This letter puts forward a method for intelligent condition diagnosis of rotating machinery using the probability density analysis and the canonical discriminant analysis (CDA) comprising the following steps. First, the noise is cancelled by statistics filter (SF), and the probability density functions (PDFs) of the vibration signals measured in each state are determined. Second, the segment values of the PDFs of the vibration signals are calculated and the integrated symptom parameters (ISPs) are combined using CDA. Third, Mahalanobis distances between the ISPs are introduced to identify the machine state. Moreover, the selecting discrimination index is optimized according to the accuracy rate of the identification. The efficacy of this novel method was confirmed by the results of the condition diagnosis for a centrifugal blower.