Abstract
In this paper, we propose a deep neural network (DNN)-based automatic modulation classification (AMC) for digital communications. While conventional AMC techniques perform well for additive white Gaussian noise (AWGN) channels, classification accuracy degrades for fading channels where the amplitude and phase of channel gain change in time. The key contributions of this paper are in two phases. First, we analyze the effectiveness of a variety of statistical features for AMC task in fading channels. We reveal that the features that are shown to be effective for fading channels are different from those known to be good for AWGN channels. Second, we introduce a new enhanced AMC technique based on DNN method. We use the extensive and diverse set of statistical features found in our study for the DNN-based classifier. The fully connected feedforward network with four hidden layers are trained to classify the modulation class for several fading scenarios. Numerical evaluation shows that the proposed technique offers significant performance gain over the existing AMC methods in fading channels.
Highlights
In digital communications, a task of classifying a modulation class from the received signal is referred to as automatic modulation classification (AMC) [1,2,3,4]
We reveal that the features found to be powerful for fading channels can be quite different from those widely used for additive white Gaussian noise (AWGN) channels
We observe that in fading channels, the performance of the proposed AMC method is slightly degraded for higher doppler frequency but significant performance gain of the proposed method is retained
Summary
A task of classifying a modulation class from the received signal is referred to as automatic modulation classification (AMC) [1,2,3,4]. While conventional communication systems use training signal or control channels to provide the information on the modulation type to the receiver, there are some scenarios where such information is not available and the modulation type should be blindly estimated from the received data. The first approach is to build a probabilistic model for the received signal and classify a modulation class based on some optimality criterion such as maximum likelihood function. Though these methods offer optimal detection performance for the given model, its classification accuracy degrades in the presence of model mismatch. The alternative approach is the machine learning-based approach, where the machine is trained to classify the modulation type based on the training data in off-line and the trained machine is deployed to apply to the real
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
