Extended-Lattice MAP Decoding for Non-Uniform Symbol Signals

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Current decoding methods for information symbols with non-uniform probability distributions, such as those encountered in network coding, are either overly complex like MAP decoding or sub-optimal like ML decoding, leading to high error rates and computational challenges.

Innovation Solution

A simplified MAP decoding method that models the sum symbol using a Gaussian distribution and employs an augmented network to find the nearest neighbor of the received signal, reducing the search space and using a sphere decoder or battery-powered decoder to efficiently estimate the received symbol.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If MAP decoding is used for non-uniform information sources, then decoding accuracy is improved, but computational complexity increases significantly

Engineering Contradiction:
Improvedecoding accuracyVSAvoidcomputational complexity
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The patent segments the complex MAP decoding problem into two parts: (1) transforming the non-uniform probability distribution into a uniform one through a mapping function, and (2) applying simplified ML decoding on the transformed uniform distribution. This segmentation allows the system to achieve near-MAP accuracy with ML-level computational complexity.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent changes the parameter space by applying a mapping function that transforms symbols from a non-uniform probability distribution into a uniform probability distribution. This parameter transformation allows standard ML decoding algorithms to be applied effectively, resolving the contradiction between accuracy and complexity.

Inventive Principle:
Principle #35Parameter changes

2Device complexity

If ML decoding is used for simplicity, then computational complexity is reduced, but decoding accuracy deteriorates for non-uniform information sources

Engineering Contradiction:
Improvecomputational complexityVSAvoiddecoding accuracy
Core Design Contradiction:
Device complexityVSMeasurement precision

Solution Approach 1:

Instead of applying complex MAP decoding directly to the original non-uniform distribution, the patent inverts the approach by first transforming the non-uniform distribution into a uniform one through a mapping function, then applying simple ML decoding. This inversion allows ML decoding to achieve accuracy comparable to MAP decoding.

Inventive Principle:
Principle #13The other way round (Inversion)

3Measurement precision

If exhaustive search is performed for optimal decoding, then decoding accuracy is improved, but processing time increases significantly

Engineering Contradiction:
Improvedecoding accuracyVSAvoidprocessing time
Core Design Contradiction:
Measurement precisionVSLoss of time

Solution Approach 1:

The patent performs a preliminary action by pre-computing the mapping function that transforms the non-uniform probability distribution into a uniform one. This preliminary transformation allows subsequent decoding to be performed efficiently using simple ML decoding without requiring exhaustive search, thus reducing processing time while maintaining accuracy.

Inventive Principle:
Principle #10Preliminary action

Data Source

PatentEP3039805B1Map decoding method using extended lattice
Publication Date: 2019.07.17 INSTITUT MINES TELECOM TELECOM BRETAGNE
  • EP3039805B1 patent drawingFigure 1
  • EP3039805B1 patent drawingFigure 2
  • EP3039805B1 patent drawingFigure 3

AI summary

The invention relates to a method for MAP decoding of a signal received through a noisy channel. The signal is made up of alphabet symbols having non-uniform probability distribution. The symbols are represented by points from a point array (Λ). The probability distribution of the symbols is approximated by a Gaussian distribution. An expanded point array (Λaug) is constructed from the point array (Λ) and the ratio (β) between the noise variance and the variance in Gaussian distribution of the symbols. The proposed MAP decoding method then goes back to carrying out decoding according to ML criteria by searching for the point, from the expanded array, that is nearest to a point representing the received signal (yaug).