Fall detection method based on convolutional neural network and multi-discrimination features

Abstract

The invention provides a tumble detection method based on a convolutional neural network CNN and multi-discrimination features. Firstly, the algorithm uses CNN to extract human body joint point coordinates and skeleton information in a video sequence; secondly, secondary processing is conducted on joint point coordinates by means of the transformation relation between points and vectors, and the included angles between the spine, the left shank and the right shank of the human body and the ground are calculated to serve as a multi-feature extraction structure so that rich falling information can be conveniently collected; and finally, comprehensively analyzing the comparison result of the included angle and the threshold value and the height-width ratio change of the human body calibration frame to realize a fall detection function. In addition, the invention designs an IoT (Internet of Things) system framework for processing the video sequence at the cloud, so as to relieve the problem of insufficient computing power of the user terminal. Compared with a traditional fall detection algorithm, the method has higher accuracy and better universality.

Description

technical field [0001] The invention relates to image processing technology, in particular to a fall detection method based on convolutional neural network and multi-discriminant features. Background technique [0002] According to the research results of the World Health Organization, falls have become the main safety hazard for the elderly over 65 years old, and 30% of the elderly group fall at least once a year. If you can't get timely rescue after falling, it is very easy to cause secondary injury, even life-threatening. Therefore, fall detection has become a research hotspot of scholars at home and abroad. Existing fall detection algorithms can be divided into two categories: non-visual and visual. Non-vision-based methods rely on the advantages of cheap, portable, and easy-to-arrange sensor devices to embed sensors into wearable devices or install them in indoor environments, such as accelerometers, Gyroscopes, vibration sensors, pressure sensors, etc. are used to ob...

AI Technical Summary

Problems solved by technology

However, due to factors such as the need to wear the device for a long time, the limited detection range, and external vibration interference, the detection accuracy is low, and it is difficult to promote it on a large scale.

Images