Falling-down detection method and system based on convolution neural network

Abstract

The invention discloses a falling-down detection method and system based on a convolution neural network. The method comprises the steps that acceleration of three axes, an inclination angle of the body, and a movement direction are collected, and denoising is conducted on data; the data is segmented, labeling and precoding ranking are conducted on the data of each axis, and then discrete fourier transform is constructed; the convolution neural network is built based on the transformed data, training is conducted on the convolution neural network, and a network model of action is obtained; model matching is conducted on the convolution neural network, and whether or not a user falls down is judged. The falling-down detection method and system based on the convolution neural network are suitable for family health and safety monitoring, can identify a complex behavior by the convolution neural network, and makes accurate judgment and a real-time alarm when an aged man falls down.

Description

Technical field [0001] The invention relates to a fall detection method and system based on a convolutional neural network. Background technique [0002] In recent years, social life has developed towards digitization, networking, and intelligence. People have begun to pay attention to intelligent family health and safety monitoring and protection. Among them, the fall detection technology is an indispensable part of the home monitoring system. An accurate and effective fall detection method is of great significance to safety monitoring. It can not only effectively prevent the elderly from falling but also reduce a series of effects after a fall (such as Paralysis, death, etc.) can also reduce the occupation of medical resources, which has far-reaching significance for the entire family and society. [0003] At present, fall detection systems are mainly based on video monitoring, audio monitoring, and wearable sensor monitoring. The cost of video monitoring is relatively large and...

AI Technical Summary

Problems solved by technology

The various detection methods mentioned above have their own advantages and disadvantages. For example, the detection method based on the threshold is simple and low in complexity, but the accuracy is not high; the detection method based on machine learning is complex and can be well identified, but cannot describe some complex behaviors.

Images