ABSTRACT
Specific features of breath sounds, not otherwise visible in the graphic records, can be identified by correlation methods. Some practical considerations in the calculation of the cross-correlations should be noted. The sampling rate must correspond to the required time resolution. The interpretation of cross-correlation graphs can provide information on the frequency content of the signals. Statistical methods are necessary to characterize signals that are nondeterministic — that is, lacking the repetitive features and structure needed for describing the signal in an exact manner with a mathematical formula. Correlation analyses are used to determine relationships within the signal, and between two signals that are related physically and recorded simultaneously. The determination of the amplitude of breath sounds is not as simple as one might think. The evaluation of the statistical properties of breath sounds is important, since most signal-processing algorithms make hidden or explicit assumptions regarding these properties.
