Estimation of background ambient noise levels from the spectral analysis of time series with application to cw propagation-loss measurements

Abstract

An application of percentile measurements is presented as a method for estimating background ambient noise levels from spectral density functions. The advantage of a percentile method over an average method is that it allows an accurate measure of background noise containing a sine wave, that is essentially independent of the sine-wave amplitude. This feature makes the percentile method very useful in the measurement of cw propagation loss, where it is frequently necessary to subtract an estimate of the background noise from a signal-plus-noise measurement, including the case when the noise contains sine waves close in frequency to the cw signal. A percentile-level estimate may be easily used to calculate an average level estimate for chi-square data and it is found that a percentile estimator is just as accurate and essentially just as stable as an average estimator. Theory based on chi-square statistics and experimental data from the ambient sound field in the deep ocean are shown to be in good agreement.