Comparison of signal preprocessing techniques for avoiding spectral leakage in auditory steady-state responses
Felipe Antunes, Leonardo Bonato Félix · Research on Biomedical Engineering · 2019
The auditory steady-state response (ASSR) can be detected with the magnitude-squared coherence (MSC)—which is an objective response detector in the frequency domain. The performance of detection techniques is affected by the spectral leakage that arises from the Fourier analysis. This study aimed at investigating two preprocessing techniques designed to mitigate spectral leakage: windowing and bandpass filtering. These two procedures were applied prior to the application of the MSC in the detection of ASSRs in the electroencephalogram of healthy volunteers. The ASSRs were evoked by amplitude modulated tones. Preprocessing techniques usually improve the performance of MSC, but windowing procedures were worse when compared to filtering. The filtering preprocessing improved the detection rate up to 145.7%. The false positive rates remained close to the significance level of the tests. In order to mitigate the spectral leakage effects on the performance of MSC in detecting ASSR, bandpass filtering is preferred to windowing. The best results were obtained by 8th order IIR filters (Butterworth, Type 1 Chebyshev, and Elliptic).