Colored-noise performance of the order-truncate-average-ratio (OTAR) filter
Jeffrey H. Shapiro, Thomas J. Green · The Journal of the Acoustical Society of America · 1999
Spectrogram background normalization for narrow-band passive acoustic detection systems has a long history. This presentation addresses a more complicated scenario in which broadband background whitening is still desired, with simultaneous preservation of intermediate-bandwidth spectral features and suppression of narrow-band interference. Toward this end, a nonlinear processor—the order-truncate-average-ratio (OTAR) spectral filter—is presented and analyzed. While much is known about order statistics, closed-form expressions for the mean and covariance of the output of this filter appear to be unattainable. However, for white Gaussian noise-input time-series with no spectral averaging (conditionally exponential spectral statistics), approximate expressions for the output mean and covariance have been derived. Moreover, for scenarios employing spectral averaging on colored Gaussian noise-input time-series, expressions for the output mean have been derived. All of these analytical results agree well with simulation of the OTAR filter. The analysis has been applied to evaluate an appropriate transfer function for the filter and to illustrate a number of other interesting properties. [Work supported by SPAWAR, under Air Force Contract F19628-95-C-0002.]