Table 6.

Literature result (Part 1 of 5): summary table of papers reporting on objective, analysis methods, and obtained results.

Author(s)	Objective	Analysis Methods	Results
Ahn et al., 2017 [65]	Detect when a person is about to enter a vehicle by analysing the movement trajectory of the smartphone	Fuzzy Inference System, electromagnetic field (EMF) fluctuations	91.1% to 94.0% accuracy; maintains at least 87.8% accuracy regardless of smartphone position and vehicle type
Ahn et al., 2019 [66]	Classify users into drivers and passengers and whether they have entered a vehicle	Bayesian classifier	Identifies the driver’s smartphone with 89.1% average accuracy
Albert et al., 2016 [67]	Identify smartphone apps that have the greatest potential to reduce risky driving behaviour	Apps mapping, Analytic Hierarchy Process (APH)	Texting prevention and Green Box are unlikely to be accepted and used; collision warning and voice control are expected to gain public support
Alqudah et al., 2021 [68]	Classify driving events such as high speed, low speed, stop, and U-turn using smartphone sensors	SVM, decision trees, Discriminate Analysis, Naïve Bayes, KNN, ensembles	Classify events with over 98% accuracy using decision trees
Baheti et al., 2018 [69]	Detect distracted drivers and the type of distraction, such as texting, talking on a mobile phone, eating, or drinking	CNN (VGG-16 architecture)	94.44% accuracy on test set; adding dropout, L2 weight regularisation, and batch normalisation increases accuracy to 96.31% on test set
Bergasa et al., 2014 [54]	Detect inattentive driving and provide feedback to the driver, assessing their driving and warning them if their behaviour is unsafe	Drowsiness score uses lane drifting and lane weaving signals to infer drowsiness; distraction score based on sudden longitudinal and lateral movements	Data from 12 drivers in two different studies; detects some inattentive driving behaviours and achieves an overall accuracy of 82% with a recall of 92%
Berri et al., 2014 [71]	Present an algorithm for extracting features from images to detect the use of mobile phones by drivers	Computer vision and machine learning (SVM for classification)	Average accuracy of 91.57% for the set of images analysed
Bo et al., 2013b [72]	Detect drivers and passengers, and whether a smartphone is being used for texting	Classification with hidden Markov model (HMM)	Classification accuracy of 87% and precision of 96.67% across 20 different driving and parking cases
Bortnik and Lavrenovs, 2021 [73]	Identify the driver’s interaction with the smartphone, such as app activity, call activity, or screen activity	Android dumpsys diagnostic data	N/A
Caird et al., 2014 [74]	Presents a meta-study on texting and driving	N/A	N/A
Castignani et al., 2015 [75]	Detect events related to driving style and scores drivers	Fuzzy logic, principal component analysis (PCA)	The developed system shows more than 90% accuracy in detecting events in an experiment with 10 drivers along a predefined route
Chen et al., 2015 [76]	Detect and differentiate between different vehicle steering patterns, such as lane changes, turns, and driving on winding roads	Signal processing, Kalman filter	High detection accuracy: 100% for right and left turns, 93% for lane changes, 97% for curvy roads
Chu et al., 2014 [77]	Detect whether a smartphone user in a vehicle is the driver or a passenger	Machine learning approach	Early prototypes on Android and iOS show over 85% accuracy with 6 users in 2 different cars
Chuang et al., 2014 [78]	Estimate driver gaze direction to detect driver distraction	Multi-class linear support vector machine (SVM) classifier	Classification accuracy between 86.4% and 97.4%.
Dai et al., 2019 [79]	Identification of the driver’s direction of speech (namely, front, right, and rear)	K-means clustering algorithm	95% accuracy on average for different phone placements, at least 92.2% accuracy for three scenarios, 90.3% accuracy when the window is open in the presence of outside noise