Intensity Discrimination of Fluctuating Signals

Abstract

The theoretical aspects of the fluctuations inherent in random noise are discussed, insofar as these influence detection of random signals. Brief elucidations of the fundamentals of statistical detection theory and of sampling theory are given. The model developed for the detection process closely follows a proposal by Green, but with the major exception that a kind of internal noise is supposed to be involved in the process. The auditory mechanism is assumed to measure the average intensity of the signal presented over a certain time T. In doing so, the evaluation is hampered by random activity so that the measurement is carried out less accurately than possible. The detection finally involves a likelihood-ratio decision procedure. The theory explains experimental thresholds of random-noise signals very welL Agreement with data on thresholds of very short noise bursts is less convincing. Finally, the theory explains very elegantly why experimental psychometric functions for wideband signals have the same slope as those for narrow-band signals. Instances in which higher slopes are reported for experimental psychometric functions for wide-band signals can most likely be traced down to having involved experimental procedures that leave too much uncertainty in the listener''s observations.