We present a set of RFI analysis methods based on the PCA and FFT. We combine these methods to develop an RFI classifier that could categorize RFI into three different general types: impulsive, periodic, non-stationary. Our method relies on principal decomposition to separate the signals of different RFI sources, allows us to detect weaker RFI than transitional threshold-based method (such as rfifind). Then we perform Fourier analysis to detect periodic RFI that is detrimental to pulsar searches. Finally, we classify the RFI based on the features extracted by PCA and Fourier analysis, then use these extracted features to build a more complete picture about the telescope RFI environment. By correlating the RFI analysis results with the telescope meta information, we could perform forensics to track down the RFI sources.
