Convolutional Neural Filtering for Intelligent Communications Signal Processing in Harsh Environments

Zhuo Sun,Jinpo Fan,Jingjing Li

doi:10.1109/access.2021.3049950

Abstract

Aiming at utilizing artificial neural networks to enhance intelligent filtering for interfered wireless communication signal in harsh environments, a new method named convolutional neural filtering is designed and presented in this paper. This method is based on model-driven deep learning princeple, by analyzing the theoretical connection between the filter model and the convolutional neural layer, it attempts to use one-dimensional convolution kernels to learn a matched or bandpass filter. Moreover, the model introduces a kernel-wise attention mechanism between different convolution kernels to selectively emphasize informative filters. The results show that in terms of interference and noise suppression for received wireless signal, the filtering method has highlighted dynamic adaptability to variation of signals and interference, and it also reveals that the performance is affected by the initialization parameters and the number of convolution kernels. Based on this method an embeddable filtering unit fully based on neural network is provided, which can be easily integrated into a deep learning network targeting such as wireless signal detection and recognition applications, avoiding complex preprocessing for end-to-end wireless signal learning.

Highlights

For communications in harsh environments, an intelligent physical (PHY) layer is fundamental and inevitable to achieving the envisioned communication requirements
In order to further optimize the effectiveness of the filter in the convolutional layer, we introduced an attention mechanism on the convolution kernels based on its contribution to the filtering performance, and proposed a convolutional neural filtering unit that combines the channel attention mechanism
As a popular structure of artificial neural networks, the convolutional neural layer is naturally consistent with the filter in signal processing in terms of mathematical model representation

Summary

INTRODUCTION

For communications in harsh environments, an intelligent physical (PHY) layer is fundamental and inevitable to achieving the envisioned communication requirements. Machine learning at PHY layer holds the potential to perform intelligent signal processing that can offer significant performance enhancements over traditional approaches [1] These approaches can be classified into two generic groups: data-driven [2] and model-driven [3]. To cope with the degraded generalization ability of deep learning under actual wireless transmission conditions, conventional solutions are to preprocess sampled signal, such as channel equalization and interference suppression by through adaptive filtering methods. As a popular structure of artificial neural networks, the convolutional neural layer is naturally consistent with the filter in signal processing in terms of mathematical model representation This has led to some research in deep learning applications outside of communication signals, such as audio or electroencephalo-graph(EEG) signals, to learn filtering for signal enhancement. We will innovatively use the linear modeling theoretical model to supervise our utilization of the convolutional neural layer on a different target, that is, the communication signal, to realize the process of intelligent filtering

FROM LINEAR FILTERING TO CONVOLUTIONAL

LEARNING FILTERING WITH SINGLE CONVOLUTIONAL LAYER

CONVOLUTIONAL NEURAL FILTER ENHANCED BY

Findings

CONCLUSION