TY  - JOUR
T1  - Autoencoder-based Channel Denoiser for Block Codes
AU  - Cho, Yeji 
AU  - Kwon, Nahyeon 
AU  - Kim, Junghyun 
AU  - Song, Hong-Yeop 
JO  - The Journal of Korean Institute of Communications and Information Sciences
PY  - 2024
DA  - 2024/1/1
DO  - 10.7840/kics.2024.49.11.1607
KW  - Channel noise
KW  - Denoiser
KW  - Autoencoder
KW  - Block codes
KW  - Deep learning
AB  - This paper proposes a new neural network-based channel denoiser designed to remove wireless channel noise before data decoding, thereby improving block code decoding performance. The proposed channel denoiser utilizes an autoencoder structure known to be effective in noise removal within image data. For performance comparison with the existing multi-layer perceptron-based model, the same neural network decoder was applied to Polar codes, LDPC codes, and BCH codes, and the denoiser effect was tested in AWGN and Rayleigh fading channels. Experiments confirmed that the proposed model utilizing an autoencoder-based denoiser significantly outperforms the traditional model using a multi-layer perceptron-based denoiser regarding block error rate across all assumed channel environments and channel codes.