Evaluating speech-like distortion product otoacoustic emissions: a comparison of time- and frequency-domain metrics across three datasets

Magnitude-squared coherence and band power best separated in-ear speech-like DPOAE recordings from out-of-ear controls (separability ~0.9), while peak magnitude performed worst across 14 stimulus types.

• Why it matters: Different stimulus designs and analysis metrics have made prior speech-like DPOAE findings hard to compare, limiting how reliably attentional effects can be studied in the cochlea.
• What they did: The authors reanalyzed three existing datasets (106 participants, 14 stimulus types) using one pipeline and compared six metrics, including band power, MSC, and four complex cross-correlation measures (peak magnitude, SNR, peak area, peak width).
• The result: Band power, MSC, and SNR separated best (mean separability ~0.9), MSC was robust to emission latency while cross-correlation provided latency, and emissions below ~700 Hz were least reliably measured.

AI summary of the abstract · may contain errors · not medical advice · read the paper.