Enhanced signal processing using modified cyclic shift tree denoising

Abstract

The cortical pyramidal neurons in the cerebral cortex, which are positioned perpendicularly to the brain’s surface, are assumed to be the primary source of the electroencephalogram (EEG) reading. The EEG reading generated by the brainstem in response to auditory impulses is known as the Auditory Brainstem Response (ABR). The identification of wave V in ABR is now regarded as the most efficient method for audiology testing. The ABR signal is modest in amplitude and is lost in the background noise. The traditional approach of retrieving the underlying wave V, which employs an averaging methodology, necessitates more attempts. This results in a protracted length of screening time, which causes the subject discomfort. For the detection of wave V, this paper uses Kalman filtering and Cyclic Shift Tree Denoising (CSTD). In state space form, we applied Markov process modeling of ABR dynamics. The Kalman filter, which is optimum in the mean-square sense, is used to estimate the clean ABRs. To save time and effort, discrete wavelet transform (DWT) coefficients are employed as features instead of filtering the raw ABR signal. The results show that even with a smaller number of epochs, the wave is still visible and the morphology of the ABR signal is preserved.