Distribution based definition of impulse noise

Existing definitions of impulse noise are based on estimates of either the decay of the noise envelope, the repetition rate of the peaks, or a comparison of single peak level to rms level. Because each of these measures fails to fully characterize at least one of the relevant characteristics of an impulse, many exceptions to these classifications can be found. An improved alternative classification should be independent of absolute pressure magnitude, impulse duration, and repetition rate; yet, it should incorporate all these on a relative basis. The kurtosis (b2) of the amplitude distribution of a sampled impulse is such a descriptor. Examination of distribution kurtosis over samples of 1, 5, 11, and 42 s suggests that the “impulsiveness” of an exposure may be characterized by this statistic. The 11‐s window also relates the statistic to click‐induced TTS processes in the ear. Normally distributed data exhibit a kurtosis of zero. For comparison, average values of b2 calculated for the 11‐s samples are: b2 = − 1.5 for a sinusoid, b2 = 65.4 for barrel recycling, b2 = 3.0 for coin blanking, b2 = 44.0 for forging, and b2 = 24.6 for stamping. Analysis of kutosis and crest factor shows that they are related measures of the distribution. In limited cases, therefore, crest factor may have utility as an interim indicator of impulsiveness.