Publication Date

Spring 2019

Degree Type

Master's Project

Degree Name

Master of Science (MS)

Department

Computer Science

First Advisor

Robert Chun

Second Advisor

Nada Attar

Third Advisor

Adil Khan

Keywords

Human activity recognition, machine learning, mobile sensors, accelerometer, gyroscope, feature selection, RNN

Abstract

It is a significant technical and computational task to provide precise information regarding the activity performed by a human and find patterns of their behavior. Countless applications can be molded and various problems in domains of virtual reality, health and medical, entertainment and security can be solved with advancements in human activity recognition (HAR) systems. HAR is an active field for research for more than a decade, but certain aspects need to be addressed to improve the system and revolutionize the way humans interact with smartphones. This research provides a holistic view of human activity recognition system architecture and discusses various problems associated with the design aspects. It further attempts to showcase the reduction in computational cost and significant achievement in accuracy by methods of feature selection. It also attempts to introduce the use of recurrent neural networks to learn features from the long sequences of time series data, which can contribute towards improving accuracy and reducing dependency on domain knowledge for feature extraction and engineering.

Download

Share

COinS