Skip to main content


Table 2 Features characterizing driving fatigue in literature

From: Feature selection for driving fatigue characterization and detection using visual- and signal-based sensors

Id Feature descriptions
\(\hbox {F}_{1}\) PERCLOS (\(\hbox {F}_{1}\) ) (Bergasa and Nuevo 2006; Ji et al. 2004; Jin et al. 2013; Kong et al. 2015; Friedrichs and Yang 2010; Darshana et al. 2015; Sun et al. 2017): represents the percentage of eyelid closure at 75%: \( \text{PERCLOS}_{75}=\frac{N_{\text{closure}75}}{N} \)
\(F_{2}\) Eye closure duration (ECD) (\(F_{2}\)) (Azim et al. 2014; Bergasa and Nuevo 2006; Kong et al. 2015; Sun et al. 2017, 2015): represents the number of frames where the eye is totally closed. \(\text{ECD}=\sum N_{\text{Total closure}}\)
\(F_{3}\) Pupil diameter (PupilDiam) (\(F_{3}\)) (Miyaji and Kawanaka 2010): represents the diameter of eye iris contour. \(\text{PupilDiam}=\text{average}(P\_\text{diam}_{\text{l}},P\_\text{diam}_{\text{r}})\)
\(F_{4}\) Percentage of saccade (Persac) (\(F_{4}\)) (Ji et al. 2004; Liang et al. 2007): represents the percentage of eye pupil saccade
\(F_{5}\) Average eye closure speed (AECS) (\(F_{5}\)) (Bergasa and Nuevo 2006; Ji et al. 2004; Friedrichs and Yang 2010): represents the number of frames where the eye changed its state (opened/closed) \(\text{AECS}=\frac{N_{\text{Total closure}}+ N_{\text{Total opening}}}{N_{\text{Transition}}}\)
\(F_{6}\) Microsleep (\(F_{6}\)) (Friedrichs and Yang 2010): Microsleep during 30 s \(\text{Microsleep}=\sum N_{\text{Total closure}}/N\)
\(F_{7}\) Eye opened (EOP) (\(F_{7}\)) (Sun et al. 2017, 2015): eye opened during 30 s \(\text{EOP}=\sum N_{\text{Total opening}}/N\)
\(F_{8}\) Gaze angle distribution (\(\text{Gaz}_{\text{dis}}\)) (Ji et al. 2004; Engström et al. 2005; Jin et al. 2013; Darshana et al. 2015; Miyaji and Kawanaka 2010): \(\text{Gaze}= \sqrt{\varphi ^{2}+\psi ^{2}}\)
\(F_{9}\) Gaze angle variation (\(\text{Gaz}_{\text{var}}\)) (Ji et al. 2004): \(\text{Gaz}_{\text{var}}=\sigma (\text{Gaze}_{\text{dis}})\)
\(F_{10}\) FixedGaze (Bergasa and Nuevo 2006; Liang et al. 2007): it is a Boolean variable, if \(\text{Gaze}_{\text{dis}}\) is stable during 30 s \(\text{FixedGaze}= 1\), otherwise \(\text{FixedGaze}= 0\)
\(F_{11}\) Blink frequency (\(\text{Blink}_{\text{freq}}\)) (Bergasa and Nuevo 2006; Jin et al. 2013; Friedrichs and Yang 2010; Sun et al. 2017, 2015): \(\text{Blink}_{\text{freq}}=\frac{N_{\text{Total closure}}}{N}\)
\(F_{12}\) Blink duration (Blink-dur) (Friedrichs and Yang 2010; Hu and Zheng 2009) : represents the blink time amount
\(F_{13}\) Distribution of head tilt angle (Head-tilts-dis) (Bergasa and Nuevo 2006; Ji et al. 2004): \(\text{HeadTilts}= \sqrt{\varphi _{\text{head}}^{2}+\psi _{\text{head}}^{2}}\)
\(F_{14}\) Variation of head tilt angle (Head-tilts-var) (Bergasa and Nuevo 2006): \(\text{HeadTiltsVar}=\sigma (\text{HeadTilts}) \)
\(F_{15}\) Head nodding frequency (Head-nod-freq) (Bergasa and Nuevo 2006; Ji et al. 2004): represents the number of peaks in the head tilt angle distribution
\(F_{16}\) FacePose (Bergasa and Nuevo 2006): represents the number of frames where driver head is in frontal position \(\sum N_{\text{head frontal position}}\)
\(F_{17}\) Steering wheel movement (SWM) (Engström et al. 2005; Joly et al. 2018; Thiffault and Bergeron 2003; Sahayadhas et al. 2012; Ingre et al. 2006): represents the variability in steering wheel movements