BioSystems 97 (2009) 15–27 Contents lists available at ScienceDirect BioSystems journal homepage: www.elsevier.com/locate/biosystems Hidden pattern discovery on event related potential EEG signals Kam Swee Ng, Hyung-Jeong Yang ∗ , Sun-Hee Kim Department of Computer Science, Chonnam National University, Gwangju 500-757, South Korea article info Article history: Received 5 January 2009 Received in revised form 18 March 2009 Accepted 24 March 2009 Keywords: ERP EEG Hidden variables Principal Component Analysis abstract EEG signals are important to capture brain disorders. They are useful for analyzing the cognitive activity of the brain and diagnosing types of seizure and potential mental health problems. The Event Related Potential can be measured through the EEG signal. However, it is always difficult to interpret due to its low amplitude and sensitivity to changes of the mental activity. In this paper, we propose a novel approach to incrementally detect the pattern of this kind of EEG signal. This approach successfully summarizes the whole stream of the EEG signal by finding the correlations across the electrodes and discriminates the signals corresponding to various tasks into different patterns. It is also able to detect the transition period between different EEG signals and identify the electrodes which contribute the most to these signals. The experimental results show that the proposed method allows the significant meaning of the EEG signal to be obtained from the extracted pattern. © 2009 Elsevier Ireland Ltd. All rights reserved. 1. Introduction Electroencephalography (EEG) is a process that measures and records the electrical activity of the brain (Rhodes, 2008). Elec- trodes, which act as the sensors to detect the electrical activity, are attached to the surface of the cerebral cortex and are con- nected by wires to a computer to capture the brain activities. EEG is non-invasive and provides information which is retrievable, easily recorded and is obtainable using inexpensive technology compared with other brain activity diagnosis methods. The reason that the EEG signal is retrieved is because it is useful in recognizing brain disorders, such as by identifying the type of seizure occurring in the brain, and in diagnosing epilepsy (Semmlow, 2004). We can examine the problem of dementia, which is difficult to diagnose by ordinary clinical diagnosis. Furthermore, it provides a practical method of identifying persons who suffer from mental health problems or are in a comatose condition, because it provides really useful information about the mental state of the brain. Aside from the clinical field, EEG is also applied in the neuro- physiological area (Cahn and Polich, 2006; Luu, 2004). EEG is also being studied in an attempt to provide a new way of communica- tion between the human brain and computers by reading the signal extracted from the brain via the computer interface (Peterson et al.,2005; Ebrahimi et al., 2003; WolPaw et al., 2000). The EEG signal can be obtained in various ways. It can be gen- erated by measuring the response to a thought or perception. The ∗ Corresponding author. Tel.: +82 62 530 3436; fax: +82 62 530 3439. E-mail addresses: kamswee@gmail.com (K.S. Ng), hjyang@chonnam.ac.kr (H.-J. Yang), wkdal749@hanmail.net (S.-H. Kim). recording of the EEG signal can be carried out when a person is reading a sentence in a book, viewing a picture in an art gallery or listening to music on the radio (Osterhoust and Holcomb, 2008). These kinds of electrical activities are called Event Related Poten- tial (ERP) EEG signals. Unfortunately, the ERP EEG is of relatively small amplitude. As a result, it cannot be readily seen in a raw EEG tracing. ERP EEG signals have not been proven to be specific or sen- sitive enough to distinguish between changes in mental activity. For example, the ERP EEG signal might look similar while reading a sentence, viewing a picture or listening to music. Therefore, the study of pattern discovery in the cognitive activity of the ERP EEG signal is necessary. Various analysis techniques have been proposed in the literature to process EEG signals. The EEG signal is transformed into the form of changes in the frequency domain over time using time–frequency transforms. The use of the Fourier transform was proposed in the early days to isolate the individual components of an EEG signal (Xiao and Hu, 2008). It is based on the observation that the EEG characteristic falls within four frequency bands, viz. delta (1–4 Hz), theta (4–8 Hz), alpha (8–13 Hz) and beta (13–30 Hz). However, it is well suited only to the study of stationary signals where all frequen- cies have an infinite coherence time. In order to obtain favorable results, a suitable window is required for the Fourier transform. Another time frequency transform is the wavelet transform which extracts the EEG signal from the wavelet coefficient (Jahankhani et al., 2006, 2007; Inuso et al., 2007). It uses a multi-resolution technique by which different frequencies are analyzed with dif- ferent resolutions. Nevertheless, the resulting wavelet transform suffers from shift sensitivity. The wavelet transform of a signal and of the time-shifted version of the same signal are not simply shifted versions of each other. 0303-2647/$ – see front matter © 2009 Elsevier Ireland Ltd. All rights reserved. doi:10.1016/j.biosystems.2009.03.007