Deep-Waveform: A Learned OFDM Receiver Based on Deep Complex-valued Convolutional Networks

Abstract

Recent explorations of Deep Learning in the physical layer (PHY) of wireless communication have shown the capabilities of Deep Neuron Networks in tasks like channel coding, modulation, and parametric estimation. However, it is unclear if Deep Neuron Networks could also learn the advanced waveforms of current and next-generation wireless networks, and potentially create new ones. In this paper, a Deep Complex Convolutional Network (DCCN) without explicit Discrete Fourier Transform (DFT) is developed as an Orthogonal Frequency-Division Multiplexing (OFDM) receiver. Compared to existing deep neuron network receivers composed of fully-connected layers followed by non-linear activations, the developed DCCN not only contains convolutional layers but is also almost (and could be fully) linear. Moreover, the developed DCCN not only learns to convert OFDM waveform with Quadrature Amplitude Modulation (QAM) into bits under noisy and Rayleigh channels, but also outperforms expert OFDM receiver based on Linear Minimum Mean Square Error channel estimator with prior channel knowledge in the low to middle Signal-to-Noise Ratios of Rayleigh channels. It shows that linear Deep Neuron Networks could learn transformations in signal processing, thus master advanced waveforms and wireless channels.

A Deep Learning Wireless Transceiver with Fully Learned Modulation and Synchronization

“…Signal parameters are estimated with a separate DNNs and the signal is then processed by manually programmed operations. The authors in [8] and [9] propose to learn merely the receiver side of the system and afterwards apply manual signal processing to support synchronization.…”

Robust Adversarial Attacks Against DNN-Based Wireless Communication Systems

“…They use expert feature based methods instead of naively learned features to improve the performance of the classification. Signal Detection in OFDM Systems: [38] and [39] deploy DNNs for channel estimation and signal detection in OFDM systems in an end-to-end manner. In [39], Zhao et al use CNNs to design an OFDM receiver outperforming the conventional OFDM receivers that are based on a Linear Minimum Mean Square Error channel estimator.…”

“…Signal Detection in OFDM Systems: [38] and [39] deploy DNNs for channel estimation and signal detection in OFDM systems in an end-to-end manner. In [39], Zhao et al use CNNs to design an OFDM receiver outperforming the conventional OFDM receivers that are based on a Linear Minimum Mean Square Error channel estimator. Ye et al [38] use DNNs to estimate Channel State Information (CSI) implicitly and recover the transmitted symbols directly instead of estimating CSI explicitly and detecting the transmitted symbols using the estimated CSI.…”

“…Neural Networks (DNNs) have recently become a part of various wireless communication systems. Example DNN-based wireless communication systems include autoencoder wireless communication [11], [13], [19], [23], [24], [26], modulation recognition (radio signal classification) [21], [27], [28], [30], [37], and OFDM channel estimation and signal detection [38], [39] which leverage the high utility of DNNs to improve the performance of their underlying wireless applications. However, DNNs are known to be susceptible to adversarial examples, which are small perturbations added to the inputs of a DNN causing it to misclassify the perturbed inputs.…”

“…Evaluation on major wireless systems. We have implemented and evaluated our attacks on three major classes of DNN-based wireless systems, specifically, autoencoder communication systems [11], [13], [19], [23], [24], [26], radio signal classification [21], [27], [28], [30], [37], and OFDM channel estimation and signal detection [38], [39]. We show that for all three applications, our attack is highly effective in corrupting the functionality of the underlying wireless systems, and at the same time offers strong undetectability and robustness.…”