Improved transient evoked otoacoustic emission screening test using simple regression model and window optimization

The underlying problem in detection of transient evoked otoacoustic emission (TEOAE) is its extremely low level which is much under the level of noise that appears in the auditory canal. The algorithms based on wavelet transformation (WT) and frequency (e.g. scale) dependent windows are proved to have higher specificity and sensitivity of TEOAE test in comparison to the FFT-based algorithms using single analysis window. In this paper, a new algorithm for TEOAE screening test with improved performances is proposed. It is based on three improvements. The first one is based on simple linear regression model applied to series of signal-to-noise ratio (SNR) estimates and correction of the reproducibility estimate by mean square error criterion. The second one exploits inter-subjects variability of TEOAE latency by selecting the best window position. The third is based on the use of specific window shape developed from approximate of minimum mean square error criterion. The performance of the proposed TEOAE algorithm was tested on real TEOAE measurements embedded in artificial noise as well as on pure noise extracted from TEOAE measurements. The results provided evidence for the superiority of its performance in comparison with two referential algorithms.