MIMO Radar Target Count Estimation via Statistical Moments
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing radar systems, particularly multiple-input multiple-output (MIMO) radar systems, face challenges in estimating the number of targets in a range-Doppler bin, which is essential for advanced angle estimation algorithms like MUSIC and ESPRIT, due to computationally intensive methods that are not suitable for automotive applications with high-resolution radars.
Innovation Solution
A neural network trained machine learning module is used to estimate the number of detected objects by processing frequency domain data from each virtual channel, reducing the data input and processing load, and capable of being installed in automobiles, even with a large number of virtual channels.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If linear algebra methods (covariance matrix calculation, eigenvalue decomposition) are used to estimate the number of targets, then measurement precision is improved, but device complexity and processing time increase significantly
Solution Approach 1:
The patent changes the fundamental parameters of the estimation method from traditional linear algebra approaches (covariance matrix, eigenvalue decomposition) to a statistical moment-based approach. By using raw moments and central moments of the signal distribution, the method achieves accurate target number estimation without requiring complex matrix operations, thus reducing computational complexity while maintaining measurement precision
Solution Approach 2:
The patent employs a lightweight estimation algorithm that can be rapidly computed and discarded for each new radar frame, replacing the need for heavy, persistent matrix decompositions. The moment-based method uses simple statistical calculations that are computationally inexpensive and can be quickly recalculated as new data arrives, making the system more suitable for real-time automotive applications
2Measurement precision
If multiple snapshots of data are acquired to increase signal-to-noise ratio, then measurement precision is improved, but loss of time increases
Solution Approach 1:
The patent performs preliminary statistical calculations (raw moments and central moments) directly from the current radar data frame without requiring multiple snapshots. By pre-computing these statistical parameters from the available data, the method achieves noise robustness equivalent to multiple snapshots while maintaining real-time processing capability and avoiding time delays
Solution Approach 2:
The patent skips the traditional step of acquiring multiple data snapshots by rushing through to a direct moment-based estimation approach. The method processes the current single frame of radar data immediately by calculating statistical moments, thereby achieving fast processing without the time-consuming data accumulation step required by conventional methods
3Measurement precision
If eigenvalue decomposition on large matrices (e.g., 128×128) is performed, then measurement precision is improved, but productivity decreases
Solution Approach 1:
The patent extracts only the essential statistical information (raw moments and central moments) from the radar data, removing the need to perform full eigenvalue decomposition on large covariance matrices. By extracting and using only these key statistical parameters, the method achieves accurate target detection while dramatically reducing the computational burden and improving processing speed
Solution Approach 2:
The patent segments the complex target detection problem into simpler statistical moment calculations rather than treating it as a single large matrix decomposition problem. By dividing the task into computing individual moments (which are simple statistical summaries) rather than decomposing a huge matrix, the method achieves the same detection accuracy with much higher processing productivity
Data Source
AI summary
According to at least one embodiment, a MIMO radar arrangement includes a radar receiver configured to generate radar reception data from radio receive signals received by a plurality of radar receive antennas. The arrangement further includes one or more signal processors configured to: generate frequency domain data for a range-Doppler bin based on the radar reception data and determine one or more peaks from the generated frequency domain data. The radar arrangement further includes a trained machine learning module configured to generate, using frequency domain data corresponding to each of the of the one or more determined peaks as input, one or more output values indicating a number of detected objects within each range-Doppler bin.