Real-time fall monitoring using body sensor network

Name: Natthapon Pannurat

keyword: Body sensor network

; Pre-fall

; Fall detection

; Accelerometer

; Sensor position

; Activity information

; Temporal reasoning

; Signal processing

Abstract: Falls represent one of the leading causes of deaths and injuries in the elderly population. More than one-third of home-dwelling people aged 65 or above and two-thirds of those in residential cares fall one or more times each year. More than two-thirds of people who have experienced falls are prone to falling again. The psychological consequences often lead to decreased mobility and independence among elderly population. Falls can occur on level surfaces, mostly in living rooms, bedrooms, kitchens, bathrooms, or hallways. Damages caused by falls include tissue injuries, lacerations, joint dislocations, bone fractures, and head trauma. This dissertation presents comprehensive reviews of state-of-the-art fall monitoring systems and a hybrid framework for real-time fall monitoring. The review focuses on fall monitoring studies using wearable sensors. The proposed fall monitoring framework incorporates the strengths of both machine learning and rule-based knowledge representation with an add-on mechanism for performing temporal reasoning using a standard rule-based inference engine. The machine learning component allows low-level classification tasks, e.g., activity classification, to be carried out efficiently for different sensor positions. The model parameters can be systematically tailored to different data sources regardless of the difference in coordinate systems. Rule-based representation facilitates the use of high-level information and incorporation of expert knowledge into a decision making process. Through the rule sets in the knowledge base, this dissertation illustrates how the activity information can be used to enhance the accuracy of fall detection. The proposed framework was validated for fall detection performance, false alarm evaluation, and comparison with a highly-cited baseline method. The fall detection performance was evaluated based on a large number of fall types and fall-like activities, i.e., 14 types of falls and 11 types of activities of daily living (ADLs). The dataset was acquired from 16 young subjects (12 males and 4 females). The false alarm rate was evaluated using a dataset acquired from 13 subjects, who were asked to perform a sequence of 4 lying postures, i.e., supine, lying on the left side, prone, and lying on the right side, and 4 ADLs, i.e., sitting, standing, walking, and jumping. All subjects were asked to place tri-axial accelerometer sensors on 8 different parts of subjects’ bodies

Thammasat University. Thammasat University Library

Address: BANGKOK

Email: preserv@tu.ac.th

Name: Ekawit Nantajeewarawat

Role: advisor

Name: Surapa Thiemjarus

Role: co-advisor

Created: 2016

Modified: 2021-06-29

Issued: 2021-06-29

วิทยานิพนธ์/Thesis

application/pdf

DOI: 10.14457/TU.the.2016.1683

eng

DegreeName: Doctor of Philosophy

Level: Doctoral Degree

Descipline: Technology

Grantor: Thammasat University

©copyrights Thammasat University

ลำดับที่.	ชื่อแฟ้มข้อมูล	ขนาดแฟ้มข้อมูล	จำนวนเข้าถึง	วัน-เวลาเข้าถึงล่าสุด
1	3052natthapon.pdf	7.03 MB	12	2026-05-26 18:16:01

ใช้เวลา

0.026576 วินาที