Detection of auditory brainstem responses by adaptive filtering

Abstract

In evoked potential studies, the Auditory Brainstem Response (ABR) isrecovered from the noise background by filtering methods. Auditory EvokedPotentials are measured by far field recording techniques with non-invasive scalpelectrodes; the potential generated in the brainstem provides the most accurateinfonnation about the integrity of the Auditory System.In this thesis, EEG recordings from human brain (real data) are processedby three different filtering algorithms which are developed for the detection of ABRburried in EEG data. These methods are the Adaptive Filtering CAF) algorithm andAveraging & Adaptive (AAF) algorithm. The third method is the Adaptive LineEnhancement (ALE) algorithm and has a rather different filtering structure.The algorithms are implemented on a IBM compatible PC and comparedfor their rate of convergence. Their performance is evaluated in terms of the MeanResidual Error (MRE) , the Integrated Mean Square Error (IMSE), Distortion Index(D!) and the correlation coefficient between the filter output and the template signal.The template signal is chosen as the 1024 averaged data. It is shown that ALEconverges faster than the AF and AAF algorithms. The correlation between thereference signal and the template is an important criteria for the convergence speed.AF adapts itself slowly when compared with AAF and ALE, on the other hand, itconverges faster than the averaging algorithm.Key Words: Electroencephalogram, Evoked Potential, Signal processing,Averaging, Adaptive Filtering.