Bayesian Signal Detection for Weak Target Separation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for separating desired electromagnetic signals from noise in receivers, such as in radar and communication systems, face challenges in distinguishing between signal and noise, especially when signal energy levels are low compared to noise, leading to high false alarm rates and difficulty in detecting weak targets.
Innovation Solution
A method involving a Bayesian statistical target model and a customized optimization algorithm that ranks measurements by reliability, calculates initial probability densities, and selects candidate measurements to associate with targets based on increased probability densities, thereby outputting an organized subset of measurements likely related to the same target.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If filtering methods (band pass filter, matched filter) are used to separate desired signal from noise, then signal separation capability is improved, but false alarm rate increases and weak signals remain undetectable when signal energy is low compared to noise
Solution Approach 1:
The patent segments the received signal into multiple independent measurements or samples, then applies statistical analysis to the collection of measurements rather than treating each individually. This segmentation allows the system to accumulate evidence across multiple measurements, improving detection reliability while maintaining low false alarm rates even when individual signal energy levels are low.
Solution Approach 2:
The patent transitions from analyzing signals in the traditional time-frequency domain to incorporating a statistical probability dimension. By calculating probability densities and comparing them against thresholds, the system adds a probabilistic assessment layer that distinguishes true signals from noise more reliably, reducing false alarms while detecting weak signals that filtering alone would miss.
2Object-affected harmful factors
If traditional filtering is applied to reject signals outside the pass band, then noise rejection is improved, but weak desired signals with similar noise characteristics cannot be separated
Solution Approach 1:
The patent performs preliminary statistical characterization of the noise environment by analyzing multiple measurements before making detection decisions. By establishing baseline statistical properties (mean, variance, probability density) from the measurements, the system creates a reference model that helps distinguish weak desired signals from noise that mimics signal characteristics, improving detection accuracy without sacrificing noise rejection.
3Measurement precision
If detection threshold is lowered to detect weaker signals, then detection sensitivity is improved, but false alarm rate increases significantly
Solution Approach 1:
The patent implements a feedback mechanism where the statistical properties derived from multiple measurements are continuously refined and used to adjust detection decisions. The system calculates probability densities for each measurement, compares them against established thresholds, and uses the accumulated statistical evidence to make reliable detection decisions even at low thresholds, thereby maintaining high detection sensitivity without proportionally increasing false alarm rates.
Data Source
AI summary
The invention provides a method and an arrangement for detecting moving point-targets within a large set of noisy measurements. The method is based on Bayesian model selection where the measurements containing targets are modeled with their physical trajectories and the non-target measurements are modeled with the statistical distribution of measurements containing no targets. An a posteriori probability density function is utilized together with a optimization algorithm specifically designed for this problem. Advantages of the invention involve a numerically efficient formulation of the a posteriori probability density, combined with the optimization algorithm. The main applications of the invention are in detecting moving targets within e.g., radar, sonar, lidar and telescopic measurements. The method is also applicable for multi-instrument data fusion.