Extracting and supplementing the new feature parameters is an effective way to solve the sorting problem of radar emitter signal in present complex systems. Because the ambiguity function can display the intrinsic information of signal waveform well, the main hill center and the main peak distribution radius of the three-dimensional graphs of ambiguity function are extracted from the main and side views as the feature vector of the signal sorting of radar emitter. The kernelized fuzzy C-means clustering experiment shows that the success rate of sorting six typical signals is 100{%