Radar Motion Classification Using 1-D Time Series for Fast Gesture Recognition

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Existing gesture recognition systems based on camera images face issues with illumination requirements, occlusions, and privacy concerns, while radar-based systems suffer from limited accuracy and slow reaction times.

Innovation Solution

A method using radar measurements to classify motions by obtaining one-dimensional time series from multiple frames, employing a classification algorithm and a convolutional neural network trained with a loss function that penalizes and rewards Euclidean distances to improve accuracy and reaction time.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If camera-based gesture classification is used, then motion recognition capability is provided, but illumination requirements and privacy issues arise

Engineering Contradiction:

Improvemotion recognition capabilityVSAvoidillumination requirements and privacy intrusion

Core Design Contradiction:

ReliabilityVSObject-affected harmful factors

Solution Approach 1:

The patent replaces camera-based optical detection with radar-based electromagnetic wave detection. The radar system transmits electromagnetic signals and processes the reflected signals to extract motion information, eliminating dependency on visible light and providing privacy-preserving gesture recognition through range-Doppler images rather than visual images

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

Solution Approach 2:

The patent changes the detection parameter from optical reflection (camera) to electromagnetic wave reflection (radar). By operating in the microwave frequency range and processing range-Doppler spectra, the system achieves illumination-invariant detection and extracts motion parameters (range, velocity) without capturing visual appearance, thus resolving privacy concerns

Inventive Principle:

Principle #35Parameter changes

2Object-affected harmful factors

If traditional radar-based gesture recognition is used, then privacy is preserved and illumination independence is achieved, but accuracy and reaction time are limited

Engineering Contradiction:

Improveprivacy preservation and illumination independenceVSAvoidrecognition accuracy and reaction time

Core Design Contradiction:

Object-affected harmful factorsVSMeasurement precision

Solution Approach 1:

The patent segments the gesture recognition process into distinct stages: range-Doppler image generation from radar measurements, feature extraction from range-velocity spectra, and classification using a trained neural network. This segmentation allows optimized processing at each stage, improving both accuracy and reaction time while maintaining privacy preservation

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

The patent performs preliminary training of a convolutional neural network classifier using labeled gesture data before deployment. The pre-trained model enables rapid inference with high accuracy, overcoming the slow reaction time limitation of traditional radar-based gesture recognition while maintaining the privacy-preserving advantages

Inventive Principle:

Principle #10Preliminary action

Data Source

PatentUS12529777B2Radar-based motion classification using one or more time series

Publication Date: 2026.01.20 INFINEON TECHNOLOGIES AG

AI summary

In accordance with an embodiment, a computer-implemented method includes obtaining a time sequence of measurement frames of a radar measurement of a scene comprising an object; based on multiple subsequent measurement frames of the time sequence of measurement frames, determining one or more one-dimensional (1-D) time series of respective observables of the radar measurement associated with the object; and based on the one or more 1-D time series, determining a motion class of a motion performed by the object using a classification algorithm.