*2.2. Apparatus*

The Empatica E4 wristband (EDA sensor) was used for physiological signal collection in our experiment. The wristband [42] is a wearable and wireless device designed for comfortable, continuous, and real-time data acquisition in daily life. It contains three sensors: (1) PPG sensor that detects cardiovascular features such as blood volume pulse and heart rate variability, (2) EDA sensor that measures an arousal of a sympathetic nervous system and derive features related to stress, engagement, and excitement, and (3) three-axis accelerometer that captures motion-based activity. The wristband operates either in a streaming mode for real-time data viewing on a mobile device, using Bluetooth Low Energy or in recording mode, using its internal memory. In our study, EDA data were collected

using the recording mode. The EDA sensor has the following characteristics: sampling frequency of 4 Hz, resolution of one digit –900 pSiemens, range of 0.01–100 μSiemens, and alternating current (8 Hz frequency) with a maximum peak to a peak value of 100 μAmps (at 100 μSiemens) [37].

To record the vehicle movements and on-road situations, a standard on-board diagnostic system (OBD-II) was used. OBD-II is a computer-based system built into modern passenger cars, which monitors emission-related controls, and performance of the engine and also detects malfunctions. OBD-II systems provide access to the health information of a vehicle along with numerous parameters and sensors from the engine control unit (ECU). The OBD-II system offers valuable information, including diagnostic trouble codes, when troubleshooting problems [43]. The technical characteristics of OBD-II can be found in the standard signaling protocols for interfaces.

A mid-sized sedan, Hyundai Grandeur (Azera in the U.S.) 6th generation (Hybrid), was used in this study. The humidity values inside the vehicle were maintained between 1016 and 1030 Mbar throughout the experiment.

#### *2.3. Experimental Conditions*

In this study, an experiment was conducted during real vehicle driving between the cities of Incheon and Seoul in South Korea, which is a fifty-kilometer route consisting of five main segments: City 1 (Incheon in Korea), Highway 1, Highway 2, Highway 3, and City 2 (Seoul in Korea) with two Tollgate points (TG) (Figure 2). The total time for a one-way drive was approximately 60 min. During this route, two types of recorded data were collected, such as EDA data measured by the Empatica E4 sensor (sampling rate = 4 Hz) and speed signal measured by OBD-II.

All studied data were collected for a month from 13 December 2017 to 13 January 2018. Datasets for 25 weekdays were obtained. Data from four days were omitted because they were inaccurate and unclear, such as unrecorded time data, device failure and zero values of EDA data. The total number of datasets (days) used in the analysis was 21.

**Figure 2.** Experiment on the driving route between Incheon and Seoul (captured by Google Earth).
