TY  - JOUR
T1  - Effective Detection and Classification of Drowsiness using Clustering and
Support Vector Machines
AU - Alajmi, Masoud AU - Asharindavida, Fayas AU - Khammari, Hedi AU - Ahmed, Irfan AU - Masud, Mehedi 
JO  - Journal of Engineering and Applied Sciences
VL  - 14
IS  - 6
SP  - 1867
EP  - 1874
PY  - 2019
DA  - 2001/08/19
SN  - 1816-949x
DO  - jeasci.2019.1867.1874
UR  - https://makhillpublications.co/view-article.php?doi=jeasci.2019.1867.1874
KW  - Drowsiness
KW  -EEG signals
KW  -k-means clustering
KW  -SVM classification
KW  -decompose
KW  -Daubechies
AB  - The study presents drowsiness classification by analyzing the Electroencephalograph (EEG) signals.
The EEG signals of different drowsiness situations such as active, drowsy and sleep are captured from many
individuals using a multichannel electrode system. We consider the dominant electrode pair FPZ-CZ and PZ-OZ
in this study. Denoising of the signal is done for clarity and band pass filtering is done with the required cutoff
frequency. In order to get the five sub bands (delta, theta, alpha, beta and gamma) of EEG signal, third order
Daubechies wavelet transform is used to decompose. Statistical analysis and energy features are calculated of
these sub bands for the effective classification. Classification is done in two stages which starts from clustering
of the feature data using K-means and later effective classification using the Support Vector Machine (SVM)
which uses the Gaussian kernel. Work has been carried out in the famous Physiobank Sleep Database. Result
is very promising and it overcomes many limitations by using the unsupervised learning and the effective
classification.
ER  -