Noise Learning Based Denoising Autoencoder

TL;DR

This paper proposes a noise learning based denoising autoencoder (nlDAE) that learns noise patterns for improved denoising, demonstrating effectiveness in signal restoration, demodulation, and localization with less data and smaller models.

Contribution

The novel nlDAE modifies the traditional DAE by explicitly learning noise, enhancing denoising performance especially when noise is simpler to regenerate.

Findings

Requires smaller latent space dimension

Needs less training data

Outperforms traditional DAE in case studies

Abstract

This letter introduces a new denoiser that modifies the structure of denoising autoencoder (DAE), namely noise learning based DAE (nlDAE). The proposed nlDAE learns the noise of the input data. Then, the denoising is performed by subtracting the regenerated noise from the noisy input. Hence, nlDAE is more effective than DAE when the noise is simpler to regenerate than the original data. To validate the performance of nlDAE, we provide three case studies: signal restoration, symbol demodulation, and precise localization. Numerical results suggest that nlDAE requires smaller latent space dimension and smaller training dataset compared to DAE.