On Reliable Time-Frequency Characterization and Delay Estimation of Stimulus Frequency Otoacoustic Emissions

The limited information on origin and nature of stimulus frequency otoacoustic emissions (SFOAEs) necessitates a thorough reexamination into SFOAE analysis procedures. This will lead to a better understanding of the generation of SFOAEs. The SFOAE response waveform in the time domain can be interpreted as a summation of amplitude modulated and frequency modulated component waveforms. The efficiency of a technique to segregate these components is critical to describe the nature of SFOAEs. Recent advancements in robust time-frequency analysis algorithms have staked claims on the more accurate extraction of these components, from composite signals buried in noise. However, their potential has not been fully explored for SFOAEs analysis. Indifference to distinct information, due to nature of these analysis techniques, may impact the scientific conclusions. This paper attempts to bridge this gap in literature by evaluating the performance of three linear time-frequency analysis algorithms: short-time Fourier transform (STFT), continuous Wavelet transform (CWT), S-transform (ST) and two nonlinear algorithms: Hilbert-Huang Transform (HHT), synchrosqueezed Wavelet transform (SWT). We revisit the extraction of constituent components and estimation of their magnitude and delay, by carefully evaluating the impact of variation in analysis parameters. The performance of HHT and SWT from the perspective of time-frequency filtering and delay estimation were found to be relatively less efficient for analyzing SFOAEs. The intrinsic mode functions of HHT does not completely characterize the reflection components and hence IMF based filtering alone, is not recommended for segregating principal emission from multiple reflection components. We found STFT, WT, and ST to be suitable for canceling multiple internal reflection components with marginal altering in SFOAE.