Looking for breakthrough ideas for innovation challenges? Try Patsnap Eureka!

Fall detection method and system based on wearable sensor and video monitoring

A technology for video surveillance and detection methods, applied in the fields of signal processing and deep learning, can solve the problems of detecting personnel falling, not processing sensor data at the same time, etc., achieving the effect of fast running speed, low overall equipment cost, and low memory usage

Pending Publication Date: 2022-02-18
SHANDONG UNIV
View PDF0 Cites 0 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

In some other places, people do not wear wearable terminals, but surveillance cameras are installed in the surrounding space, and it is generally only based on the surveillance video to judge whether a person has fallen
In other places, people are equipped with wearable terminals, and surveillance cameras are also installed in surrounding places. At present, there is no model and system that can simultaneously process sensor data and video data to detect people's falls, especially for the coexistence of many people. Scenarios, in the case of limited computing resources, how to accurately detect people's falls is an urgent problem to be solved

Method used

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
View more

Image

Smart Image Click on the blue labels to locate them in the text.
Viewing Examples
Smart Image
  • Fall detection method and system based on wearable sensor and video monitoring
  • Fall detection method and system based on wearable sensor and video monitoring
  • Fall detection method and system based on wearable sensor and video monitoring

Examples

Experimental program
Comparison scheme
Effect test

Embodiment 1

[0046] A fall detection method based on wearable sensors and video surveillance consists of cascading two independent models, such as figure 1 As shown, they are respectively based on the Gated Recurrent Unit (Gated Recurrent Unit, GRU) sensor data analysis one-stage GRU model, and the two-stage TSM surveillance video processing model based on the time shift module TSM surveillance video analysis; in the first stage In the GRU model, GRU is a variant of LSTM, which can maintain comparable performance to LSTM while reducing the amount of calculation. In the two-stage TSM surveillance video processing model, an adaptive channel shift strategy is added to the time shift module TSM combined with channel attention, and the improved TSM is inserted into the improved feature combined with the convolutional block attention mechanism The extraction network CSPDarknet53 is used to improve the accuracy and robustness of video detection.

[0047] A fall detection method based on wearable...

Embodiment 2

[0055] According to a kind of fall detection method based on wearable sensor and video monitoring described in embodiment 1, its difference is:

[0056] In step 2, start the first-stage algorithm of fall detection based on GRU for analysis, output the result in a probabilistic manner, and initially judge whether a fall behavior occurs, specifically refers to: preprocessing the posture parameters to be detected obtained in step 1 (according to step (3 ) preprocessing method) and then input the trained one-stage GRU model for detection, wherein, the trained one-stage GRU model construction and training process are as follows:

[0057] (1) Build a one-stage GRU model: a one-stage GRU model includes multi-layer GRU, and each layer of GRU adopts 100 memory units; in the selection of the model, the present invention considers that this is the first-stage algorithm, and it is necessary to ensure a low amount of calculation At the same time, to ensure good accuracy, a multi-layer GRU ...

Embodiment 3

[0072] According to a kind of fall detection method based on wearable sensor and video monitoring described in embodiment 1, its difference is:

[0073] In step 4, use the two-stage TSM surveillance video analysis algorithm to judge whether a fall occurs, specifically refers to: input the trained two-stage TSM after preprocessing the surveillance video to be detected (according to the preprocessing method of step c) The monitoring video processing model is used for detection. Among them, the trained two-stage TSM monitoring video processing model construction and training process are as follows:

[0074] a. Construct a two-stage TSM surveillance video processing model:

[0075] like Figure 4As shown, the two-stage TSM surveillance video processing model includes a series of CMBL and CSP1-X. Through a series of CMBL operations and CSP1-X operations of different sizes, the spatio-temporal features are extracted repeatedly, and the feature map is pooled by SPP spatial pyramid p...

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More

PUM

No PUM Login to View More

Abstract

The invention relates to a fall detection method and system based on a wearable sensor and video monitoring. The system is formed by cascading two independent models which are respectively a sensing data analysis first-stage GRU model based on a gated loop unit network and a monitoring video analysis second-stage TSM monitoring video processing model based on an improved time shift module TSM. In the first-stage GRU model, GRU is a variant of LSTM and can maintain performance comparable to LSTM while reducing computations. In the second-stage TSM monitoring video processing model, an improved TSM is inserted into a feature extraction network CSPDarknet53 which can be comparable to Resnet152, and a target area in an extracted feature map is enhanced by using convolution block attention (CBAM), so that the accuracy and robustness of video detection are improved.

Description

technical field [0001] The invention relates to a fall detection method and system based on wearable sensors and video monitoring, and belongs to the technical fields of deep learning and signal processing. Background technique [0002] Hospitals, nursing homes, home care and other fields have continuously increased the demand for detection of accidental falls. For patients, the elderly and other specific people, when a fall occurs, it is necessary to detect it in time and provide corresponding assistance, otherwise serious physical damage may occur. In recent years, the fall detection technology based on deep learning has been continuously developed. Long Short-Term Memory (LSTM) and its variants have been widely used in fall detection based on wearable sensor data. It can effectively Solve the long-term dependence problem in time series analysis and improve the accuracy of detection. The Temporal Shift Module (TSM) adopts the complexity of 2D convolution to achieve the pe...

Claims

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More

Application Information

Patent Timeline
no application Login to View More
Patent Type & Authority Applications(China)
IPC IPC(8): G06V40/20G06K9/00G06N3/04G06N3/08G06V20/40G06V20/52
CPCG06N3/04G06N3/08G06F2218/12
Inventor 翟超倪志祥郑丽娜
Owner SHANDONG UNIV
Who we serve
  • R&D Engineer
  • R&D Manager
  • IP Professional
Why Patsnap Eureka
  • Industry Leading Data Capabilities
  • Powerful AI technology
  • Patent DNA Extraction
Social media
Patsnap Eureka Blog
Learn More
PatSnap group products

Browse by: Latest US Patents, China's latest patents, Technical Efficacy Thesaurus, Application Domain, Technology Topic, Popular Technical Reports.

© 2024 PatSnap. All rights reserved.Legal|Privacy policy|Modern Slavery Act Transparency Statement|Sitemap|About US| Contact US: help@patsnap.com