An Effective Hybrid Model for EEG-Based Drowsiness Detection

An Effective Hybrid Model for EEG-Based Drowsiness Detection

Table of Contents





Abstract

Early detection of driver drowsiness and the development of a functioning driver alertness system may support the prevention of numerous vehicular accidents worldwide. Wearable sensors and camera-based systems are generally employed in the driver drowsiness detection. Electroencephalogram (or EEG) is considered another effective option for driver drowsiness detection. Various EEG-based drowsiness detection systems have been proposed to date. In this paper, EEG signals are also used for the detection of drowsiness, with the proposed method being composed of three main building blocks. Both raw EEG signals and their corresponding spectrograms are used in the proposed building blocks. In the first building block, while energy distribution and zero-crossing distribution features are calculated from the raw EEG signals, spectral entropy and instantaneous frequency features are extracted from the EEG spectrogram images. In the second building block, deep feature extraction is employed directly on the EEG spectrogram images using pre-trained AlexNet and VGGNet. In the third building block, the tunable Q-factor wavelet transform (TQWT) is used to decompose the EEG signals into related sub-bands. The spectrogram images of the obtained sub-bands and statistical features, such as mean and standard deviation of the sub-bands’ instantaneous frequencies, are then calculated. Each feature group from each building block is fed to a long-short term memory (LSTM) network for the purposes of classification. The obtained results from the LSTM networks are then fused with a majority voting layer. The MIT-BIH Polysomnographic database was used in the experimental works. The evaluation of the proposed method was carried out with a ten-fold cross-validation test and the average accuracy represented accordingly. The obtained average accuracy score was 94.31%. The obtained result was also compared with other results to be found in the literature. The comparison shows that the proposed method’s achievement was found to be better than the compared results.

Author Keywords

  • Drowsiness detection,
  • EEG signals,
  • signal processing,
  • deep feature extraction,
  • LSTM network

IEEE Keywords

  • Electroencephalography,
  • Feature extraction,
  • Vehicles,
  • Spectrogram,
  • Wavelet transforms,
  • Databases

Introduction

DROWSINESS is defined as feeling sleepy or being unable to keep one’s eyes open [1]. Due to the low level of consciousness during the state of drowsiness, a person’s instantaneous reflex is weakened, which negatively affects their ability to process quick decisions. Drowsiness is seen as a major cause of vehicular accidents worldwide. There are various statistics from around the world that indicate most vehicular accidents to be directly caused by driver fatigue [2]. Car companies invest millions of dollars incorporating drowsiness alert systems into their new models [3]. Driver-based drowsiness detection is generally carried out with cameras, wearable sensors, and EEG signals. Camera-based drowsiness detection systems employ various computer vision methods to detect the abnormal behaviors of drivers. Wearable sensor-based approaches utilize various signal processing and machine learning techniques in order to detect driver drowsiness. EEG-based driver drowsiness detection systems rely on the processing of EEG signals; an important brain state indicator used in sleep analysis [4]. To date, various methods have been proposed for EEG-based drowsiness detection. Garcés Correa et al. [5] proposed a driver drowsiness detection method based on power spectral density (PDS) and wavelets. Various frequency-based features were calculated from 18 EEG recordings. The artificial neural networks (ANN) classifier was used and an 84.1% accuracy score was obtained by the authors. Garces Correa and Laciar Leber [6] then investigated another driver drowsiness detection method where time and spectral analysis, and wavelets were used for feature extraction. The ANN classifier was again utilized, and an 85.66% accuracy score was reported by the authors. Belakhdar et al. [7] proposed windowed FFT, ANN, and support vector machines (SVM) classifiers for driver drowsiness detection. The authors used the polysomnography database to evaluate their proposals, and an accuracy score of 84.75% was obtained. Silveira et al. [8] proposed an effective method for drowsiness detection that used the m-terms from wavelet decomposition of EEG signals. Especially, the alpha and beta rhythms of the EEG signals were considered in discrimination of the subjects’ awake and drowsy states. The PhysioNet Sleep EEG dataset was used in their experiment and a 98.7% accuracy score was reported by the authors. Chen et al. [9] proposed an EEG-based drowsiness detection system based on wavelet transform, Fourier Transform (FT), and eyelid movements.

Conclusion

In this paper, an efficient hybrid method is proposed for EEG-based drowsiness detection. The proposed method employs three feature extraction mechanisms for the robust characterization of the drowsiness EEG signals. In doing so, energy and zero-crossing distributions, and spectral entropy and instantaneous frequency features are extracted in the first feature extraction mechanism. The second feature extraction mechanism extracts deep features from pre-trained AlexNet and VGG16 models. Finally, the statistical features of the instantaneous frequency of TQWT decomposed EEG signals are extracted in the third feature extraction mechanism. These features are then either classified individually with LSTM networks or the outputs of all LSTM networks are combined on a majority voting mechanism. The obtained results show that individual classification results ranged from 86.46% to 88.47%. In the individual models, the TQWT-based features achieved a higher accuracy score than the other two individual models. After the majority voting, the obtained accuracy score was 94.31%, which is an improvement on the individual achievements. To the best of our knowledge, this accuracy score is higher than the previously reported accuracy scores on the MIT-BIH Polysomnographic database. In the feature works, as TQWT-based features produce better accuracy scores, the multi-resolution signal decomposition models such as EMD and WT will be used in feature extraction.

About KSRA

The Kavian Scientific Research Association (KSRA) is a non-profit research organization to provide research / educational services in December 2013. The members of the community had formed a virtual group on the Viber social network. The core of the Kavian Scientific Association was formed with these members as founders. These individuals, led by Professor Siavosh Kaviani, decided to launch a scientific / research association with an emphasis on education.

KSRA research association, as a non-profit research firm, is committed to providing research services in the field of knowledge. The main beneficiaries of this association are public or private knowledge-based companies, students, researchers, researchers, professors, universities, and industrial and semi-industrial centers around the world.

Our main services Based on Education for all Spectrum people in the world. We want to make an integration between researches and educations. We believe education is the main right of Human beings. So our services should be concentrated on inclusive education.

The KSRA team partners with local under-served communities around the world to improve the access to and quality of knowledge based on education, amplify and augment learning programs where they exist, and create new opportunities for e-learning where traditional education systems are lacking or non-existent.

FULL Paper PDF file:

An Effective Hybrid Model for EEG-Based Drowsiness Detection

Bibliography

author

U. Budak, V. Bajaj, Y. Akbulut, O. Atila and A. Sengur,

Year

2020

Title

An Effective Hybrid Model for EEG-Based Drowsiness Detection

Publish in

in IEEE Sensors Journal, vol. 19, no. 17, pp. 7624-7631, 1 Sept.1, 2019

Doi

10.1109/JSEN.2019.2917850

PDF reference and original file: Click here

+ posts

Somayeh Nosrati was born in 1982 in Tehran. She holds a Master's degree in artificial intelligence from Khatam University of Tehran.