Effects of spectral composition on the detectability of pulse‐train signals

Abstract

The detectability of a pulse‐train signal as a function of its spectral composition was investigated. A two‐interval forced‐choice detection task was employed with the level of the masking noise controlled by a PEST program. In all conditions, the signal was a pulse train composed of 60 100‐μsec positive or negative polarity pulses with an interpulse period of 2 msec. Polarity manipulations and filtering were used to vary both the overall bandwidth of the signal and the distribution of energy within that bandwidth. Conditions were run in which pulse polarity was a random variable or was varied according to some fixed (periodic) pattern. A variety of low‐pass and bandpass filter settings were used. Both monaural (NmSm) and binaural (N0Sm) conditions were run. Detectabilities were found to be a function of both the overall bandwidth and the spectral distribution within that bandwidth. Results indicated that although pulse‐train signals are broad‐band signals, it is the low‐frequency regions that are most important for detection. The relative importance of particular frequency regions was found to vary with the spectral distribution and with the mode (monaural or binaural) of presentation. Relationships to previous critical band and multicomponent signal detection research are discussed.