List of available solutions to detect patient fall from bed

[Afroz Chakure]

Solutions to track patient fall :

1. Using Background Subtractor: Objects of interest can be separated from the background in a video. If the background of a scene remains unchanged the detection of foreground objects would be easy.
2. Feature descriptor algorithms such as histograms of oriented gradients can be trained to identify certain features of the human body and detect when a person is not present in the frame (Adaptive background mixture models could be used for real-time tracking).
3. Other methods include using Kalman Filters, KNNs and using Transfer learning (See Paper 3).
4. Email communication alerts using Mutt email client for Linux.

Resources
Github Links :
Link 1: Video-fall-detection-using-opencv
Link 2: Fall Detection with CNNs and Optical Flow
Link 3: Fall detection ver1 and ver2

Research Papers :
Paper 1
Paper 2
Paper 3

For alerts (Email communication/ Telegram alert) :
Mutt email client
Linux Telegram Messenger CLI

[Afroz Chakure]

Steps to Implement the method:

1. For each frame, it will read and the video will be converted into gray.
2. The background is removed.
3. We find and draw the contours.
4. If the height of the contour is lower than width, it may be a fall and we add 1 to a count, if the count is greater than 10, it will be drawn as a rectangle to the possible person fallen.

Note: The red bounding boxes detect a fall while the green ones detect a person standing in the frame.

Dummy Implementation code:
Link to Code