A figure of merit for vocal attack time measurement.

Objectives: To develop and evaluate a figure of merit (FOM) for the measurement of vocal attack time (VAT) for sustained /a/. The FOM assesses a critical assumption of vocal startup on which the VAT measure is based and therefore represents integrity of the derived measure.

Methods: Pearson's correlation coefficient (0 ≤ |r| ≤ 1) was determined from amplitude features of sound pressure (SP) and electroglottographic (EGG) signals (n=1,033 tokens) to serve as a candidate FOM for the study. SP and EGG signals from 102 tokens were visually inspected to empirically derive a criterion level of FOM less than 0.75 to indicate when an assumption underlying a measurement had failed and the VAT measure should be disregarded.

Results: The median FOM value for all 1,033 tokens was 0.975. Thirty-two tokens (2.8%) were rejected on the basis of the FOM criterion. There was no correlation between VAT and FOM.

Conclusions: A companion metric, Pearson's r, can be readily obtained to assess the quality of VAT measurement. Decisions to retain or disregard VAT measures may therefore be made on the basis of measurement quality rather than on the basis of group statistics.