Objective measurement of cochlear tuning factor by means of time-frequency analysis of oto-acoustic emissions

A new technique is proposed for the objective estimate of cochlear tuning, starting from measurements of the delay function of Stimulus Frequency (SF) and Transiently Evoked (TE) Otoacoustic Emissions (OAEs). The technique is quick and reliable, also in not cooperating subjects, while the psychoacoustic tuning measurements are time consuming and based on a large number of assumptions. It is well known that OAEs originate from two main backscattering mechanisms: coherent reflection and nonlinear wave-fixed distortion. The main TEO-AE and SFOAE sources are supposed to be linear reflection from the peak region, due to randomly distributed roughness. Recent experiments found evidence of OAE sources more basally located with respect to the CP (characteristic place) on the basilar membrane (BM). The origin of the basal sources is due to the multiple-peak nature of the coherent local reflectivity function generated by the roughness. The OAE components generated at different places of the BM can be effectively separated in the time-frequency domain, being characterized by different phase-gradient delay. A time-frequency technique was proposed to identify the curved time-frequency region corresponding to single-reflection SFOAE and TEOAE components, to get, for each frequency, a weighted average of the delay over this region, weighted by the square of the wavelet coefficient. This average delay is assumed to scale as the square root of the tuning factor. The estimated spectral tuning values turned out to decrease significantly with increasing stimulus level, confirming that at high stimulus levels a saturation process occurs in which a widening of the BM excitation patterns takes place. This increase of the BM response width increases the relative importance of the shorter-delay more basal peaks of the reflectance and, consequently, a reduction of the average delay. The proposed technique is based on the idea that a smooth relation exists between the average delay and the BM tuning, which is correctly exploited to get reliable and stable tuning estimates only if: 1) multiple reflections are filtered out, and 2) a weighted average of the delay is considered instead of a single delay value associated with the most intense of the OAE components, which is that picked up by standard measurements of the phase-gradient delay.