*3.1. Heart Signal*

We have used several sensors in this work to measure stress objectively. One of the vital signs most used in research on driving is the heart rate variability (HRV). There are numerous studies where this biosignal is measured due to its strong correlation with stress, and the fact that it can be obtained in a non-intrusive way [5].

Heart rate variability can be analyzed in two di fferent domains: time and frequency. Time domain analysis of the HRV signal consists of measuring the mean or standard deviation of the time intervals between consecutive heartbeats. Frequency domain analysis is a method based on the amount of heart signal found in two di fferent frequency bands. In the case of heart rate analysis, the ranges are (0.04–0.15 Hz) and (0.15–0.4 Hz). In this work, we use the following measurements obtained from the heart rate signal that have been widely used in the literature [59]:

• pNN50 (%): this is the number of consecutive heartbeats di ffering more than 50 ms divided by the total number of measured heartbeats and expressed as a percentage. This variable decreases when driving stress is high.

• LF/HF: this is the low-frequency (LF) power (0.04–0.15 Hz) modulated by the sympathetic and parasympathetic nervous system divided by the high-frequency (HF) power (0.15–0.4 Hz) associated with the parasympathetic nerve activity. This ratio captures the global sympathovagal balance [25]. A high LF/HF ratio means sympathetic dominance, which happens when driving stress is elevated.

There are several non-intrusive devices which obtain the heart rate. Many of them are based on the photo-plethysmography sensor that allows us to measure the blood volume pulse. This type of device has improved significantly in recent years. However, they are very sensitive to movement and the level of pressure on the skin [60]. Another type of sensor that allows measuring the heart rate non-intrusively is a chest band with electrodes. These solutions achieve higher accuracy than the photo-plethysmography sensor [61,62].

In our case, the heart signal is obtained using a Polar H10 chest band. Polar H10 is the successor to the Polar H7 device. Polar H10 introduces improvements to measure heart rate variability. This device can o ffer precision for measuring the time between successive heartbeats (also called RR interval) similar to that obtained by a Holter ECG [61]. Polar H10 was connected wirelessly to the EliteHRV © app [63] running on a Google Pixel 3a. This app directly receives the RR intervals and uses a proprietary algorithm to correct artifacts such as ectopic beats or signal noise to present a more valid signal for heart rate analysis (HRV). We exported the calculated HRV values (pNN50 and LF/HF ratio) from the web dashboard provided by this app. It is also important to mention that before starting the heart rate measurement, there is a 30 s sensor stabilization period. Table 1 shows the Polar H10 specifications.


