Enhanced Image Captioning Using Features Concatenation and Efficient Pre-Trained Word Embedding

One of the issues in Computer Vision is the automatic development of descriptions for images,sometimes known as image captioning.Deep Learning techniques have made significant progress in this area.The typical architecture of image captioning systems consists mainly of an image feature extractor subsystem followed by a caption generation lingual subsystem.This paper aims to find optimized models for these two subsystems.For the image feature extraction subsystem,the research tested eight different concatenations of pairs of vision models to get among them the most expressive extracted feature vector of the image.For the caption generation lingual subsystem,this paper tested three different pre-trained language embedding models:Glove(Global Vectors for Word Representation),BERT(Bidirectional Encoder Representations from Transformers),and TaCL(Token-aware Contrastive Learning),to select from them the most accurate pre-trained language embedding model.Our experiments showed that building an image captioning system that uses a concatenation of the two Transformer based models SWIN(Shiftedwindow)and PVT(PyramidVision Transformer)as an image feature extractor,combined with the TaCL language embedding model is the best result among the other combinations.

De-Noising Brain MRI Images by Mixing Concatenation and Residual Learning(MCR)

Brain magnetic resonance images(MRI)are used to diagnose the different diseases of the brain,such as swelling and tumor detection.The quality of the brain MR images is degraded by different noises,usually salt&pepper and Gaussian noises,which are added to the MR images during the acquisition process.In the presence of these noises,medical experts are facing problems in diagnosing diseases from noisy brain MR images.Therefore,we have proposed a de-noising method by mixing concatenation,and residual deep learning techniques called the MCR de-noising method.Our proposed MCR method is to eliminate salt&pepper and gaussian noises as much as possible from the brain MRI images.The MCR method has been trained and tested on the noise quantity levels 2%to 20%for both salt&pepper and gaussian noise.The experiments have been done on publically available brain MRI image datasets,which can easily be accessible in the experiments and result section.The Structure Similarity Index Measure(SSIM)and Peak Signal-to-Noise Ratio(PSNR)calculate the similarity score between the denoised images by the proposed MCR method and the original clean images.Also,the Mean Squared Error(MSE)measures the error or difference between generated denoised and the original images.The proposed MCR denoising method has a 0.9763 SSIM score,84.3182 PSNR,and 0.0004 MSE for salt&pepper noise;similarly,0.7402 SSIM score,72.7601 PSNR,and 0.0041 MSE for Gaussian noise at the highest level of 20%noise.In the end,we have compared the MCR method with the state-of-the-art de-noising filters such as median and wiener de-noising filters.

基于1D-Concatenate的信道估计DNN模型优化方法

为提高DNN模型在无线通信中信道估计精度,提出一种基于1D-Concatenate的信道估计DNN模型优化方法。该方法将Concatenate进行一维(1D)数据转换,以跳跃连接的方式引入DNN模型,抑制梯度消失问题,运用1D-Concatenate恢复网络训练过程中丢失的数据特征,提高DNN信道估计精度。为验证优化方法的有效性,选取较典型的基于DNN的无线通信信道估计模型进行对比仿真实验。实验结果表明,本文提出的优化方法对已有DNN模型的估计增益提升可达77.10%,在高信噪比下信道增益提升可达3 dB。该优化方法能有效提高DNN模型在无线通信中的信道估计精度,特别是高信噪比下提升效果显著。

LLR Processing of Polar Codes in Concatenation Systems

This paper studies the decoding performance of low-density parity-check(LDPC)codes in a serial concatenation system with polar codes employing the successive cancellation(SC)decoding.It is known that the absolute incorrect log-likelihood ratio(LLR)values from the SC decoding can be very large.This phenomenon dramatically deteriorates the error correcting performance of the outer LDPC codes.In this paper,the LLR values of polar codes are regulated by a log processing before being sent to the LDPC decoder.Simulation results show that the log processing is an efficient approach with a low optimization complexity compared with the existing procedures to improve the performance of the serial concatenation systems.

