Signal Detection Scheme Based on Adaptive Ensemble Deep Learning Model

Accurate signal detection is one of the most important requirements of wireless communication systems. The two most important processes of the signal detection are channel estimation and compensation. Since an orthogonal frequency-division multiplexing (OFDM) system among wireless communication systems uses orthogonal sub-carriers, the system has advantages of high bandwidth efficiency and simple channel compensation process compared with a single carrier system. However, due to the imperfect channel estimation and amplification of receiving noise by the channel compensation process, the reliability performance of the system is deteriorated according to the channel condition. This paper aims to introduce a new signal detection scheme based on deep learning. To address the challenges of signal detection, we propose the method which integrates the ensemble deep learning with the acquired received signals from multi-path channel according to the channel condition. The channel estimation corresponds to learning the deep neural network, and the channel compensation corresponds to assigning the received data to the learned network. Experiments on the OFDM symbol classification demonstrate that the proposed scheme has dramatically improved reliability performance compared with the conventional scheme.