A Video Captioning Method by Semantic Topic-Guided Generation

In the video captioning methods based on an encoder-decoder,limited visual features are extracted by an encoder,and a natural sentence of the video content is generated using a decoder.However,this kind ofmethod is dependent on a single video input source and few visual labels,and there is a problem with semantic alignment between video contents and generated natural sentences,which are not suitable for accurately comprehending and describing the video contents.To address this issue,this paper proposes a video captioning method by semantic topic-guided generation.First,a 3D convolutional neural network is utilized to extract the spatiotemporal features of videos during the encoding.Then,the semantic topics of video data are extracted using the visual labels retrieved from similar video data.In the decoding,a decoder is constructed by combining a novel Enhance-TopK sampling algorithm with a Generative Pre-trained Transformer-2 deep neural network,which decreases the influence of“deviation”in the semantic mapping process between videos and texts by jointly decoding a baseline and semantic topics of video contents.During this process,the designed Enhance-TopK sampling algorithm can alleviate a long-tail problem by dynamically adjusting the probability distribution of the predicted words.Finally,the experiments are conducted on two publicly used Microsoft Research Video Description andMicrosoft Research-Video to Text datasets.The experimental results demonstrate that the proposed method outperforms several state-of-art approaches.Specifically,the performance indicators Bilingual Evaluation Understudy,Metric for Evaluation of Translation with Explicit Ordering,Recall Oriented Understudy for Gisting Evaluation-longest common subsequence,and Consensus-based Image Description Evaluation of the proposed method are improved by 1.2%,0.1%,0.3%,and 2.4% on the Microsoft Research Video Description dataset,and 0.1%,1.0%,0.1%,and 2.8% on the Microsoft Research-Video to Text dataset,respectively,compared with the existing video captioning methods.As a result,the proposed method can generate video captioning that is more closely aligned with human natural language expression habits.

Low complexity joint source-channel decoding for transmission of wavelet compressed images

To utilize residual redundancy to reduce the error induced by fading channels and decrease the complexity of the field model to describe the probability structure for residual redundancy, a simplified statistical model for residual redundancy and a low complexity joint source-channel decoding（JSCD） algorithm are proposed. The complicated residual redundancy in wavelet compressed images is decomposed into several independent 1-D probability check equations composed of Markov chains and it is regarded as a natural channel code with a structure similar to the low density parity check （LDPC） code. A parallel sum-product （SP） and iterative JSCD algorithm is proposed. Simulation results show that the proposed JSCD algorithm can make full use of residual redundancy in different directions to correct errors and improve the peak signal noise ratio （PSNR） of the reconstructed image and reduce the complexity and delay of JSCD. The performance of JSCD is more robust than the traditional separated encoding system with arithmetic coding in the same data rate.

JOINT ITERATIVE DECODING FOR SIMPLE-ENCODING SYSTEMATIC IRREGULAR-LDPC-BASED CODED COOPERATION IN NON-IDEAL RELAY CHANNEL

In this paper, a new kind of simple-encoding irregular systematic LDPC codes suitable for one-relay coded cooperation is designed, where the proposed joint iterative decoding is effectively performed in the destination which is in accordance with the corresponding joint Tanner graph characterizing two different component LDPC codes used by the source and relay in ideal and non-ideal relay cooperations. The theoretical analysis and simulations show that the coded cooperation scheme obviously outperforms the coded non-cooperation one under the same code rate and decoding complex. The significant performance improvement can be virtually credited to the additional mutual exchange of the extrinsic information resulted by the LDPC code employed by the source and its counterpart used by the relay in both ideal and non-ideal cooperations.

Integrated Joint Source-Channel Symbol-by-Symbol Decoding of Variable-Length Codes Using 3-D MAP Sequence Estimation

Most of multimedia schemes employ variable-length codes （VLCs） like Huffman code as core components in obtaining high compression rates. However VLC methods are very sensitive to channel noise. The goal of this paper is to salvage as many data from the damaged packets as possible for higher audiovisual quality. This paper proposes an integrated joint source-channel decoder （I-JSCD） at a symbol-level using three-dimensional （3-D） trellis representation for first-order Markov sources encoded with VLC source code and convolutional channel code. This method combines source code and channel code state-spaces and bit-lengths to construct a two-dimensional （2-D） state-space, and then develops a 3-D trellis and a maximum a-posterior （MAP） algorithm to estimate the source sequence symbol by symbol. Experiment results demonstrate that our method results in significant improvement in decoding performance, it can salvage at least half of （50%） data in any channel error rate, and can provide additional error resilience to VLC stream like image, audio, video stream over high error rate links.

