In this paper, we design a new class of high-efficiency deep joint source-channel coding methods to achieve end-to-end video transmission over wireless channels. The proposed methods exploit nonlinear transform and conditional coding architecture to adaptively extract semantic features across video frames, and transmit semantic feature domain representations over wireless channels via deep joint source-channel coding. Our framework is collected under the name deep video semantic transmission (DVST). In particular, benefiting from the strong temporal prior provided by the feature domain context, the learned nonlinear transform function becomes temporally adaptive, resulting in a richer and more accurate entropy model guiding the transmission of current frame. Accordingly, a novel rate adaptive transmission mechanism is developed to customize deep joint source-channel coding for video sources. It learns to allocate the limited channel bandwidth within and among video frames to maximize the overall transmission performance. The whole DVST design is formulated as an optimization problem whose goal is to minimize the end-to-end transmission rate-distortion performance under perceptual quality metrics or machine vision task performance metrics. Across standard video source test sequences and various communication scenarios, experiments show that our DVST can generally surpass traditional wireless video coded transmission schemes. The proposed DVST framework can well support future semantic communications due to its video content-aware and machine vision task integration abilities.

Gaussian approximation (GA) is widely used to construct polar codes. However, when the code length is long, the subchannel selection inaccuracy due to the calculation error of conventional approximate GA (AGA), which uses a two-segment approximation function, results in a catastrophic performance loss. In this paper, new principles to design the GA approximation functions for polar codes are proposed. First, we introduce the concepts of polarization violation set (PVS) and polarization reversal set (PRS) to explain the essential reasons that the conventional AGA scheme cannot work well for the long-length polar code construction. In fact, these two sets will lead to the rank error of subsequent subchannels, which means that the orders of subchannels are misaligned, which is a severe problem for polar code construction. Second, we propose a new metric, named cumulative-logarithmic error (CLE), to quantitatively evaluate the remainder approximation error of AGA in the logarithm. We derive the upper bound of CLE to simplify its calculation. Finally, guided by PVS, PRS, and CLE bound analysis, we propose new construction rules based on a multi-segment approximation function, which obviously improve the calculation accuracy of AGA so as to ensure the excellent performance of polar codes especially for the long code lengths. Numerical and simulation results indicate that the proposed AGA schemes are critical to constructing high-performance polar codes.

Non-orthogonal multiple access (NOMA) is one of the key techniques to address the high spectral efficiency and massive connectivity requirements for the fifth generation (5G) wireless networks. To efficiently realize NOMA, we propose a framework combining the binary polar coding and the NOMA transmission, which adequately utilizes the reliability distinctions among users. In this polar-coded NOMA (PC-NOMA) system, the NOMA transmission is recognized from the novel perspective of channel polarization. Combing the binary polar coding and signal modulation, the original NOMA channel is decomposed into multiple bit polarized channels by using a three-stage channel transform, that is, user$\rightarrow$signal $\rightarrow$bit partitions. Specifically, for the first stage channel transform, we design two schemes, namely the sequential user partition and the parallel user partition (PUP). For the SUP based PC-NOMA system, a “worst-goes-first” strategy is proposed to determine the user partition order which improves the system performance by the enhanced polarization effect among the user synthesized channels. Also, its performance is analyzed by using the channel polarization principle. In the receiver, a joint successive cancellation detection and decoding scheme is developed. For the PUP based PC-NOMA system, a parallel detection scheme is exploited to reduce the processing latency. Furthermore, a round-robin user scheduling is given for the above two PC-NOMA systems, which ensures the user fairness over multiple data blocks. The block error ratio and throughput performances indicate that the proposed PC-NOMA systems obviously outperform the state-of-the-art turbo-coded NOMA systems due to the advantages of joint design between the polar coding and the NOMA transmission.

In this paper, a novel wireless energy harvesting cooperative jammer (EH-CJ)-aided transmission scheme, which can enhance the security for cooperative cognitive radio networks (CCRNs), is proposed. The EH-CJ first harvests energy from ambient radio frequency (RF) signals transmitted from the primary transmitter (PT). This energy is then used to transmit jamming noise (JN), so that the performance of eavesdroppers' receptions is degraded, and also to deliver additional power to the secondary energy receiver (ER). In this way, the proposed scheme can enhance the security and extend the lifetime for CCRNs. For this system, we formulate an optimization problem for maximizing the secrecy rate of the secondary system under transmit power constraints at the secondary transmitter as well as the EH-CJ, minimum harvested energy at the secondary ER and information rate at the primary receiver (PR). Since this optimization problem is not convex, we propose a two-stage optimization algorithm based upon the Charnes-Cooper transformation and rank-one relaxation to solve it. Computer simulation results have demonstrated that the proposed EH-CJ-aided scheme achieves better secrecy rate performance, as compared to an isotropic JN scheme as well as compared to an equivalent scheme which does not use the aid of EH-CJ.

The polar neural decoder has a performance improvement at the cost of additional storage and computation. In this letter, we first propose a ResNet-like BP architecture to improve the standard polar BP decoding. Each iteration inherits a portion of information from the previous iteration as additional input and passes its output to the next iteration. Second, trainable damping factors, which can be learned through deep learning techniques, are applied to determine the inheritance ratio in the ResNet-like BP. We allocate the damping factors flexibly and demonstrate the optimization of a shared constant damping factor provides the same level of performance with the individual optimization of the damping factor at each state. Moreover, we qualitatively explain the performance gain by analyzing the mutual information between the input and output of the decoder. Numerical and simulation results indicate that the ResNet-like BP decoder can effectively improve the error correction performance and accelerate convergence compared with the standard polar BP algorithm.