Using Kalman Filter to Process Auditory Brainstem Responses in Subjects with Normal Hearing

摘要

Auditory brainstem response (ABR) audiometry is a widely used tool for detecting hearing loss. The ABR is generated within the first 10-15 ms in the brainstem as responses to external sound stimulations following the stimulus onset. However, due to the small amplitude of the ABR signal, a large number of repeated averaging trials are usually needed to obtain meaningful ABR waveform, leading to long time consumption in the clinical practices. The subjects need to be sedated in special cases to keep them still for the long recording time. In this study, the method of Kalman filter was used to adaptively separate noises from ABR signals so that less trials were necessary to improve the efficiency. The performance of the Kalman filter method was compared with the traditional averaging methods under different conditions. The results showed that Kalman filter could obtain reliable ABR waveforms by wisely filtering out noises with a reduced number of repeated trials as less as 500 times, and the morphology converged to a smooth and clean curve as the number of trials increased. The study showed that the Kalman filter might provide a useful tool to improve the efficiency and signal quality of the ABR measurements.