International Journal of Computer Applications (0975 – 8887) Volume 60– No.4, December 2012 50 Orthogonal and Biorthogonal Wavelet Analysis of Visual Evoked Potentials Ahmed Fadhil Hassoney Faculty of Physical Science UTM, Johor, Malaysia. Abd Khamim Ismail Faculty of Physical Science UTM, Johor, Malaysia. Hentabli Hamza Faculty of Comp. Science and Infor. System, UTM, Johor, Malaysia. ABSTRACT In the present work the performance of orthogonal and Biorthogonal wavelet filters were analyzed for visual evoked potentials (VEP) on a variety of noisy signals. The signals were analyzed at different signal to noise ratio (SNR). This research proposed a method for the selection of the best analysis. The proposed method used longest common subsequence (LCS) and basic local alignment search tool (BLAST) to measure the analysis performance objectively and visual quality subjectively of the signal analysis. It was found that orthogonal wavelets outperform the biorthogonal ones in both the criteria especially at high noisy signal. General Terms Signal analysis and processing. Keywords Wavelet transforms, longest common subsequence, Basic Local Alignment Search Tool 1. INTRODUCTION The wide application of visual evoked potentials (VEPs) urged the researcher to repeat the same question, how can the picked up signals be improved in order to give a better view to the observer. Wavelet transform is the newest technique to replace the traditional time-frequency by time-scale signal processing. The major advantage of wavelet is the ability to perform local analysis that is, to analyze a localized area of a larger signal; however time varying non-stationary waveforms are decomposed using wavelet analysis. Neuroelectric waveforms are non-stationary signals and wavelet techniques analyze such signals by providing excellent joint time frequency resolution. Wavelet analysis had been successfully applied for analysis of EEG potentials and spike detection [2]. Some studies used wavelet transforms with their BCI system based on P300 response [3]. Steady state sweep visual evoked potentials wavelet decomposition and multi-resolution decomposition and denoising of VEP and AEP are correctly done by using biorthogonal wavelets [4][5]. Quiroga et. al. show a good example of wavelet transform in the analysis of the frequency composition of evoked potentials, it showed a better performance of wavelet decomposition as compared to the Fourier based method [6]. Classification of the signals with the help of the wavelet functions have been shown in many studies, likewise the study used second order Gaussian wavelet kernels with multifocal visual evoked potentials (mfVEP) [7]. Daubechies wavelet (db4) was used to distinguish the normal and abnormal VEP responses from each other [8]. VEP signal decomposition and denoising were studied using symlet 5 [9]. Db4 and coif3 were also tested to extract feature of P300 oaaball [10]. Coiflet wavelet was used for classification of EEG for brain computer interface [11]. This present research introduces a comparative study of one dimensional discrete wavelets function between orthogonal (Dubechies (db), Symlet (sym) and Coiflet (cif)) and biorthogonal wavelets of visual evoked potentials to find the best analysis that matches perfectly with the original signal. The basic measure of the performance of the analysis algorithm is the longest common subsequence (LCS) and basic local alignment search tool (BLAST), which are defined by matching the string of the two sequences; the analysis of various wavelet families for signal processing on a variety of signals with additional known noise and then process it to compare the performance of wavelets. According to this analysis, the selection of the best wavelet for VEP signal processing taking into account improvement in the signal to noise ratio (SNR) was shown. 2. WAVELET FAMILIES Wavelet families with filter can be divided into two main categories, orthogonal and Biorthogonal wavelets, which have different properties of basic functions. Orthogonality decorrelates the transform coefficients by minimizing redundancy. Symmetry provides linear phase and minimize border arti-facts. Other Important properties of wavelet functions in signal processing applications are compact support, symmetry, regularity and degree of smoothness [12]. 3. QUALITY MEASURES In this research the performances of signal processing techniques are mainly analyzed on the basis of two measures: Longest common subsequence (LCS) and basic local alignment search tool (BLAST). LCS is defined as finding the longest subsequence common to all sequences in a set of sequences. Note that, subsequence is different from a substring and the longest matching subsequence between two strings; the analyzed one as compared to the original. BLAST enables a researcher to compare a query sequence with a library or database of sequences and identify library sequences that resemble the query sequence above a certain threshold. Visual quality of the signal is also considered as subjective quality measures. 4. EXPERIMENTAL-RESULTS, ANALYSIS AND COMPARISON Orthogonal and Biorthogonal wavelet families were analyzed for visual evoked potentials signal and their results were compared. First, transient visual evoked potentials signal (VEP) were simulated based on the standard ISCEV [13], then white Gaussian noise was added to the original signal at five different signal to noise ratio (0, 5, 10, -5, and -10 db). Biorthogonal and orthogonal (Daubchies, symlet and coiflet) wavelet analysis were applied to the noisy signal in order to improve signal to noise ratio and get back the original signal. Figure 1 shows the original signal and signal with additional