memory usage while improving performance.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Adaptability or versatility

If scalable codecs are used to enable low-bitrate transmission and storage, then adaptability to different networks is improved, but quality at low bitrates deteriorates compared to non-scalable codecs

Engineering Contradiction:
Improveadaptability to different networksVSAvoidquality at low bitrates
Core Design Contradiction:
Adaptability or versatilityVSManufacturing precision

Solution Approach 1:

The patent segments the quantization process into multiple refinement stages, where each stage adds incremental precision to the reconstructed signal. The first stage provides a coarse quantization suitable for low-bitrate transmission, while subsequent stages progressively refine the quality. This segmentation allows the system to adapt to different network conditions by transmitting only the necessary number of refinement stages, thereby maintaining adaptability while improving quality at low bitrates compared to conventional scalable codecs.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent performs preliminary quantization in the first stage to establish a baseline representation that is sufficient for low-bitrate applications. This preliminary action ensures that even if only the first stage is transmitted, the receiver obtains a usable low-quality signal. Subsequent stages then build upon this preliminary representation to provide enhanced quality when bandwidth permits, thus resolving the contradiction between adaptability and quality.

Inventive Principle:
Principle #10Preliminary action

2Manufacturing precision

If larger codebooks are used in vector quantization to improve approximation quality, then distortion is reduced, but storage requirements and computational complexity increase

Engineering Contradiction:
Improveapproximation qualityVSAvoidcomputational complexity
Core Design Contradiction:
Manufacturing precisionVSDevice complexity

Solution Approach 1:

The patent divides the codebook into multiple refinement stages, where each stage uses a smaller, more manageable codebook. The first stage uses a coarse codebook with fewer entries, reducing storage and computational requirements. Subsequent stages use progressively finer codebooks that refine the approximation without requiring the entire large codebook to be stored and searched simultaneously. This segmentation maintains approximation quality while dramatically reducing device complexity.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent implements dynamic codebook selection where the number of refinement stages and the size of codebooks at each stage can be adapted based on available bandwidth and quality requirements. This dynamic approach allows the system to use smaller codebooks when resources are constrained while maintaining the option to use larger effective codebooks when quality is prioritized, thus resolving the contradiction between approximation quality and computational complexity.

Inventive Principle:
Principle #15Dynamics

3Reliability

If multiple descriptions are used in scalable coding to provide robustness, then reliability is improved, but redundancy increases leading to inefficiency

Engineering Contradiction:
ImproverobustnessVSAvoidredundancy
Core Design Contradiction:
ReliabilityVSLoss of substance

Solution Approach 1:

The patent segments the information into hierarchical refinement stages where each stage contains incremental information building upon the previous stage. This segmentation creates a natural progression from essential to enhanced information, providing robustness because the base layer can be decoded independently, while avoiding excessive redundancy because each subsequent layer contains only the differential information needed to improve quality, not a complete duplicate of previous layers.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent implements a nested structure where refinement stages are organized hierarchically, with each stage nested within and building upon the previous stage. The first stage contains the most essential information, and subsequent stages are nested within this framework, adding layers of detail. This nesting provides robustness (like multiple descriptions) while minimizing redundancy by ensuring each nested layer contains only the additional information necessary at that level of refinement.

Inventive Principle:
Principle #7Nested doll (Nesting)

Data Source

PatentUS8340450B2Successively refinable lattice vector quantization
Publication Date: 2012.12.25 TELEFONAKTIEBOLAGET LM ERICSSON (PUBL)
  • US8340450B2 patent drawing
  • US8340450B2 patent drawing
  • US8340450B2 patent drawing

AI summary

A vector quantizer includes a lattice quantizer (10) approximating a vector x by a lattice vector belonging to a lattice Λ0. A lattice vector decomposer (14) connected to the lattice quantizer successively decomposes the lattice vector into a sequence of quotient vectors y, and a sequence of remainder vectors ri on successive lattices ΛI−1 by lattice division with a corresponding predetermined sequence of integers pi≧2, where i=1 . . . k and k is a positive integer representing the number of elements in each sequence.