Range-Doppler Neural Processing for Unambiguous Road User Velocity
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current radar sensors using range-Doppler matrices struggle to accurately determine object parameters like velocity and classification due to ambiguity and limited unambiguous range, which hinders the differentiation of road users such as pedestrians, cyclists, and vehicles, especially in traffic monitoring and classification tasks.
Innovation Solution
The method involves generating a range-Doppler matrix and transferring it to a neural network for parameter identification, allowing for the expansion of the useful Doppler evaluation range and eliminating ambiguities by interpreting zones of reflected energy rather than individual local maxima, enabling unambiguous velocity determination and classification.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If different types of frequency ramp are transmitted to increase useful Doppler evaluation range, then Doppler unambiguous range increases, but device complexity increases
Solution Approach 1:
The patent makes the radar system multi-functional by enabling it to operate with different frequency ramp configurations depending on the application requirements. The same hardware infrastructure can perform both simple distance measurement and extended Doppler evaluation by adjusting the transmission signal parameters, eliminating the need for separate specialized systems for different measurement ranges.
Solution Approach 2:
The patent dynamically adjusts transmission signal parameters (frequency ramp type, slope, deviation) based on the desired measurement range and target characteristics. By changing these parameters adaptively, the system achieves extended Doppler evaluation range when needed while maintaining simpler operation for basic applications, thus managing device complexity through flexible parameter control rather than fixed complex hardware.
2Adaptability or versatility
If neural networks are used to classify objects based on frequency spectrum, then classification capability improves, but ambiguity in measurement signals prevents accurate learning
Solution Approach 1:
The patent performs preliminary processing of the radar signals to create a more informative representation before feeding data to the neural network. By pre-processing the range-Doppler matrix to highlight zones of reflected energy and their spatial distribution characteristics, the system prepares the data in a form that reduces ambiguity and enhances the discriminative features available for classification, allowing the neural network to learn more effectively.
Solution Approach 2:
The patent introduces an intermediary processing stage that transforms the ambiguous frequency spectrum data into a more structured representation emphasizing spatial-temporal energy distribution patterns. This intermediary representation serves as a bridge between the raw radar signals and the neural network classifier, preserving critical information that would otherwise be lost in ambiguous spectral data and enabling accurate object classification.
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
This approach allows for the complete elimination of ambiguities in the Doppler dimension, enabling accurate classification and velocity determination of road users, including trucks, within the range of interest, and improves the angular resolution for orientation and direction of travel, enhancing traffic monitoring and classification capabilities.
Implementation Method 1
the velocity, determined via the Doppler shift of the reflected radar waves
Implementation Method 2
A sensor that can be used to conduct such a method emits a transmission signal in the form of radar beams
Data Source
AI summary
The invention relates to a method for determining at least one parameter of an object, wherein the method comprises the following steps:a. provision of a range-Doppler matrix,b. transfer of at least one part of the range-Doppler matrix to a neural network andc. identification of the at least one parameter by the neural network.