Comparison of different feature sets for respiratory sound classifiers

In this study, a comparison is made between the performances of k-NN classifiers with different feature sets derived from respiratory sound data acquired from four different fixed locations on the posterior chest area. The two class recognition problem between healthy and pathological subjects is addressed. Each subject is represented by a single respiration cycle divided into sixty segments from which three different feature sets consisting of 6th order AR model coefficients, percentile frequency parameters and principle components, respectively, are extracted. Performances of k-NN classifiers for these feature sets for four different microphone locations are considered in segment-wise and subject-wise results.