Automatic detection by a wearable camera
Inventors
Velipasalar, Senem • ALMAGAMBETOV, Akhan • CASARES, Mauricio
Assignees
Syracuse University • National Science Foundation NSF
Publication Number
US-9571723-B2
Publication Date
2017-02-14
Expiration Date
2032-11-16
Interested in licensing this patent?
MTEC can help explore whether this patent might be available for licensing for your application.
Abstract
There is set forth herein a system including a camera device. In one embodiment the system is operative to perform image processing for detection of an event involving a human subject. There is set forth herein in one embodiment, a camera equipped system employed for fall detection.
Core Innovation
The invention sets forth a system including a wearable camera device that performs image processing to detect events involving a human subject, particularly fall detection. The camera device processes images by comparing a subsequently captured image to an earlier captured image, focusing on analyzing the surroundings of the subject rather than the subject themselves. The device is capable of wirelessly transmitting messages containing event notifications and images to external destinations, such as emergency responders.
The problem being addressed is the high incidence and severe consequences of falls among elderly patients, which are a leading cause of injury and death in that population. Existing fall detection methods face challenges like trade-offs among detection accuracy, processing power, and user privacy. Current devices often require the user to press a button to call for help, which is ineffective if the user is unconscious after a fall. Fixed cameras violate privacy and limit monitoring to specific areas, while accelerometer-based devices face accuracy or intrusiveness issues.
The invention solves these issues by using a wearable embedded wireless smart camera that captures images not of the wearer, but of their surroundings, protecting privacy and enabling monitoring everywhere the subject goes, including outdoors. Image processing onboard the device uses Histograms of Oriented Gradients (HOG) and correlation-based dissimilarity distances to classify human actions including walking, sitting, laying down, and falling, discriminating falls from other activities. The system only transmits data when a fall or target event is detected, reducing privacy concerns and data transmission needs.
Claims Coverage
The patent includes three independent claims covering a wearable camera device with image processing capabilities, a method using such a device, and a computer program product implementing the detection method.
Wearable camera device adapted for event detection and messaging
A camera device wearable by a human that captures images of the surroundings, processes images to detect human subject actions by comparing sequential images, and wirelessly transmits messages upon event detection, including events such as laying down, sitting down, and falling.
Method for detecting human actions using a wearable camera directed away from the subject
Positioning a camera on a human subject so captured images represent surroundings, processing these images to detect actions—including laying down, sitting down, and falling—by comparing subsequent images to prior ones, and outputting human-observable indicators responsively.
Computer program product for action detection from wearable camera images
Software stored on a non-transitory medium, executable on a processor, that processes images representing surroundings of a human subject captured by a wearable camera to determine occurrences of actions including laying down, sitting down, and falling, and outputs a human-observable indicator correspondingly.
The claims collectively cover a wearable camera system, the associated detection methods based on image processing of surrounding scenes, and software implementing these methods, focusing on fall and related event detection with privacy-preserving features and event-responsive communication.
Stated Advantages
Privacy is preserved as the wearable camera is directed away from the subject and images are only transmitted upon event detection.
The system allows continuous monitoring wherever the subject goes, including outdoors, unlike fixed cameras limited to specific locations.
Low power and low cost are achieved by using embedded smart camera hardware optimized for efficient image processing.
The detection algorithm effectively distinguishes falls from other similar activities such as sitting and laying down, reducing false alarms.
The system provides timely alerts with images sent to emergency personnel, aiding in rapid location and assistance of the subject.
Documented Applications
Fall detection and alert for elderly patients to facilitate immediate medical assistance upon a fall.
Monitoring and classification of human activities including walking, sitting, and laying down for health assessment.
Privacy-preserving monitoring via wearable cameras that do not capture the subject but their surroundings.
Early diagnosis and statistics gathering by analyzing time spent in various activities or rooms to detect health deterioration.
Sending images and alerts to emergency response personnel to improve response and subject location.
Interested in licensing this patent?