Detection of tones in low-noise noise : further evidence for the role of envelope fluctuations

This paper investigates the role of envelope fluctuations in simultaneous masking conditions. Thresholds for tones in noise with a flat temporal envelope (low-noise noise, LNN) were compared with those in Gaussian noise. All measurements were performed with a running-noise presentation of 500-ms maskers. The sinusoidal signal was spectrally and temporally centered in the masker. The main findings were: (a) The 5.5-dB threshold difference between 100-Hz-wide Gaussian and LNN maskers at 1 kHz that was previously observed using frozen noise (cf. Hartmann and Pumplin [J. Acoust. Soc. Am. 83. 2277-2289 (1988)]) is also apparant for running noise, although thresholds are generally higher in the latter condition. (b) The threshold difference between Gaussian and LNN maskers at 1 kHz reaches a maximum of 9.4 dB at a masker bandwidth of 25 Hz, while at 10 kHz, the difference reaches a maximum of 15 dB at bandwidths of 50 and 100 Hz. For a 100-Hz-wide masker presented at different center frequencies, there is no advantage for LNN maskers below 1 kHz. Towards higher frequencies, the difference between the two noises increases and reaches about 15 dB at 10 kHz. (c) At 1 kHz with a 100-Hz bandwidth, decreasing the signal duration from 500 to 20 ms increases the threshold difference to 7.6 dB. (d) Thresholds in a dichotic condition, in which the masker is in phase and the signal is out of phase, lie within 2 dB for the two noise types, and are nearly constant for masker bandwidths between 5 and 100 Hz. It is argued that the primary detection cue in LNN is not an increase in energy, but rather an increase in envelope fluctuations due to the addition of the signal. This hypotheses is supported by simulations with an auditory-filterbank model. The simulations further suggest that, for a large LNN advantage, it is not sufficient that the LNN envelope is flat at the output of the on-frequency filter. In addition, it is crucial that off-frequency filters also yield a flat temporal envelope.