Presentation Title
On Bed Posture Recognition With Pressure Sensor Array System
Presentation Type
Poster Presentation/Art Exihibt
College
College of Natural Sciences
Major
School of Computer Science and Engineering
Location
Event Center BC
Start Date
5-18-2017 11:00 AM
End Date
5-18-2017 12:00 PM
Abstract
Automated on bed posture recognition is of major importance to physical rehabilitation exercises. This paper presents a pressure sensor array system to unobtrusively recognize on bed postures for patients recovering from illness, injury and surgery. In this paper, we propose to use limb clusters to recognize on bed postures instead of using the whole body pressure map since the shapes and positions of limbs represent the postures in nature. To measure the similarities of limb clusters, we propose to use a comprehensive distance metric considering both physical distance and pressure difference. In the recognition phase, a weighted limb based recognition method is applied. The experiments are conducted with 15 subjects ranging from various gender, age and weight. The experiments have demonstrated that our proposed system and method can achieve a 97% recognition rate for typical on bed postures.
On Bed Posture Recognition With Pressure Sensor Array System
Event Center BC
Automated on bed posture recognition is of major importance to physical rehabilitation exercises. This paper presents a pressure sensor array system to unobtrusively recognize on bed postures for patients recovering from illness, injury and surgery. In this paper, we propose to use limb clusters to recognize on bed postures instead of using the whole body pressure map since the shapes and positions of limbs represent the postures in nature. To measure the similarities of limb clusters, we propose to use a comprehensive distance metric considering both physical distance and pressure difference. In the recognition phase, a weighted limb based recognition method is applied. The experiments are conducted with 15 subjects ranging from various gender, age and weight. The experiments have demonstrated that our proposed system and method can achieve a 97% recognition rate for typical on bed postures.