Nonlinear separation of crackles and squawks from vesicular sounds using third-order statistics

A combination of nonlinear filtering with third-order statistics is presented, resulting to a modification of the nonlinear digital ST-NST filter, proposed by Arakawa et al, for improving its performance in noisy environments. The use of third-order statistics in estimating the AR-model's order and coefficients provides more reliable estimates of the stationary part of the input signal. The implementation of this modified ST-NST filter in fine crackles, coarse crackles and squawks was examined. Experiments have shown that a reliable and accurate separation of these adventitious sounds from vesicular sounds can be achieved, using this method, even in the presence of additive Gaussian noise.