Wireless Channel Data Filtering for AI Model Training
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Wireless communication systems, particularly at the physical layer, suffer from low-quality channel data due to environmental noise, affecting the performance of AI models trained with such data.
Innovation Solution
A communication method and device that determine the validity of channel data based on metrics like RSRP, RSRQ, and power distribution in the delay domain to filter out low-quality data, ensuring only high-quality data is used for model training.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Quantity of substance
If channel data is collected from wireless communication physical layer, then data quantity is sufficient for AI training, but data quality deteriorates due to environmental noise
Solution Approach 1:
The patent extracts and removes low-quality channel data from the dataset by evaluating each data sample against predefined quality criteria (RSRP thresholds, RSRQ thresholds, power distribution characteristics). Only data samples meeting the quality standards are retained for model training, while noisy or invalid data is discarded. This extraction process resolves the contradiction by separating high-quality data from the mixed dataset.
Solution Approach 2:
The patent applies different quality evaluation criteria to different aspects of channel data: RSRP for signal strength, RSRQ for signal quality, and power distribution for temporal characteristics. Each data dimension is evaluated with appropriate thresholds and standards, allowing nuanced quality assessment that preserves locally high-quality data while filtering out problematic samples.
2Quantity of substance
If all collected channel data is used for model training, then training data volume is maximized, but model performance deteriorates due to noisy data
Solution Approach 1:
The patent extracts only the high-quality portion of channel data for model training by evaluating each data sample against quality criteria. Data samples failing to meet RSRP thresholds, RSRQ thresholds, or power distribution requirements are excluded from the training set. This ensures the model trains on clean, reliable data while maintaining sufficient training volume.
Solution Approach 2:
The patent changes the parameter composition of the training dataset by selectively including or excluding data samples based on multiple parameters (RSRP, RSRQ, power distribution). This parameter-based filtering transforms the raw collected data into a refined training dataset with optimized quality characteristics for AI model training.
3Ease of manufacture
If channel data with low quality is used for training, then data collection process is simple, but training effect deteriorates
Solution Approach 1:
The patent performs preliminary quality evaluation and filtering of channel data before it is used for model training. By pre-assessing data quality using RSRP, RSRQ, and power distribution criteria, the system ensures only high-quality data enters the training pipeline. This preliminary action prevents poor-quality data from degrading training effects while maintaining collection simplicity.
Solution Approach 2:
The patent implements a feedback mechanism where channel data quality is continuously evaluated against predefined criteria, and evaluation results feed into the decision of whether to include data in the training set. This closed-loop quality control ensures training effectiveness while keeping the collection process straightforward through automated quality assessment.
Data Source
AI summary
A communication method includes the following. A first communication device determines whether first channel data is valid or invalid, where the first channel data is usable for training a first model when the first channel data is valid, and the first channel data is unusable for training the first model when the first channel data is invalid.