Abstract—Reliable classification of drowsy stage in EEG
signals have attracted attentions from researchers for many
years because of large amounts of brain signal noise. Recent
studies have demonstrated that the analysis of EEG signals can
get benefits from wavelet transform (WT). Despite of this,
experiments do not support the effective use of wavelet features
for the discrimination of EEG signals because there is much
redundant and irrelevant information contained in wavelet
coefficients. Furthermore, extraction of useful features from
EEG signals for classification is still an open research question.
The novel method present in this paper is to extract useful
features for classification of EEG signals based on wavelet
transform. This method basically consists of two major steps.
The first step is extracting energy coefficients from wavelet
transform based on Parseval’s theorem to represent the
distribution of brain signals. The second step focuses on revising
weights of energy coefficients to facilitate a classification
method. We show that the energy-based features not only
capture meaningful information of wavelet transform, but also
are useful for classification. We evaluate the proposed method
by using the energy-based features to train a neural network for
classification of drowsy and alert signals in EEG records. The
experimental results conducted on the MIT-BIH
Polysomnographic database have shown that the proposed
method achieves 90.27% of accuracy compared to
wavelet-based methods.
Index Terms—EEG, drowsiness, alertness, wavelet transform,
energy distribution, neural network, classification.
The authors are with the Department of Computer and Information
Science, King Mongkut's University of Technology North Bangkok,
Bangkok, Thailand 10800 (e-mail: getaaun9@gmail.com,
suwatchaik@kmutnb.ac.th, luepolp@kmutnb.ac.th).
Cite: Naiyana Boonnak, Suwatchai Kamonsantiroj, and Luepol Pipanmaekaporn, "Wavelet Transform Enhancement for Drowsiness Classification in EEG Records Using Energy Coefficient Distribution and Neural Network," International Journal of Machine Learning and Computing vol. 5, no. 4, pp. 288-293, 2015.