Application Of The Wavelet Transform And Higher-Order Cumulant-Based Inverse Filtering To The Quantification Of Non- Gaussian Noise For Use In Estimating Hearing Hazards

The foundation of the ISO-1999 standard for estimating NIHL in noise-exposed groups is tied to two postulates that form the basis of: (a) an age correction procedure and (b) the use of an energy metric (L eq) to quantify the exposure. This leads to a straightforward relation; NIHL = f(age, L eq). Implicit in such a formulation is that temporal variables associated with a noise exposure are not significant for the production of NIHL. This is contrary to the accumulating experimental data that suggest that energy may be a necessary but not sufficient metric for the evaluation of noise exposures. For example, Lei et al. 1 have shown that the statistical properties of a nonGaussian noise, reflected in the kurtosis statistic, which incorporates both temporal and peak properties of a signal in the time domain [β(t)] and spectral effects when computed on the frequency domain signal [β(f)], are highly correlated with the magnitude and distribution of sensory cell loss. In this paper we describe an approach to and present a rationale for the analysis of noise environments that incorporates the wavelet transform and higher-order cumulant based inverse filtering to obtain β(t), β(f), and the joint peak-interval histogram.