The state characterization of cavitation is significant not only for preventing catastrophic faults in industrial applications but also for keeping equipment stealthiness for military purposes. Many works concerning mechanical diagnostics search for modulation frequencies. However, this diagnosis strategy has so far been hindered in cavitation detection of fluid machinery. This results from that the first-order and second-order cyclostationary noise in the monitoring signal likely have the same modulation frequencies as the cyclostationary components caused by cavitation. To deal with this dilemma, the present paper proposes a novel strategy—cavitation characterization by carrier distribution. First, a cyclostationary model of a single cavitating blade is established. On this basis, the mathematical connection between spectral correlation, carrier power spectral density, and modulation quantities is elaborated. Finally, attached cavity and unattached cavity are identified qualitatively from carrier distribution by combining cavitation mechanism. The first paper is also a prelude to the second paper where cavitation quantitative characterization is achieved by modulation intensity indicators.