i-POIS (Intelligent Posture Identification System)

Date of Publication

2011

Document Type

Bachelor's Thesis

Degree Name

Bachelor of Science in Computer Science

College

College of Computer Studies

Department/Unit

Computer Science

Thesis Adviser

Clement Y. Ong

Defense Panel Member

Gregory G. Cu
Merlin Teodosia C. Suarez
Ana Marian M. Pedro

Abstract/Summary

Sitting postures have been correlated with psychological aspects of a person such as emotions and affective states. It has also been further utilized in fields such as human-computer interaction and ergonomics. Existing posture identification systems use pressure sensors that require several connections which keep their resolution limited in practice (under 100x100). Moreover, pressure sensors are expensive.

The Intelligent Posture Identification System (i-POIS) is a low-cost posture identification system that detects nine sitting postures through Frustrated Total Internal Reflection (FTIR): Sitting Upright, Leaning Forward, Leaning Forward Left, Leaning Forward Right, Leaning Back, Leaning Back Left, Leaning Back Right, Slumping Back, and Sitting on the Edge. Trough FTIR, pressure distribution information from both seat pan and back rest of the chair are captured by low-cost cameras. The system is implemented on a four-legged chair. Image processing techniques are used to extract blob features and several machine learning algorithms in the WEKA package were tested to classify these into their corresponding posture labels. Feature selection is also performed to draw out the most important attributes that the best describe a sitting posture and to avoid the problem of over-fitting.

The training set of the system is composed of data from 50 subjects, amounting to a total of 2250 instances. The accuracy rate of the system based on the training set is done using WEKA. Correlation-Based Feature Selection was used as the attribute selector and Greedy Stepwise was used as search algorithm. KStar Algorithm, along with attribute selection, produced the best results with the detection rate of 87.7778%. The confusion matrix on the other hand, shows that there is confusion between the classification of Sitting Upright and Leaning Forward. Real time performance show that there are confusions in terms of postures that use the back rest of the chair, since a user’s lumbar region has lighter weight than the gluteal region. Moreover, the results of the system depend on the following factors: the person’s weight and height, clothing used, and the way he/she perceives a sitting posture.

Abstract Format

html

Language

English

Format

Print

Accession Number

TU16029

Shelf Location

Archives, The Learning Commons, 12F, Henry Sy Sr. Hall

Physical Description

1 v. (various foliations) ; 28 cm.

Keywords

Sitting position; Posture; Computer vision; Image processing; Pattern recognition systems

This document is currently not available here.

Share

COinS