Modulation classification based on spectrogram

Three spectrogram-based modulation classification methods are proposed in this paper. Their recognition scope and performance is investigated or evaluated by theoretical analysis and simulation studies. The method taking moment-like features is robust to frequency offset while the other two, both of which make use of principal component analysis (PCA) but with different forms of inputs, can achieve higher accuracy at low SNR (as low as 2 dB). Due to the expressive capability of spectrogram and the image preprocessing steps, all of our methods are insensitive to unknown phase and frequency offsets, timing errors, and the arriving sequence of symbols.