Deep Learning De-aliasing Radar Signal Processing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Radar systems often experience aliasing issues due to under-sampling, leading to uncertainty in object detection, which affects the accuracy and reliability of vehicle operation systems.
Innovation Solution
The method employs deep learning to de-alias radar signals by training neural networks using un-aliased and aliased data samples, allowing for the selection of optimal configurations in ambiguous radar systems to achieve accurate object detection without incurring accuracy penalties.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Use of energy by moving object
If under-sampling is performed to reduce resource consumption, then resource efficiency is improved, but aliasing occurs leading to detection uncertainty
Solution Approach 1:
A neural network is introduced as an intermediary component between the under-sampled radar signals and the object detection process. The neural network processes the aliased data cube to recover accurate target information, enabling the system to maintain high detection reliability while using reduced sampling rates for resource efficiency
Solution Approach 2:
The system creates a virtual representation (data cube) from the under-sampled radar signals and processes this copy through neural network operations. The neural network learns to map the aliased data cube to the unaliased data cube, effectively copying the desired signal characteristics without requiring the original full-rate samples
2Measurement precision
If multiple neural network configurations are trained and evaluated, then detection accuracy is improved, but system complexity increases
Solution Approach 1:
Instead of exhaustively evaluating all possible neural network configurations, the system trains multiple configurations and selects the best performing one based on evaluation metrics. This partial exploration approach achieves high detection accuracy without the excessive complexity of exhaustive search, finding sufficient solutions through targeted training of multiple configurations
Solution Approach 2:
Multiple neural network configurations are trained in advance during an offline phase, and their performance is evaluated beforehand. The best configuration is selected prior to deployment, allowing the system to achieve high detection accuracy without the complexity of real-time configuration selection during operation
Data Source
AI summary
Deep learning in a radar system includes obtaining unaliased time samples from a first radar system. A method includes under-sampling the un-aliased time samples to obtain aliased time samples of a first configuration, matched filtering the un-aliased time samples to obtain an un-aliased data cube and the aliased time samples to obtain an aliased data cube, and using a first neural network to obtain a de-aliased data cube. A first neural network is trained to obtain a trained first neural network. The under-sampling of the un-aliased time samples is repeated to obtain second aliased time samples of a second configuration. The method includes training a second neural network to obtain a trained second neural network, comparing results to choose a selected neural network corresponding with a selected configuration, and using the selected neural network with a second radar system that has the selected configuration to detect one or more objects.