The invention discloses a method for automatically identifying and learning abnormal
radio signal types. The method is characterized by completing
feature extraction of a
radio signal in virtue of analysis and actual measurement of the
frequency spectrum data of the radio data in combination with experiential knowledge of radio monitoring experts and a
feature extraction method. In the
characteristic space of the
radio signal, clustering analysis is executed on actually-measured sweep-frequency interference,
broadband interference,
narrowband interference, and invalid intervention signals by using a clustering
analysis method. By means of a result of the clustering analysis, the method may provide a radio monitoring device with a capability of automatically identifying abnormal radio
signal types. With accumulation of the
frequency spectrum data of the actually-measured radio data and increase in identification errors, the method provides a capability of automatically learning the results of the clustering analysis regularly. According to an identification result in combination with a
communication device, the method provides an automatic alarming capability for a monitoring device. The method, used in a radio monitoring device, improves an
information processing capability of the radio monitoring device, may achieve unmanned operation of the radio monitoring device, and decreases the
workload of monitoring technicians.