Text

theses

The basic necessity of sleep in our life is critically important to ensure our wellbeing. Sufficient sleep of good quality is highly desired in order to have enough energy to live. One of the main factors to measure sleep quality is the amount of body motion during sleep. In-bed motion detection is an important technique that can enable an array of applications, among which are sleep monitoring and abnormal movement detection. When detection is combined with classification, it can be used to detect, notify, and recognize specific events, enabling us to focus on critical tasks. In this study, we present a low-cost, low-overhead, and highly robust system for in-bed movement detection and classification that uses low-end load cells. By observing the forces sensed by the load cells, placed under each bed leg, we can detect many different types of movements, and further classify them as big or small depending on magnitude of the force changes on the load cells. We have designed three different features, which we refer to as Log-Peak, Energy-Peak, and ZeroX-Valley, that can effectively extract body movement signals from load cell data that is collected through wireless links in an energy-efficient manner. After establishing feature values, we employ a simple threshold-based algorithm to detect and classify movements. We have conducted a thorough evaluation, that involves collecting data from 30 subjects who perform 27 pre-defined movements in an experiment. By comparing our detection and classification results against the ground truth captured by a video camera, we show the Log-Peak strategy can detect these 27 types of movements at an error rate of 6.3% while classifying them as big or small movements at an error rate of 4.2%. In the second part of this dissertation, we set out to achieve much finer body motion classification. Towards this goal, we define 9 classes of movements, and design a machine learning algorithm using Support Vector Machine (SVM) and Random Forest techniques to classify a movement into one of these 9 classes. In this way, we can find out which body parts are involved in every movement. For every movement, we have extracted 24 features and used them in our model. This movement classification system was evaluated on data collected from 40 subjects who performed 35 predefined movements in each experiment. The accuracy of our model is not the same for all classes of movements. On average, it correctly classifies 90% of movements. This model can be used conveniently for long-term home monitoring. To improve the classification accuracy, we investigate more machine learning techniques. We use Random Forest and XGBoost as additional classification tools. We apply multiple tree topologies for each technique to reach their best results. After examining various combinations, we achieve the final classification accuracy of 91.5%. Lastly, another in-bed motion detection system is built. We use a geophone sensor to detect body motions in bed, which we call MotionPhone. MotionPhone is more accurate in detecting motion but not efficient for classification purposes. We thus believe combining these two systems can give us better results. Both systems are unobtrusive, low-cost, and private, which can thus enable a large array of important applications.

ETD_8217

Ph.D.

Includes bibliographical references

by Musaab Adil Alaziz

doi:10.7282/T33X89QH

ETD doctoral

The author owns the copyright to this work.