CFR Parameter Prediction Using Machine Learning for New Use Cases

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Legacy crest factor reduction (CFR) optimization techniques are inefficient for online optimization and cannot predict optimal parameter values for new or un-predetermined use cases, leading to performance degradation and failure to meet spectral emission mask limits.

Innovation Solution

Employing machine learning models to learn the relationship between CFR use case attributes and optimal parameter values, enabling prediction of near-optimal configurations for various use cases, thus avoiding iterative optimization procedures and ensuring efficient performance across different scenarios.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If legacy iterative optimization techniques are used for CFR parameter optimization, then optimization can be performed offline, but the process is time-consuming and cannot provide rapid predictions for new use cases

Engineering Contradiction:
Improveoptimization accuracyVSAvoidoptimization time
Core Design Contradiction:
Measurement precisionVSLoss of time

Solution Approach 1:

The patent applies preliminary action by training machine learning models offline using legacy optimization techniques to generate labeled training data. This pre-computed knowledge is stored and later used for rapid predictions without requiring iterative optimization at runtime, thus resolving the contradiction between optimization accuracy and optimization time.

Inventive Principle:
Principle #10Preliminary action

Solution Approach 2:

The patent creates a copy of the optimization process by training machine learning models to replicate the behavior of legacy iterative optimization techniques. The trained models serve as lightweight copies that can predict optimal CFR parameters instantly without performing the actual iterative optimization, thereby eliminating time loss while maintaining accuracy.

Inventive Principle:
Principle #26Copying

2Reliability

If legacy optimization techniques are used, then known use cases can be optimized, but new or un-predetermined use cases cannot be handled, leading to performance degradation

Engineering Contradiction:
Improveperformance reliabilityVSAvoiduse case adaptability
Core Design Contradiction:
ReliabilityVSAdaptability or versatility

Solution Approach 1:

The patent implements universality by training multiple machine learning models with different supervised learning algorithms (e.g., random forest, gradient boosting, neural networks). These diverse models collectively handle various use case types, enabling the system to adapt to both known and new use cases while maintaining reliable performance across different scenarios.

Inventive Principle:
Principle #6Universality (Multi-functionality)

Solution Approach 2:

The patent applies parameter changes by using machine learning models to predict CFR parameters (such as clipping threshold, window size, peak cancellation coefficients) based on input signal characteristics. This allows the system to dynamically adjust parameters for new use cases without requiring re-optimization, thereby improving adaptability while maintaining performance reliability.

Inventive Principle:
Principle #35Parameter changes

3Measurement precision

If machine learning models are trained with multiple algorithms, then model selection can be optimized, but the training process becomes more complex

Engineering Contradiction:
Improveprediction accuracyVSAvoidtraining complexity
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The patent applies self-service by implementing an automated model selection mechanism that evaluates multiple trained machine learning models and automatically selects the best-performing one for each use case. This self-selecting process reduces the need for manual intervention and simplifies the complexity of managing multiple algorithms, while still achieving high prediction accuracy through systematic comparison.

Inventive Principle:
Principle #25Self-service

Data Source

PatentEP4708794A1Machine learning model for predicting crest factor reduction configuration parameter values
Publication Date: 2026.03.11 NOKIA SOLUTIONS & NETWORKS OY
  • EP4708794A1 patent drawingFigure 1~2
  • EP4708794A1 patent drawingFigure 3
  • EP4708794A1 patent drawingFigure 4

AI summary

Disclosed is a method comprising selecting one or more crest factor reduction processing characteristics to be optimized for a crest factor reduction technique; collecting use case attribute data comprising a set of values of one or more use case attributes associated with the crest factor reduction technique; determining, based on the use case attribute data, a set of optimized configuration parameter values for the one or more crest factor reduction processing characteristics; generating a set of labelled training data based on the use case attribute data and the set of optimized configuration parameter values; and training, based on the set of labelled training data, a machine learning model for predicting configuration parameter values for the one or more crest factor reduction processing characteristics.