Frame Synchronization by Using Convolutional Neural Network

Abstract

Background/ObjectivesIn packet-based wireless communication system, when the packet starting point is unknown to the receiver, frame synchronization must be performed to find the starting point.The purpose of this paper is to develop a new frame synchronization method by using convolutional neural network (CNN) to improve the performance at low signal to noise ratio (SNR).

Methods/Statistical analysis: CNN, one of the deep learning methods, are widely used because of its excellent performance in image processing. In this paper, we propose a new frame synchronization method by transforming the frame synchronization problem into CNN problem. Specifically, the correlator output for frame synchronization is transformed into a 2D image and input to the CNN, and the CNN finds a frame or packet starting point.

Findings: Conventional frame synchronization methods recognize the start of a frame when the correlator output exceeds a certain threshold. However, since the optimal threshold must be determined as a function of SNR, it is a burden of estimating SNR of the received signal before performing frame synchronization. However, the proposed frame synchronization based on CNN does not need SNR estimation and shows better performance than the conventional method.

Improvements/ApplicationsWe compare false detection probability of the proposed CNN method with the convolutional technique through computer simulation. According to the result, the proposed CNN method has about 2dB performance gain compared to the conventional method in AWGN environments.

Keywords: CNN, 2D transformation, Frame synchronization, Deep learning, Synchronized communication system