Abstract: The subtle features extraction of signal is the key to specific communication emitter identification. Considering the nonlinearity and nonstationarity of measured emitter signals, several nonlinear dynamic characteristics are extracted as fingerprint features for specific emitters identification, which measure the nonlinear complexity of signals. Firstly, the raw signal is decomposed with Intrinsic Time-scale Decomposition (ITD) algorithm and several proper components are extracted by correlation coefficient. Then, permutation entropy, approximate entropy and sample entropy of each signal component are extracted as feature vector. The choices of relevant parameters are optimized by experiment and Support Vector Machine (SVM) is used for the classification of signals. Subtle features extraction and classification are conducted utilizing measured ship communication signals, and compared with several feature extraction methods, proposed method gets better identification performance with increase of signal types, which proved the validity of multi-scale analysis for extracting non-linear parameters.