Authors:
Xiangxiang Zhang
and
Tao Luo
Affiliation:
Beijing University of Posts and Telecommunications, China
Keyword(s):
DHDV decoder, DnCNN, Convolutional code, Correlated noise.
Related
Ontology
Subjects/Areas/Topics:
Artificial Intelligence
;
Computational Intelligence
;
Evolutionary Computing
;
Knowledge Discovery and Information Retrieval
;
Knowledge-Based Systems
;
Machine Learning
;
Soft Computing
;
Symbolic Systems
Abstract:
Traditional Viterbi decoding algorithm degrades performance under correlated noise. Inspired by the powerful learning from data of deep neural networks, we apply denoising convolutional neural network (DnCNN) to channel decoding under correlated noise. In this paper, we propose a DnCNN hard decision Viterbi decoder (short for DHDV decoder) architecture to enhance the performance of convolutional codes under correlated noise. The architecture applies DnCNN to denoise and improves the decoding SNR (signal to noise ratio) through the correlation of noise correlation. Then, the Viterbi Algorithm decoder decodes from the denoised data. DHDV decoder obtains greater BER performance improvement under stronger noise correlation, and it is robust under different convolutional codes and correlated noise models. Comparing the complexity of DnCNN network and matrix multiplication whitening method, DnCNN network complexity is lower when the code length is longer.