JOINT SOURCE-CHANNEL DECODING OF HUFFMAN CODES WITH LDPC CODES

In this paper, we present a Joint Source-Channel Decoding algorithm (JSCD) for Low-Density Parity Check (LDPC) codes by modifying the Sum-Product Algorithm (SPA) to account for the source redun-dancy, which results from the neighbouring Huffman coded bits. Simulations demonstrate that in the presence of source redundancy, the proposed algorithm gives better performance than the Separate Source and Channel Decoding algorithm (SSCD).

Improved List Sphere Decoder for Multiple Antenna Systems

An improved list sphere decoder （ILSD） is proposed based on the conventional list sphere decoder （LSD） and the reduced- complexity maximum likelihood sphere-decoding algorithm. Unlike the conventional LSD with fixed initial radius, the ILSD adopts an adaptive radius to accelerate the list cdnstruction. Characterized by low-complexity and radius-insensitivity, the proposed algorithm makes iterative joint detection and decoding more realizable in multiple-antenna systems. Simulation results show that computational savings of ILSD over LSD are more apparent with more transmit antennas or larger constellations, and with no performance degradation. Because the complexity of the ILSD algorithm almost keeps invariant with the increasing of initial radius, the BER performance can be improved by selecting a sufficiently large radius.

Joint Source-Channel Decoding Scheme for Image Transmission over Wireless Channel

We improve the iterative decoding algorithm by utilizing the “leaked” residual redundancy at the output of the source encoder without changing the encoder structure for the noisy channel. The experimental results show that using the residual redundancy of the compressed source in channel decoding is an effective method to improve the error correction performance.

Integrated Joint Source-Channel Decoding for H.264 Video Transmission

This paper proposes an integrated joint source-channel decoder （I-JSCD） using Max-Log-MAP method for sources encoded with exp-Golomb codes and convolutional codes, and proposes a system applying this method to decoding the VLC data, e.g. motion vector differences （MVDs）, of H.264 across an AWGN channel. This method combines the source code state-space and the channel code state-space together to construct a joint state-space, develops a 3-D trellis and a maximum a-posterior （MAP） algorithm to estimate the source sequence symbol by symbol, and then uses max-log approximation to simplify the algorithm. Experiments indicate that the proposed system gives significant improvements on peak signal-to-noise ratio （PSNR） （maximum about 15 dB） than a separate scheme. This also leads to a higher visual quality of video stream over a highly noisy channel.

Joint Iterative Decoding for Network-Coding-Based Multisource LDPC-Coded Cooperative MIMO

A network-coding-based multisource LDPC-coded cooperative MIMO scheme is proposed,where multiple sources transmit their messages to the destination with the assistance from a single relay.The relay cooperates with multiple sources simultaneously via network-coding.It avoids the issues of imperfect frequency/timing synchronization and large transmission delay which may be introduced by frequency-division multiple access(FDMA)/code-division multiple access(CDMA)and time-division multiple access(TDMA)manners.The proposed joint″Min-Sum″iterative decoding is effectively carried out in the destination.Such a decoding algorithm agrees with the introduced equivalent joint Tanner graph which can be used to fully characterize LDPC codes employed by the sources and relay.Theoretical analysis and numerical simulation show that the proposed scheme with joint iterative decoding can achieve significant cooperation diversity gain.Furthermore,for the relay,compared with the cascade scheme,the proposed scheme has much lower complexity of LDPC-encoding and is easier to be implemented in the hardware with similar bit error rate(BER)performance.

Context Information and Fragments Based Cross-Domain Word Segmentation

A new joint decoding strategy that combines the character-based and word-based conditional random field model is proposed.In this segmentation framework,fragments are used to generate candidate Out-of-Vocabularies(OOVs).After the initial segmentation,the segmentation fragments are divided into two classes as "combination"(combining several fragments as an unknown word) and "segregation"(segregating to some words).So,more OOVs can be recalled.Moreover,for the characteristics of the cross-domain segmentation,context information is reasonably used to guide Chinese Word Segmentation(CWS).This method is proved to be effective through several experiments on the test data from Sighan Bakeoffs 2007 and Bakeoffs 2010.The rates of OOV recall obtain better performance and the overall segmentation performances achieve a good effect.

Equalization for continuous phase modulation over frequency-selective fading channels

This paper presents an equalization algorithm for continuous phase modulation （CPM） over frequency-selective channels. A specific training sequence is first embedded in each data packet. By recursive least-squares （RLS） estimation, the channel information parameters can be acquired, and a fractionally Simulation results show that the proposed algorithm can acquire the spaced equalizer performs joint decoding and equalization. channel information parameters rapidly and accurately, and that the fractionally spaced equalizer can eliminate the intersymbol interference （ISI） effectively, and is not sensitive to timing inaccuracy, so this algorithm can be exploited for demodulation system in burst mode.

