The Effect of Listening to Music on Driving Fatigue
The Effect of Listening to Music on Driving Fatigue
Joseph Dodson Thomas Jefferson High School for Science and Technology
This article was originally published in the 2021 print edition of Teknos Science Journal.
After a long day at work, you get behind the wheel, exhausted and ready to get home. The monotony of the highway soothes you, and you begin to doze off. Blinks become longer, breaths become slower, vision becomes blurrier. Drifting between lanes in drowsiness, you hit another car. If only you had a way to keep yourself awake just long enough to make it home safely. In the Neuroscience Research Lab, I am working to find just that. Using electroencephalogram (EEG) data (measurements of brain activity) and a machine learning model, I hope to determine whether music is an effective tool to reduce driver fatigue.
With traffic increasing in nearly every major American city, the impacts of drowsy driving have only worsened. Every year, over 100,000 crashes occur due to driver fatigue and drowsy driving, resulting in at least 1,550 deaths and over 71,000 injuries [6]. Not surprisingly, 20% of Americans admitted to falling asleep behind the wheel in the past year. This number rises to more than 40% of Americans when asked if they had fallen asleep while driving at any time during their lives [5]. Unfortunately, the current solutions to drowsy driving, which include not driving or getting more sleep, are unfeasible for many people, necessitating a new solution that can temporarily help drivers make it to their destination safely.
Past research has demonstrated that music is effective for temporarily reducing feelings of fatigue and increasing reported energy levels. In Bigliassi et al. (2018), participants were told to squeeze a ring for a set amount of time, inducing light-to-medium feelings of fatigue while either listening to music or being in silence. This task was performed in an fMRI scanner, allowing the researchers to see which parts of the brain were activated. The researchers found that participants who listened to music had their attention shifted away from the fatigue-inducing exercise and also had a higher energy level than the participants listening to silence [3].
In a similar study, Bigliassi et al. (2019) used an EEG cap to measure participants’ brain activity while they walked, either listening to music, a podcast, or nothing. Participants who listened to music had the lowest perceived exertion, the highest affective valence (or energy level), and the highest reported enjoyment during the task. Though participants liked the music more than the podcast, Costas Karageorghis (personal communication, Feb. 9, 2021) stated that the authors did not adjust for this difference because it was not statistically significant. Thus, the difference in the participants’ liking of the podcast and music could not have had a statistically significant impact on their perceived exertion, energy level, or enjoyment. In addition to increasing energy levels, the authors found that listening to music downregulated theta waves, which are associated with feelings of tiredness and fatigue, while upregulating high-frequency beta waves, which inhibit theta waves and thus reduce feelings of fatigue [4].
Previous studies have found that EEG data can be used in machine learning algorithms to identify driver fatigue. Luo et al. (2019) used data from a forehead EEG that captured data as participants drove in a virtual simulator. Using this data, the authors created an algorithm to identify three main entropy features in the EEG data (parts of the data that stand out due to their complexity). Using these entropies with a support-vector machine (SVM) classifier (a type of machine learning model), the researchers were able to identify driver fatigue with a 95% accuracy [8].
Jiao et al. (2020) also used EEG taken while participants were driving in a virtual simulator, a long-short term memory (LSTM) network, a type of machine learning model, was used to identify fatigue. However, instead of using entropies, the researchers focused on alpha-wave phenomena called alpha-blocking, which occurs when someone is focused on a task, and alpha wave attenuation-disappearance, which signals drowsiness. The researchers trained the LSTM network to find these phenomena, using them to predict fatigue with a mean accuracy of 98% [7].
A third study by Ahmadi et al. (2020) again used EEG data recorded while participants drove in a virtual simulator, but used a different technique for classifying fatigue. They focused on a measurement called Gaussian copula mutual information (GCMI), which shows how much information is shared between two random wavelets, or small segments of a single EEG channel. Once this was calculated, the researchers used existing statistical methods to select features in the data. The authors then programmed their SVM classifier to focus on these features and the GCMI data when classifying drivers as fatigued or non-fatigued, achieving an accuracy of 98.1% [2].
