Application of Independent Component Analysis on Artifacts Removal in the Time-Frequency Analysis of Transient Evoked Otoacoustic Emissions

Abstract

TEOAEs (Transient evoked otoacoustic emissions) analysis has become a clinical evaluation of peripheral auditory system as a basic and safe and objective means. Currently, one of the most effective OAEs analysis methods is time-frequency analysis method. In order to study the time-frequency distribution of TEOAEs, we established an improved TEOAEs simulation model which was based on the existing international generic TEOAEs simulation and the latency of TEOAEs in the latest research.. To the new model, we compared the results which had been got by Linear class, Cohen’s class, Affine class, Rearrangement class time-frequency distribution, and come to the conclusion that the cone-shaped kernel distribution (CKD) is the best description for TEOAEs. Meanwhile, in view of the problem that the TEOAEs artifacts affect the CKD, we proposed the application of ICA (independent component analysis) to remove the artifacts so as to improve the SNR (signal to noise ratio) and verified its reliability in disposing the TEOAEs contaminated by the artifacts. Ultimately we have established a two-step analysis method called FastICA-CKD. The new method was used to investigate into the real TEOAEs which were recorded in linear recording mode under different stimulus intensities. After that, it also has been discussed that was the difference among the TEOAEs which were recorded in linear recording mode before and after FastICA and the TEOAEs which were recorded in derived nonlinear recording (DNLR) mode under different stimulus intensities. Their correlation and the integrity of these useful components of TEOAEs in the CKD have been compared.