UC Santa Cruz

This thesis proposes a novel approach for the detection and recognition of human

movements. Specically, Activities of Daily Living (ADL) such as standing up,

sitting down and walking are tracked. In order to monitor human motion, a headset

prototype with a single Inertial Measurement Unit (IMU) with accelerometer, gyroscope

and magnetometer has been developed. Accurate estimation of orientation

and velocity through computationally cheap IMU data processing allows a representative

characterization of human behavior patterns in this case. Thanks to tests

performed among dierent subjects, results show that events are robustly detected

89% of the cases and that activities are recognized with a success rate of 92%. What

is more, adequate rejection of noisy movements has also been achieved 94% of the

time. Thus, an ecient and robust solution for human physical activity tracking is

presented.