在傳統的正交分頻多工(Orthogonal Frequency-Division Multiplexing, OFDM)
解調器之前加上非線性消隱，是一種改善正交分頻多工接收器在脈衝雜訊干擾的
環境中性能的常見方法。因為這種方法實踐簡單以及在性能改善方面是有效的，
所以被廣泛應用。然而傳統的非線性消隱的方法在實踐上需要已知的環境資訊去
計算出其最佳閾值，但是環境資訊在實際應用上通常是無法已知的。
因此，我們將探討經由正交分頻多工生成後的訊號，在包含多路徑衰減
(Multipath Fading)與脈衝雜訊(Impulsive Noise)干擾的環境中，是否能透過基於循
環神經網路(Recurrent Neural Network, RNN)之非線性消隱正交分頻多工接收器，
在不需要已知環境資訊的狀況下，能接近傳統方法的理想性能。
本文將以門控循環單元(Gated Recurrent Unit, GRU)作為循環神經網路的循
環神經單元，透過預定抽樣(Scheduled Sampling)的訓練方法，訓練基於循環神經
網路的異常偵測模型。後續模擬結果使用多路徑通道及白努利-高斯(BernoulliGaussian)脈衝雜訊通道的混合通道模型，最後與傳統非線性消隱方法做位元錯誤
率(BER)的性能比較。

Adding nonlinear fading before the traditional Orthogonal Frequency-Division
Multiplexing (OFDM) demodulator is a common method for improving the
performance of OFDM receivers in environments with impulsive noise interference.
Because this method is simple to implement and effective in performance improvement,
it is widely used. However, traditional nonlinear blanking methods require known
environmental information to calculate the optimal threshold, but environmental
information is usually unknown in practical applications.
Therefore, we will explore whether the signal generated by OFDM in an
environment with Multipath Fading and Impulsive Noise interference can approach the
ideal performance of traditional methods without needing known environmental
information through a Nonlinear Fading OFDM receiver based on a Recurrent Neural
Network (RNN).
In this paper, we use the Gated Recurrent Unit (GRU) as the recurrent neural unit
of the recurrent neural network, and train the anomaly detection model based on the
recurrent neural network using scheduled sampling. The subsequent simulation results
use a hybrid channel model of multipath channels and Bernoulli-Gaussian impulsive
noise channels, and finally compare the bit error rate (BER) performance with
traditional nonlinear fading methods.

參考文獻
[1] Barazideh, Reza et al. "Impulsive Noise Detection in Ofdm-Based Systems: A Deep Learning Perspective." 2019 IEEE 9th Annual Computing and Communication Workshop and Conference (CCWC), IEEE, 2019, pp. 0937-42.
[2] Ghosh, Monisha. "Analysis of the Effect of Impulse Noise on Multicarrier and Single Carrier Qam Systems." IEEE Transactions on Communications, vol. 44, no. 2, 1996, pp. 145-47.
[3] Kassam, Saleem A. Signal Detection in Non-Gaussian Noise. Springer Science & Business Media, 2012.
[4] Cho, Kyunghyun et al. "Learning Phrase Representations Using Rnn Encoder-Decoder for Statistical Machine Translation." arXiv preprint arXiv:1406.1078, 2014.
[5] Werbos, Paul J. "Backpropagation through Time: What It Does and How to Do It." Proceedings of the IEEE, vol. 78, no. 10, 1990, pp. 1550-60.
[6] Williams, Ronald J and Jing Peng. "An Efficient Gradient-Based Algorithm for on-Line Training of Recurrent Network Trajectories." Neural computation, vol. 2, no. 4, 1990, pp. 490-501.
[7] Williams, Ronald J and David Zipser. "A Learning Algorithm for Continually Running Fully Recurrent Neural Networks." Neural computation, vol. 1, no. 2, 1989, pp. 270-80.
[8] Bengio, Samy et al. "Scheduled Sampling for Sequence Prediction with Recurrent Neural Networks." Advances in neural information processing systems, vol. 28, 2015.
[9] Malhotra, Pankaj et al. "Lstm-Based Encoder-Decoder for Multi-Sensor Anomaly Detection." arXiv preprint arXiv:1607.00148, 2016.
[10] Zhidkov, Sergey V. "Analysis and Comparison of Several Simple Impulsive Noise Mitigation Schemes for Ofdm Receivers." IEEE Transactions on Communications, vol. 56, no. 1, 2008, pp. 5-9.
[11] Ye, Hao et al. "Power of Deep Learning for Channel Estimation and Signal Detection in Ofdm Systems." IEEE Wireless Communications Letters, vol. 7, no. 1, 2017, pp. 114-17.