Inertial sensors (IMU) are widely used in the field of human activity recognition (HAR), since this source of information is the most informative time series among non-visual datasets. HAR researchers are actively exploring other approaches such as different feature extraction methods, machine learning models and classifiers, different sources of signals and sensor positioning to im- prove the performance of HAR systems.
Human physical activities have a significant impact on human body, specifically heart activity and oxygen delivery, thus, we explore heart activity related bio-signals to check if these signals are advantageous in the field of HAR research. In this thesis, we investigate the impact of combining bio-signals with dataset acquired from accelerometer on recognizing human daily activities. To achieve this aim, we used PPG-DaLiA dataset consisting of 3D-accelerometer (3D-ACC), electrocardiogram (ECG), photoplethysmogram (PPG) signals acquired from 15 individuals while per- forming daily activities. We extracted hand-crafted time and frequency domain features, then we applied correlation-based feature selection approach to reduce feature-set dimensionality. After introducing early fusion scenarios, we trained and tested random forest models with subject-dependent and subject-independent setups. Our results indicate that combining features extracted from 3D- ACC signal with ECG signal improve the classifier’s performance F1-scores by 2.72% and 3.00% (from 94.07% to 96.80%, and 83.16% to 86.17%) for subject-dependent and subject-independent approaches, respectively.