ABSTRACT

This report begins with a discussion of the statistical nature of noise, giving particular attention to the difference between thermal noise and fixed-amplitude random-phase noise. Then two measures of noise power fluctuations are introduced, the fourth moment of the waveform and the fourth moment of the envelope. It is argued that the latter measure provides the more pertinent description for psychoacoustical purposes. Experiments are described suggesting that noise power fluctuations can contribute about 5 dB of masking to the familiar signal in noise detection experiment and that these fluctuations are the basis for the discrimination of spectral density for small noise bandwidths.