The final study by Acı et al. (2019) did not focus specifically on classifying between fatigued and non-fatigued participants; rather, their goal was in classifying which of three mental states -- focused, unfocused, and asleep -- the driver was in. Using EEG data taken while participants drove a train in a virtual simulator, the researchers first performed feature extraction, which consists of using different statistical techniques to locate the most important features in the EEG data. In this study, the authors used a short-time Fourier transform and the Blackman window, both standard feature extraction methods. Once the feature extraction was complete, the researchers trained two SVM classifiers: one to detect the focused state against all of the other states and another to detect the unfocused state against all of the others. If a certain sample was not classified as focused or unfocused by one of the SVM models, then it represented the asleep category. Using this combination of feature extraction and machine learning models, the researchers achieved an average accuracy of 91.72% [1].
In my project, I am applying these studies to examine music as a way to increase energy levels and EEG data as an accurate way to predict fatigue. My design will first train a neural network called EEGNET on fatigued and unfatigued driver EEG data, teaching the model to classify fatigue. Then, I will run an EEG dataset of subjects listening to music through the model to generate a set of predicted fatigue classifications. Finally, I will compare these predicted classifications with the actual fatigue classifications from the driver dataset to determine whether music reduces driver fatigue in a statistically significant way.
Over the past few years, there has been an explosion in new technologies to identify fatigue, specifically for truckers [9]. It is my hope that one day, these technologies could be combined with a system that would automatically play music when it detects driver fatigue, or, even better, activate an autonomous driving system. Eventually, we will achieve a world where fatigue-induced crashes are a thing of the past.
References
[1] Acı, Ç. İ., Kaya, M., & Mishchenko, Y. (2019). Distinguishing mental attention states of humans via an EEG-based passive BCI using machine learning methods. Expert Systems With Applications, 134, 153-166. https://doi.org/10.1016/j.eswa.2019.05.057
[2] Ahmadi, A., Bazregarzadeh, H., & Kazemi, K. (2020). Automated detection of driver fatigue from electroencephalography through wavelet-based connectivity. Biocybernetics and Biomedical Engineering. https://doi.org/10.1016/j.bbe.2020.08.009
[3] Bigliassi, M., Karageorghis, C. I., Bishop, D. T., Nowicky, A. V., & Wright, M. J. (2018). Cerebral effects of music during isometric exercise: An fMRI study. International Journal of Psychophysiology, 133, 131-139. https://doi.org/10.1016/j.ijpsycho.2018.07.475
[4] Bigliassi, M., Karageorghis, C. I., Hoy, G. K., & Layne, G. S. (2019). The way you make me feel: Psychological and cerebral responses to music during real-life physical activity. Psychology of Sport and Exercise, 41, 211-217. https://doi.org/10.1016/j.psychsport.2018.01.010
[5] Drivers are Falling Asleep Behind the Wheel. (n.d.). National Safety Council. Retrieved January 21, 2021, from https://www.nsc.org/road-safety/safety-topics/fatigued-driving
[6] Facts and Stats. (n.d.). Drowsy Driving. Retrieved January 21, 2021, from https://drowsydriving.org/about/facts-and-stats/
[7] Jiao, Y., Deng, Y., Luo, Y., & Lu, B.-L. (2020). Driver sleepiness detection from EEG and EOG signals using GAN and LSTM networks. Neurocomputing, 408, 100-111. https://doi.org/10.1016/j.neucom.2019.05.108
[8] Luo, H., Qiu, T., Liu, C., & Huang, P. (2019). Research on fatigue driving detection using forehead EEG based on adaptive multi-scale entropy. Biomedical Signal Processing and Control, 51, 50-58. https://doi.org/10.1016/j.bspc.2019.02.005
[9] Weed, J. (2020, February 6). Wearable Tech That Tells Drowsy Truckers It's Time to Pull Over. The New York Times. Retrieved January 21, 2021, from https://www.nytimes.com/2020/02/06/business/drowsy-driving-truckers.html