Neural Network Interference Estimation in Wireless Systems
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing wireless communication systems face challenges in efficiently estimating and removing interference signals, particularly due to the complexity of evaluating multiple interference combinations, which affects system throughput and reception performance.
Innovation Solution
A method and apparatus utilizing a trained neural network to estimate interference parameters in wireless communication systems, where the neural network is trained using input vectors comprising received signals and channel matrices, allowing for efficient determination of interference parameters without exhaustive search, even with increasing complexity.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional interference estimation methods are used, then interference can be removed from received signals, but the computational complexity increases significantly when evaluating multiple interference combinations
Solution Approach 1:
The patent replaces traditional mechanical/computational interference estimation methods with a machine learning-based system. A neural network model is trained offline to learn the relationship between received signals and interference parameters, then uses this trained model to rapidly estimate interference without exhaustive computational search during actual operation, thus reducing real-time computational complexity while maintaining accuracy
Solution Approach 2:
The patent performs preliminary training of the machine learning model offline using labeled training data that includes various interference scenarios. This preliminary action pre-computes the optimal interference estimation mappings, so that during actual operation, the system only needs to input received signals to get rapid interference estimates without performing complex real-time calculations
2Measurement precision
If exhaustive search methods are used to evaluate multiple interference combinations, then accurate interference parameters can be determined, but the processing time increases significantly
Solution Approach 1:
The patent substitutes exhaustive search algorithms with a pre-trained neural network that directly maps received signals to interference parameters. The neural network has learned the complex relationships during offline training, enabling rapid inference without iterative search, thus dramatically reducing processing time while maintaining accurate interference parameter estimation
Solution Approach 2:
The system performs exhaustive exploration of interference combinations during the offline training phase to build accurate training data. This preliminary action allows the model to learn from all possible interference scenarios beforehand, so that during operation, interference parameters can be determined instantly without time-consuming search
3Productivity
If machine learning models are used to estimate interference parameters, then computational complexity is reduced and processing speed is improved, but the model requires extensive training data and computational resources for training
Solution Approach 1:
The patent performs the computationally intensive model training and data generation as a preliminary offline action. Training data is generated in advance using simulated or measured signals with known interference characteristics. This separates the heavy computational burden from real-time operation, allowing rapid inference during actual use while accepting the upfront cost of data preparation and model training
Data Source
AI summary
The present disclosure relates to an interference estimation method and apparatus. The apparatus includes at least one first processor, a serving channel matrix, and an interference channel matrix. The at least one first processor generates an input vector from at least one of a received signal vector corresponding to a received signal with a serving signal and an interference signal. The serving channel matrix corresponds to the serving signal. The interference channel matrix corresponds to the interference signal and a second processor. The second processor executes at least one machine learning model trained by sample input vectors and sample interference parameters. The at least one first processor provides the input vector to the second processor and determines interference parameters corresponding to the interference signal based on an output vector of the at least one machine learning model provided by the second processor.