A new method for the automated detection of spontaneous otoacoustic emissions embedded in noisy data

There is, as yet, no universally agreed upon method for the detection of spontaneous otoacoustic emissions (SOAEs). In this paper, we augment extant techniques by presenting a new automated approach. The new procedure reliably labels the spectral peaks as SOAEs, rejects noisy data such as that due to body movements or breathing, provides a subject-specific measure of the probability that a particular peak is erroneously identified as a SOAE, and can be accomplished on-line. Comparison of results with this new approach and with methods undertaken previously in this laboratory and elsewhere are made. The estimated prevalence of SOAEs for the new method is as large as that for any method previously proposed.