An automated procedure for identifying spontaneous otoacoustic emissions.

Abstract

An algorithm is described for objectively identifying and measuring spontaneous otoacoustic emissions (SOAEs) using the spectrum that results from transformation of the acoustic waveform measured in the outer ear canal. Prior to spectral analysis, the rms level is calculated for successive short segments of the waveform and only the weakest 25% of the segments are retained for the spectral analysis [the quietest 150 when using 16k-point fast Fourier transforms (FFTs)]. The resulting initial spectrum is scanned for peaks (potential SOAEs) which are then deleted from the spectrum. New values are estimated for the deleted values using linear extrapolations from frequency ranges on either side of the deleted values. The end result is a smoothed spectrum devoid of all local peaks. The initial spectrum is then compared peak-by-peak with the smoothed spectrum, and those peaks having differences that exceed an objectively determined decision criterion are identified as likely SOAEs. The effects of varying some of the important parameter values of the algorithm are described, and the sensitivity of the procedure is evaluated by measuring the detection rate for a Lorentzian peak of known amplitude added to a spectrum otherwise devoid of SOAEs.