Skip to Main Content
 

Global Search Box

 
 
 
 

Files

File List


17373.pdf (12.4 MB)

ETD Abstract Container

Abstract Header

A study on context driven human activity recognition framework

Chakraborty, Shatakshi
http://orcid.org/0000-0001-5227-4140
http://rave.ohiolink.edu/etdc/view?acc_num=ucin1439308572

Abstract Details

2015, MS, University of Cincinnati, Engineering and Applied Science: Computer Science.
In recent world, human activity recognition has drawn much attention in the field of human computer interaction. There is a growing demand of activity recognition in different areas of everyday living, such as health-care systems like patient health monitoring, home-based rehabilitation, entertainment, and many more. In this research, we are aiming to use activity recognition theory in health-care system to monitor patient behavior during the waiting time at clinical visits. In today’s health-care system, patients wait about 22 minutes on average in doctor’s offices, and more than four hours in emergency departments. As wait time increases, patient satisfaction drops. With a growing consumer-mindedness of instant gratification or satisfaction, health care providers or hospitals are looking ways to improve productivity, like shortening each patient’s path through the health care system, perhaps, adopting measures such as clinics using kiosks, and not reception desks, speedier check-in for returning patients, and taking measures to funnel visitors to the appropriate part of the clinic or hospital when appointments have been arranged earlier, while providing more attentive face-to-face care to those who are first timers to the system and in need. The purpose of this study is to investigate a computer-based means to obtain useful data on typical human behaviors during visits to clinics. A framework to implement the technology to study human behavior has been proposed by Tao Ma[15] recently. In his four-layer hierarchical framework, computer vision is used to study and understand human behavior through body movements. We explore a second framework developed by Saguna et al. [25] which uses probability theory and statistical learning methods to discover complex activity signatures. Additional modalities of information, such as speech, facial expressions, time-based contextual information can also be incorporated to interpret various human behaviors and elicit the cognitive processes used in analyzing the workflow of normal activities. Appertaining to the vast area of human behavior study, a particularly interesting setting to study body movements in analyzing human behavior is in a common venue of our daily life, in particular, typical visit to a doctor’s clinic or a hospital. No study has been conducted thus far, to our best knowledge, to understand patient’s satisfaction during a clinical visit based on real-time body movements and gestures of the patients in the waiting lounge. In this research work, we further explore the existing framework to represent the small cosmos of waiting rooms in clinic, and to apply mathematical models to derive individual complex behavior, often found in this setting. The first section of the research explores the background and theory of the two frameworks: First, is the 4-layered hierarchical framework, namely the 4 layers are: 1) Feature extraction, 2) Behavior classification, 3) Individual behavior sequence, and 4) Social interaction. Second, is the Context-Driven Activity Theory, where, we created a complex activity dataset with 12 activities performed by a patient which depicts some very simple and common activities that a patient involves in during the waiting time. We then apply the frameworks in order to validate the performance of the existing works. This study discusses how the results can be particularly beneficial for understanding a patient’s experience and make recommendations for improving the quality of patient experience in the United States. Ultimate objectives for such set of collected data and analysis include making work flow in clinics or hospitals more efficient, optimizing office staff functions, and increasing face-to-face time between physicians and patients.
Chia Han, Ph.D. (Committee Chair)
William Wee, Ph.D. (Committee Member)
Xuefu Zhou, Ph.D. (Committee Member)
110 p.
Computer Science
context driven; activity recognition; complex activities; concurrent activities; interleaved activity

Recommended Citations

Citations

  • Chakraborty, S. (2015). A study on context driven human activity recognition framework [Master's thesis, University of Cincinnati]. OhioLINK Electronic Theses and Dissertations Center. http://rave.ohiolink.edu/etdc/view?acc_num=ucin1439308572

    APA Style (7th edition)

  • Chakraborty, Shatakshi. A study on context driven human activity recognition framework. 2015. University of Cincinnati, Master's thesis. OhioLINK Electronic Theses and Dissertations Center, http://rave.ohiolink.edu/etdc/view?acc_num=ucin1439308572.

    MLA Style (8th edition)

  • Chakraborty, Shatakshi. "A study on context driven human activity recognition framework." Master's thesis, University of Cincinnati, 2015. http://rave.ohiolink.edu/etdc/view?acc_num=ucin1439308572

    Chicago Manual of Style (17th edition)

ucin1439308572
640
© 2015, all rights reserved.
This open access ETD is published by University of Cincinnati and OhioLINK.