Abstract:
The cues for detecting an S(pi) signal masked by an No noise are driven by the signal-to-masker ratio. If the detection process can resolve epochs of low-masker energy such as dips in a fluctuating masker, it would be expected that greater weight would be given to portions of the signal which occur during these epochs. This prediction was tested using the time-domain version of COSS analysis [S. Buus et al., J. Acoust. Soc. Am. 99, 2288--2297 (1996)]. The masker was a 50-Hz-wide noise centered at 500 Hz; the signal was a train of six 500-Hz tone bursts each 50 ms in duration. The level of each tone burst was independently perturbed during each trial around levels previously determined for both NoSo and NoS(pi) thresholds. For each signal phase condition, 5000 trials were collected from each of four listeners, yielding a total of 30 000 signal+masker combinations per listener. The listeners' responses were analyzed with respect to both signal level and corresponding short-term masker level. Results showed that listener weights were relatively constant across all short-term masker levels for the NoSo condition but, for the NoS(pi) condition, were higher during masker dips and lower during masker peaks. [Work supported by NIDCD.]