Title:
Intelligent reactions between human body and the environment: Design of smart clothing

Speaker:
Ding Wei (JAIST)

Abstract:
The subject is to use a statistical machine learning in smart clothing that is comfortable to wear, therefore more loose-fitting, which will allow for less than accurate sensors information to be used. This means that someone could be comfortable and still use smart clothing tools.
The aim in smart clothing design is to embed sensors and other electronic devices into clothing to collect signals from the human body and the environment in order to make intelligent reactions accordingly. One of the biggest design challenges is the difficulty of addressing both comfort and sensor accuracy in actual real-world application. Usually skin-tight sensors yield good signal accuracy but an uncomfortable feeling for the wearer whereas comfortable sensors loosely in contact with the skin lead to poor accuracy. In this work, we propose a concept that employs a statistical machine learning approach to enhance the sensor performance by integrating the information from inaccurate non-contact sensors thereby allowing for greater accuracy without making people uncomfortable. We determine the feasibility of the concept by having sensors that are not in direct contact with the skin detect body temperature and then analyzing the results. We develop several types of features from the temperature sensors and integrate them into a non-linear regression model. The experimental results show that the proposed method can improve sensor performance by over 30% as compared to simple average and linear regression methods. This proves the feasibility and potential of machine learning when seeking an optimal trade-off between sensor accuracy and comfort in smart clothing design.