Comparison of methods for analysis of cyclostationary noise
Karel Vokurka · The Journal of the Acoustical Society of America · 1998
A great number of machines work cyclically. Hence the noise they generate is cyclostationary. A traditional approach to this noise analysis is based on an assumption of its stationarity. This assumption implies continuous time averaging. Unfortunately in this way a certain amount of information concerning, e.g., time-frequency distribution, is lost. Methods have been looked for to preserve this information. They all are based on periodic time averaging and the resulting statistical characteristics are known as gated spectrum, cyclic correlation, cyclic autospectrum, double autocorrelation, double autospectrum, etc. Each of the methods has its advantages and disadvantages. These may concern processing time, hardware requirements, frequency resolution, occurrence of artifacts (cross terms), and others. In this paper the methods are analyzed and compared and possible improvements in their performance are suggested. [Work supported by the Grant Agency of the Czech Republic and by the European Commission.]