Millimeter-Wave Radar Activity Classification for Accurate Fall Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing fall detection systems for the elderly face challenges such as high cost, privacy concerns, discomfort, and susceptibility to occluded or low illumination scenarios, and struggle to differentiate between similar activities like sitting and falling, leading to false positives.
Innovation Solution
A millimeter-wave radar system uses a graph encoder to encode radar point clouds and a cadence-velocity diagram to classify activities by extracting relationships and periodicity of body parts, reducing false positives through a combination of graph convolutional neural networks and long short-term memory networks.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If vision systems are used for elderly fall detection, then detection capability is improved, but cost increases and privacy concerns arise
Solution Approach 1:
The patent replaces vision-based mechanical/optical detection systems with radar-based electromagnetic wave detection. The radar system uses electromagnetic signals to detect falls, eliminating the need for cameras and visual processing infrastructure, thereby reducing cost and privacy concerns while maintaining detection reliability
Solution Approach 2:
The patent introduces radar technology as an intermediary detection method between direct visual contact and wearable sensors. The radar system acts as a non-contact intermediary that can detect falls through electromagnetic waves without requiring direct visual line-of-sight or physical contact with the elderly person
2Measurement precision
If wearable systems are used for elderly fall detection, then detection accuracy is improved, but comfort decreases and freedom of movement is restricted
Solution Approach 1:
The patent extracts the detection function from the elderly person's body by using an external radar system instead of wearable sensors. The fall detection capability is taken out of the wearable form factor and implemented as a standalone environmental sensing system, eliminating discomfort and movement restrictions
Solution Approach 2:
The radar system provides self-service detection by automatically monitoring the environment and identifying falls without requiring the elderly person to wear or interact with any detection devices. The system serves itself by using environmental electromagnetic waves to perform detection
3Ease of manufacture
If traditional radar methods are used for activity classification, then implementation simplicity is maintained, but classification accuracy decreases leading to false positives
Solution Approach 1:
The patent combines multiple signal processing techniques (range-Doppler imaging, point cloud generation, graph convolutional neural networks, and cadence-velocity diagrams) to create a composite analysis framework. This composite approach integrates multiple features and processing stages to improve classification accuracy while maintaining reasonable implementation complexity
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 effectively distinguishes between activities like falling and sitting, providing accurate fall detection with low false positives, offering cost-effective, privacy-respecting, and unrestricted monitoring.
Implementation Method 1
receiving raw data for a scene comprising a target from a millimeter-wave radar sensor
Implementation Method 2
generating a first cadence velocity diagram indicative of a periodicity of movement of one or more parts of the target
Data Source
Figure 1
Figure 2~4
Figure 5~6B
AI summary
In an embodiment, a method includes: receiving raw data from a millimeter-wave radar sensor; generating a first radar-Doppler image based on the raw data; generating a first radar point cloud based on the first radar-Doppler image; using a graph encoder to generate a first graph representation vector indicative of one or more relationships between two or more parts of the target based on the first radar point cloud; generating a first cadence velocity diagram indicative of a periodicity of movement of one or more parts of the target based on the first radar-Doppler image; and classifying an activity of a target based on the first graph representation vector and the first cadence velocity diagram.