Blind Identification of Interleavers Using Deep Learning Neural Network

Channel encoders and interleavers play a significant role in correcting the random and burst errors introduced by the noisy channel in a digital communication system. Usually, the information regarding the type of channel encoders and interleavers along with their parameters, which are used at the transmitting end, will be available at the receiver. But in military communication systems, the encoder/interleaver type and parameters are either partially known or unknown and in case of reconfigurable receivers and adaptive modulation and coding (AMC)-based systems, blind identification of chan-nel encoders and interleavers helps in improving the spectral efficiency. In this paper, we have explored the possibility of utilizing a deep learning approach for identifying three different types of interleavers, which are block, convolutional, and helical interleavers. We employ convolutional neural network (CNN) model for the classification task and the input data to the interleavers consist of block encoded and convolutional encoded data. In both the cases, our CNN model achieves classification accuracy exceeding 95% at varying signal-to-noise ratio (SNR) values. Our findings also indicate that the accuracy improves with the input sample length and it is better for convolutional encoded data. Finally, at the same bit error rate (BER) value, our model demonstrates superior classification accuracy compared to the existing algebraic rank-based approach available in the literature.