Noise Suppression Chanel Estimation Method Using Deep Learning in IEEE 802.11p Standard

In this paper, we propose a channel estimation method based on a complex valued regression of the neural network for the IEEE 802.11p standard. It consists of the complex weighted summation optimized by feedforward neural network with backpropagation algorithm using initial estimated channel of the pilot and the long preamble. It also exploits the shift matrix in order to mitigate the effect from a systemic problems in IEEE 802.11p standards. The major problems of IEEE 802.11p standard are wide bandwidth of 10 MHz consisting of 64 subcarriers and relatively insufficient four pilot subcarriers at single ODFM symbol, which are unsuitable for a channel of vehicular environment. Despite these problems, the proposed method performs better than the conventional channel estimation methods. The performance of proposed scheme is provided with the comparison between constructed data pilots (CDP), Spectral Temporal Averaging (STA), and proposed scheme. The proposed channel estimation scheme has low mean square error (MSE) and bit error rate (BER) throughout the whole SNR region. It is the result from properly trained weight. At the low SNR region, especially, the performance of proposed scheme is much better than CDP and STA scheme. It is because of the noise suppression effect caused by a weighted summation algorithm.