Radar Actimetry Device Using Micro-Doppler Spectrograms for Fall Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current motion detection systems, both onboard and remote, face limitations such as user compliance issues, fragility, high false alarm rates, privacy concerns, and inability to provide fine-grained gait analysis for fall detection, particularly in real-time and low-computing power environments.
Innovation Solution
A software radar system emitting between 6 MHz and 250 GHz, utilizing micro-Doppler spectrogram processing and machine learning algorithms for real-time actimetry characterization, which extracts geometrical shape parameters from high-resolution images and employs a simple classification technique like SVM for efficient and accurate fall detection.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If video cameras are used for motion detection, then detailed visual information is obtained, but privacy concerns and occlusion issues arise
Solution Approach 1:
The patent introduces radar as an intermediary sensing modality that indirectly measures motion parameters without capturing visual images. The radar system detects micro-Doppler signatures and kinematic parameters through electromagnetic wave reflection, serving as a mediator between the subject and the detection system, thereby eliminating privacy concerns while maintaining measurement precision.
Solution Approach 2:
The patent replaces the optical/mechanical video camera system with an electromagnetic radar system. This substitution transitions from direct visual capture to indirect electromagnetic measurement, eliminating occlusion issues caused by line-of-sight requirements and providing reliable detection without privacy intrusion.
2Measurement precision
If onboard sensors are used for fall detection, then high fall identification accuracy is achieved, but user compliance and device fragility become problems
Solution Approach 1:
Instead of placing sensors on the subject (onboard approach), the patent inverts the configuration by placing the radar sensor in the environment and detecting the subject passively. This eliminates the need for users to wear or carry devices, dramatically improving compliance while maintaining high detection accuracy through environmental sensing.
3Measurement precision
If complex deep learning methods are used for activity classification, then classification accuracy is improved, but computing power requirements and power consumption increase
Solution Approach 1:
The patent extracts and utilizes only the essential features from radar signals - micro-Doppler signatures and kinematic parameters - rather than processing complete raw signal data. This feature extraction approach maintains classification accuracy while dramatically reducing the computational burden and power consumption compared to processing full deep learning models.
Solution Approach 2:
The patent applies a partial approach by using simplified classification algorithms (SVM, KNN, or simple neural networks) that process only the most critical extracted features rather than implementing full complex deep learning architectures. This partial processing maintains sufficient accuracy for fall detection while significantly reducing energy consumption.
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
The system achieves better performance than existing methods while ensuring smaller hardware implementation and lower power consumption, enabling real-time fall detection and gait analysis with high accuracy.
Implementation Method 1
The most common method for classifying activities is based on the extraction of characteristics resulting from micro-Doppler signatures (spectrogram). The relative motion of the structural components of an object/body generates unique pixel areas in the time-frequency domain of radar return signals.
Implementation Method 2
a radar emitting and receiving radar signals
Data Source
AI summary
The invention discloses a device (1) for characterizing in real time the actimetry of a subject, having: a radar (2) emitting and receiving radar signals, and having a software interface for configuring the shape of the signal emitted; processing and computing means (3) coupled to the radar (2), having a trained classifier (3a) using a database, said processing and computing means (3) being configured to perform in real time: —a capture of color micro-Doppler images (6) having several color channels (R, V, B), each having micro-Doppler signatures (6a) with color pixels the value of which is a function of a reflectivity and a speed of the subject; —a processing of the micro-Doppler images (6) for: computing a so-called monochromatic image having monochromatic pixels, each having a given monochromatic intensity, on the basis of the color pixels of each color micro-Doppler image; transforming the monochromatic image into a binary image by segmentation, according to a binary luminous intensity threshold, of the monochromatic pixels, producing binary pixels, the value of which is dependent on the chromatic intensity of the monochromatic pixel associated with the binary pixel, with respect to the threshold.