NEW ITERATIVE SUPER-TRELLIS DECODING WITH SOURCE A PRIORI INFORMATION FOR VLCS WITH TURBO CODES

A novel Joint Source and Channel Decoding (JSCD) scheme for Variable Length Codes (VLCs) concatenated with turbo codes utilizing a new super-trellis decoding algorithm is presented in this letter. The basic idea of our decoding algorithm is that source a priori information with the form of bit transition probabilities corresponding to the VLC tree can be derived directly from sub-state transitions in new composite-state represented super-trellis. A Maximum Likelihood (ML) decoding algorithm for VLC sequence estimations based on the proposed super-trellis is also described. Simu-lation results show that the new iterative decoding scheme can obtain obvious encoding gain especially for Reversible Variable Length Codes (RVLCs),when compared with the classical separated turbo decoding and the previous joint decoding not considering source statistical characteristics.

V-BLAST BASED LDPC-CODED RELAY COOPERATION

An efficient LDPC-coded multi-relay cooperation architecture is proposed based on virtual vertical Bell Labs layered space-time （V-BLAST） processing for uplink communication, where minimum-mean-square-error （MMSE） and BP-based joint iterative decoding based on the introduced muhi-layer Tanner graph are effectively de- signed to detect and decode the corrupted received sequence at the destination. By introducing V-BLAST transmis- sion to the coded multi-relay cooperation, relays send their streams of symbols simultaneously, which increases the data rate and significantly reduces the transmission delay. The theoretical analysis and numerical results show that the new LDPC coded cooperation scheme outperforms the coded non-cooperation under the same code rate, and it also achieves a good trade-off among the performance, signal delay, and the encoding complexity associated with the number of relays. The performance gain can be credited to the proposed V-BLAST processing architecture and BP-based joint iterative decoding by the introduced multi-layer Tanner graph at a receiver-side.

Adaptive protograph-based BICM-ID relying on the RJ-MCMC algorithm:a reliable and efficient transmission solution for plasma sheath channels

For reentry communication,owing to the influence of the highly dynamic plasma sheath(PS),the parasitic modulation effect can occur and the received phase shift keying(PSK)signal constellation can be severely rotated,leading to unacceptable demodulation performance degradation.In this work,an adaptive non-coherent bit-interleaved coded modulation with iterative decoding(BICM-ID)system with binary PSK(BPSK)modulation and protograph lowdensity parity-check under the PS channel is proposed.The proposed protograph-based BICMID(P-BICM-ID)system can achieve joint processing of demodulation and decoding,where the soft information is adaptively estimated by reversible-jump Markov chain Monte Carlo(RJMCMC)algorithms.Simulation results indicate that compared to existing algorithms,the proposed system can adapt well to the dynamic characteristics of the PS channel and can obtain a 5dB performance improvement at a bit error rate of 10^(-6).

An LDPC coded cooperative MIMO scheme over Rayleigh fading channels with unknown channel state information

This paper describes a coded cooperative multiple-input multiple-output(MIMO) scheme,where structured low-density parity-check(LDPC) codes belonging to a family of repeat-accumulate(RA) codes are employed.The outage probability of the scheme over Rayleigh fading channels is deduced.In an unknown channel state information(CSI) scenario,adaptive transversal filters based on a spatio-temporal recursive least squares(ST-RLS) algorithm are adopted in the destination to realize receive diversity gain.Also,a joint 'Min-Sum' iterative decoding is effectively carried out in the destination.Such a decoding algorithm agrees with the bilayer Tanner graph that can be used to fully characterize two distinct structured LDPC codes employed by the source and relay.Simulation results verify the effectiveness of the adopted filter in the coded cooperative MIMO scheme.Theoretical analysis and numerical simulations show that the LDPC coded cooperative MIMO scheme can well combine cooperation diversity,multi-receive diversity,and channel coding gains,and clearly outperforms coded noncooperation schemes under the same conditions.

Analysis of Outage Capacity of NOMA： SIC vs. JD

In fifth-generation wireless communication networks, Non-Orthogonal Multiple Access（NOMA） has attracted much attention in both academic and industrial fields because of its higher spectral efficiency in comparison with orthogonal multiple access. Recently, numerous uplink NOMA techniques have been proposed,some of which are based on Successive Interference Cancellation（SIC） and others on Joint Decoding（JD, or simultaneous decoding）. In this study, we analyze the outage capacities of SIC and JD in the case of single-block transmission over a two-user Gaussian multiple-access channel with partial channel state information at transmitter from the perspective of information theory. Results of the analysis and numerals show that compared to SIC, JD can achieve a sum-rate gain of up to 10% or sum-power gain of 0.8 dB.