This paper looks into the technology classification problem for a distributedwireless spectrum sensing network. First, a new data-driven model for AutomaticModulation Classification (AMC) based on long short term memory (LSTM) isproposed. The model learns from the time domain amplitude and phase informationof the modulation schemes present in the training data without requiring expertfeatures like higher order cyclic moments. Analyses show that the proposedmodel yields an average classification accuracy of close to 90% at varying SNRconditions ranging from 0dB to 20dB. Further, we explore the utility of thisLSTM model for a variable symbol rate scenario. We show that a LSTM based modelcan learn good representations of variable length time domain sequences, whichis useful in classifying modulation signals with different symbol rates. Theachieved accuracy of 75% on an input sample length of 64 for which it was nottrained, substantiates the representation power of the model. To reduce thedata communication overhead from distributed sensors, the feasibility ofclassification using averaged magnitude spectrum data, or online classificationon the low cost sensors is studied. Furthermore, quantized realizations of theproposed models are analyzed for deployment on sensors with low processingpower.
展开▼
