CONVOLUTIONAL NEURAL NETWORK-BASED CHANNEL ENCODER CLASSIFICATION

Abstract

Channel encoders are essential in digital communication systems for correcting channel-induced random errors. In
most cases, the receiver has access to details on the transmitting end's channel encoders, including their kind and
characteristics. The kinds and characteristics of encoders may only be known to a limited extent or not at all in non-
cooperative situations, such as military communication systems. In this research, we investigate the possibility of
using a deep learning strategy to categorize four distinct kinds of encoders: polar, block, convolutional, and Bose
Chaudhuri-Hocquenghem (BCH). Our suggested method achieves classification accuracy surpassing 95% up to a
bit-error-rate (BER) value of 0.03 using a convolutional neural network (CNN) model. Also, when the input sample
length increases, the accuracy improves, according to the findings.