Channel Capacity Optimization Using Dirichlet Process Estimation

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Existing communication systems struggle to optimize channel capacity when the channel model is unknown, leading to high complexity and inaccurate modeling of channel conditional probability distributions, especially in non-trivial channels like terahertz communications.

Innovation Solution

A method using a Dirichlet process and collapsed Gibbs sampling to approximate the channel conditional probability distribution, allowing for a compact and accurate representation, which is then used to optimize the input distribution for improved transmission strategies.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If an accurate description of the channel is used to learn the channel conditional probability distribution, then the accuracy of the distribution estimation is improved, but the computational complexity at the receiver increases

Engineering Contradiction:

Improveaccuracy of channel conditional probability distribution estimationVSAvoidcomputational complexity at the receiver

Core Design Contradiction:

Measurement precisionVSDevice complexity

Solution Approach 1:

The patent transforms the complex channel conditional probability distribution estimation problem into a finite mixture model with a predefined number of components. By changing the parameterization approach from general functional forms to specific mixture components with identifiable parameters, the receiver complexity is reduced while maintaining estimation accuracy through the collapsed Gibbs sampling algorithm.

Inventive Principle:

Principle #35Parameter changes

Solution Approach 2:

The patent uses training signals to create a learned representation of the channel conditional probability distribution that can be copied and reused for decoding. The finite mixture model parameters estimated during training are stored and applied during data transmission, avoiding the need to reprocess raw channel observations and reducing real-time computational complexity.

Inventive Principle:

Principle #26Copying

2Adaptability or versatility

If the number of parameters to be tracked is increased to model sophisticated channels, then the adaptability to different channel conditions is improved, but the overhead of pilots required increases

Engineering Contradiction:

Improveability to track sophisticated channel modelsVSAvoidoverhead of pilots

Core Design Contradiction:

Adaptability or versatilityVSQuantity of substance

Solution Approach 1:

The patent employs a finite mixture model with a fixed, moderate number of components that provides sufficient modeling capability for sophisticated channels without requiring excessive parameters. This partial action approach uses just enough model complexity to capture essential channel characteristics while avoiding the need for extensive pilot overhead that would be required for fully flexible high-dimensional models.

Inventive Principle:

Principle #16Partial or excessive action

Solution Approach 2:

The channel conditional probability distribution parameters are estimated in advance during a training phase using dedicated pilot signals. This preliminary action allows the system to learn and store the channel characteristics before actual data transmission, so that during normal operation, decoding can proceed with minimal additional overhead since the model parameters are already available.

Inventive Principle:

Principle #10Preliminary action

3Quantity of substance

If a compact representation of channel conditional probability distribution is used, then the overhead on the feedback link is reduced, but the accuracy of the distribution representation may be compromised

Engineering Contradiction:

Improveoverhead on feedback linkVSAvoidaccuracy of distribution representation

Core Design Contradiction:

Quantity of substanceVSMeasurement precision

Solution Approach 1:

The patent represents the channel conditional probability distribution as a finite mixture model with a fixed number of components, creating a compact parametric form that can be efficiently transmitted over the feedback link. Instead of transmitting full distribution functions or large datasets, only the mixture model parameters (weights, means, and covariances) are fed back, significantly reducing overhead while preserving essential distribution characteristics for accurate decoding.

Inventive Principle:

Principle #26Copying

Data Source

PatentUS12494947B2Method, system, and computer program for optimizing a capacity of a communication channel using a dirichlet process

Publication Date: 2025.12.09 MITSUBISHI ELECTRIC CORP

AI summary

The invention relates to a method for optimizing a capacity of a communication channel in a communication system comprising at least a transmitter, a receiver, said communication channel between the transmitter and the receiver, and a channel conditional probability distribution estimator. The transmitter transmits messages conveyed by a signal associated with a transmission probability according to an input signal probability distribution estimated by said estimator. The transmitter takes at least the messages as inputs and outputs the signal to be transmitted on the channel. The channel takes the transmitted signal as an input and outputs a received signal which is processed at the receiver in order to decode the transmitted message. The probability distribution for each possible transmitted signal is estimated from at least output signals received at the receiver, and by using a collapsed Gibbs sampling relying on a Dirichlet process.