UWB Radar Ground Truth Generation for ML Target Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing machine learning models for radar target detection require extensive and expensive manual data labeling for ground truth, which is unreliable and time-consuming, especially in environments with poor signal-to-noise ratio and multiple reflections, leading to inaccurate target detection.
Innovation Solution
Utilize the signal processing capabilities of UWB radar devices to generate real-time ground truth data by combining ranging and radar measurements, enabling online training of machine learning models using transfer learning techniques to adapt to specific environments.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If manual data labeling with expensive third-party ground truth measurement systems is used, then ground truth accuracy is improved, but training cost and time consumption increase significantly
Solution Approach 1:
The radar device generates its own ground truth data by utilizing its internal signal processing capabilities. The device processes received signals through range Doppler map generation and peak detection algorithms to automatically identify target positions, eliminating the need for external ground truth systems. This self-service approach allows the device to create training datasets independently, significantly reducing both cost and time requirements while maintaining sufficient accuracy for machine learning model training.
Solution Approach 2:
The patent creates simplified representations of ground truth data by generating range Doppler maps from radar signals and extracting peak positions as target location labels. Instead of using complex expensive measurement systems, the patent copies the essential ground truth information (target positions) through signal processing operations that replicate the function of expensive external systems at a fraction of the cost and time.
2Measurement precision
If manual data labeling with expensive third-party ground truth measurement systems is used, then ground truth accuracy is improved, but training cost increases significantly
Solution Approach 1:
The radar device generates its own ground truth data by utilizing its internal signal processing capabilities. The device processes received signals through range Doppler map generation and peak detection algorithms to automatically identify target positions, eliminating the need for external ground truth systems. This self-service approach allows the device to create training datasets independently, significantly reducing both cost and time requirements while maintaining sufficient accuracy for machine learning model training.
Solution Approach 2:
The patent creates simplified representations of ground truth data by generating range Doppler maps from radar signals and extracting peak positions as target location labels. Instead of using complex expensive measurement systems, the patent copies the essential ground truth information (target positions) through signal processing operations that replicate the function of expensive external systems at a fraction of the cost and time.
3Reliability
If peak detection approach is used in poor SNR conditions with many nearby reflections, then target detection reliability is improved, but detection accuracy deteriorates
Solution Approach 1:
The patent applies machine learning models to preprocess and analyze radar signals before final target detection. The trained ML models evaluate range Doppler maps and identify target peaks by learning from training data that includes various SNR conditions and reflection scenarios. This preliminary action of using ML-based peak detection instead of simple thresholding improves both reliability in poor SNR conditions and accuracy in distinguishing true targets from reflections.
Solution Approach 2:
The patent changes the detection parameters by using machine learning models that can adaptively adjust detection thresholds and criteria based on learned patterns from training data. Instead of fixed threshold parameters, the ML models dynamically determine optimal detection parameters based on the specific signal characteristics, improving both reliability in poor SNR and accuracy in distinguishing targets from reflections.
Data Source
AI summary
Embodiments of methods of providing a machine learning (ML) model for detection of radar targets are disclosed that include determining, using ultra-wideband circuitry, range data and location data from a range doppler map corresponding to one or more locations of one or more targets. The method may include determining, using the ML model, estimated target locations based on data from a Doppler range map and iteratively training the ML model based on differences between the estimated target locations and ground truth including a selected one of the location data or the range data.