Radar Motion Characterization via Spectrogram and Particle Filter

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Current radar-based systems for human motion analysis face challenges in reliably distinguishing human motion from other sources, such as animal motion, due to the limitations of spectral processing techniques, which often fail to accurately classify gait patterns and require high computational resources.

Innovation Solution

A system that utilizes radar data to generate spectrograms, which are then processed using a particle filter to estimate the probability density of a state vector including gait classification variables and motion parameters, allowing for real-time characterization and classification of human motion, including distinguishing between walking, running, and other types of motion, using models like the Thalmann and Vignaud models.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If spectral processing techniques are used for radar-based human motion analysis, then the system can process radar data, but it fails to reliably distinguish human motion from other sources such as animal motion

Engineering Contradiction:

Improvemotion classification accuracyVSAvoidhuman motion distinction reliability

Core Design Contradiction:

Measurement precisionVSReliability

Solution Approach 1:

The patent transforms the radar signal from the time domain to the time-frequency domain using spectrogram analysis. This parameter transformation enables the extraction of micro-Doppler features that characterize human gait patterns, allowing reliable distinction between human and non-human motion sources by analyzing frequency-time evolution rather than simple spectral content

Inventive Principle:

Principle #35Parameter changes

Solution Approach 2:

The patent introduces a time-frequency representation (spectrogram) as an additional dimension for motion analysis. By visualizing and processing radar signals in the time-frequency plane rather than单纯的 frequency domain, the system can capture the temporal evolution of motion characteristics, enabling more accurate human motion classification

Inventive Principle:

Principle #17Another dimension (Dimensionality change)

2Productivity

If feature-based approaches are used to extract gait parameters from spectrograms, then computation is fast, but it is difficult to find features providing good correspondence with human motion

Engineering Contradiction:

Improvecomputation speedVSAvoidhuman motion feature correspondence

Core Design Contradiction:

ProductivityVSMeasurement precision

Solution Approach 1:

The patent replaces manual feature engineering with automated machine learning classification. Instead of manually selecting and extracting gait features that correspond to human motion patterns, the system uses trained classifiers (such as support vector machines, neural networks, or random forests) that automatically learn the optimal features from the spectrogram data, achieving both high accuracy and computational efficiency

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

3Measurement precision

If model-based approaches are used for human motion analysis, then gait parameters can be estimated, but the computational resources required are high

Engineering Contradiction:

Improvegait parameter estimation accuracyVSAvoidcomputational resource consumption

Core Design Contradiction:

Measurement precisionVSUse of energy by moving object

Solution Approach 1:

The patent applies a two-stage processing approach where a lightweight classifier first performs preliminary motion classification to identify human targets, and only then are detailed gait parameter estimation models applied to confirmed human subjects. This partial application of computationally intensive models only when necessary reduces overall computational resource consumption while maintaining accurate gait parameter estimation for human motion

Inventive Principle:

Principle #16Partial or excessive action

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

Enables accurate real-time classification and estimation of human motion parameters, reducing computational burden and improving the distinction between human and non-human motion, suitable for low-cost, low-power sensors, enhancing surveillance and security applications.

Implementation Method 1

a radar emitting electromagnetic signals towards a target and receiving echoed signals, output as radar data as a function of time

Methodology Applied

Scientific EffectDoppler effect: Doppler Effect

Data Source

PatentUS9714797B2System for characterizing motion of an individual, notably a human individual, and associated method

Publication Date: 2017.07.25 THALES NEDERLAND BV

AI summary

System (1) for characterizing motion of at least one individual forming a target, comprising a radar emitting electromagnetic signals towards the target and receiving echoed signals, output as radar data, preprocessing means (11) receiving radar data as input and outputting a spectrogram representing time variations of the Doppler spectrum of the radar data, characterizing means (13) configured for outputting a probability density estimation of the state of a target, state means a vector of properties from a target, referred to as state vector, the system being characterized in that said state vector comprises at least one discrete target gait classification variable determining one target gait model among a set of determined target gait models, and a set of discrete and/or continuous motion parameters, the characterizing means comprising estimation means estimating the probability density of said state vector from said spectrogram.