*3.1. Signal Processing Method for HAR*

The signal processing method of HAR system that was carried out in this research is shown in Figure 3.

**Figure 3.** Signal processing method of HAR system.

According to Figure 3, the raw signals of human daily activities such as walking, walking upstairs, walking downstairs, sitting, standing and laying are acquired from the inertial sensors of smartphones. Then, the raw signals are segmented using sliding windows and filtered using median and band pass Butterworth filters to remove the irrelevant information or noise. Next, feature extraction of time and frequency domain are implemented. Thereafter, all the data of extracted features are classified with ensemble classifiers using base learners of SVM and RF. The performance of classification of is measured using performance evaluation metrics such as precision, recall, F-measure, accuracy and ROC.
