Detection and discrimination of amplitude-modulated signals in noise

Abstract

Comodulation masking release demonstrated improved detection of pure-tone signals in amplitude-comodulated noise bands [Hall et al., J. Acoust. Soc. Am. 76(1), 50–56 (1984)]. However, in most real-world situations, signals such as speech sounds are comodulated and noises such as cocktail party and traffic noises are not. Two experiments were designed to examine whether comodulation of signals can lead to improved signal detection in noise. Experiment I used a 2AFC method to measure detection threshold of a 2/3-oct-spaced, five-tone complex in pink noise. The complex tones were either amplitude comodulated or randomly modulated. Results from three normal-hearing listeners showed no difference in detection threshold between amplitude-comodulated and randomly modulated conditions. Experiment II used a method of constant stimuli to measure the psychometric function of discriminating six vowel-like stimuli in pink noise. These vowels consisted of three formant-frequency sinusoids which were unmodulated or amplitude-comodulated or randomly modulated by a 125-Hz sinusoid simulating the voice pitch. While no difference was observed between comodulated and randomly modulated conditions, the psychometric function for the nomodulation control condition was shifted by 4–6 dB toward lower signal-to-noise ratios. A possible explanation is that auditory fibers phase-lock to formant-frequencies of the unmodulated stimuli and to the fundamental frequency of the modulated stimuli.