Achieving Energy Efficiency in Wireless Sensor Networks Using Dynamic Channel Polling and Packet Concatenation

There has been a significant interest of researchers to combine different schemes focused on optimizing energy performance while developing aMAC protocol for Wireless Sensor Networks(WSNs).In this paper,we propose to integrate two cross-layer schemes:dynamic channel polling and packet concatenation using a recent asynchronous MAC protocol“Adaptive&Dynamic Polling MAC”(ADPMAC).ADP-MAC dynamically selects the polling interval distribution based on characterization of incoming traffic patterns using Coefficient of variation(CV).Packet Concatenation(PC)refers to combining the individually generated data packets into a single super packet and sending it at the polling instant.Also,the Block Acknowledgement(BA)scheme has been developed for ADP-MAC to work in conjunction with the packet concatenation.The proposed schemes have been implemented in Tiny-OS for Mica2 platform and Avrora emulator has been used for conducting experiments.Simulation results have revealed that the performance both in terms of energy&packet loss improves when ADP-MAC is used in conjunction with the additional features of PC&BA.Furthermore,the proposed scheme has been compared with a stateof-art packet concatenation primitive PiP(Packet-in-Packet).It has been observed that ADP-MAC supersedes the performance of PiP in terms of PDR(Packet Delivery Ratio)due to better management of synchronization between source and sink.

Multi-Core Virtual Concatenation Scheme Considering Inter-Core Crosstalk in Spatial Division Multiplexing Enabled Elastic Optical Networks

Spatial division multiplexing enabled elastic optical networks(SDM-EONs) are the potential implementation form of future optical transport networks, because it can curve the physical limitation of achievable transmission capacity in single-mode fiber and single-core fiber. However, spectrum fragmentation issue becomes more serious in SDM-EONs compared with simple elastic optical networks(EONs) with single mode fiber or single core fiber. In this paper, multicore virtual concatenation(MCVC) scheme is first proposed considering inter-core crosstalk to solve the spectrum fragmentation issue in SDM-EONs. Simulation results show that the proposed MCVC scheme can achieve better performance compared with the baseline scheme, i.e., single-core virtual concatenation(SCVC) scheme, in terms of blocking probability and spectrum utilization.

融合位置注意力机制与轻量化STDC网络的非结构化场景语义分割

近年来,非结构化道路分割已成为计算机视觉领域的重要研究方向之一.现有的大多数方法适合结构化道路的分割并无法满足非结构化道路分割的准确性与实时性需求.为了解决上述问题,本文对STDC网络进行改进,引入残差连接来更好地融合多尺度语义信息,还提出一种嵌入位置注意力模块的空洞空间卷积池化金字塔(PAASPP)来增强网络对道路等特定区域的位置感知能力.本文在RUGD与RELLIS-3D两个数据集上进行实验,所提出方法的MIoU在两个数据集的测试集上分别达到了50.78%和49.96%.

基于分层最小和译码的RS-LDPC级联码改进算法

RS-LDPC级联码能够有效提高数据传输的可靠性和系统的容错能力,在5G通信等领域中得到了广泛的应用,且在6G中具有很好的应用前景。但是RS-LDPC级联码与单码相比具有计算复杂度高、不易于在硬件上实现的不足,因此提出一种基于分层最小和的RS-LDPC级联码改进译码算法,将LDPC码的校验矩阵分解成多个子矩阵,在不同子矩阵层次上并行计算。此外,还引入了新的关于校验节点信息更新的简化函数,旨在保证译码性能的同时降低计算复杂度,达到易于硬件实现的目的。结果表明改进型分层最小和算法复杂度大大降低,且在性能上优于传统BP译码0.25 dB左右。

基于部分信息的SLT-LT联合码防窃听方案设计

近年来,信息的安全传输备受人们关注,现有的物理层安全技术从信息论的角度出发,将物理层安全编码与传输信道的动态物理特性进行结合,实现信息的保密传输。作为一种纠删码,LT(Luby transform)码由于其编码随机性、码率不固定等特性,使得窃听者不能直接从泄露的编码符号中得到有用信息,只要合法用户在窃听者之前接收到足够数量的编码符号,便可实现信息的安全传输。而作为一种转移LT(Shifted LT,SLT)码,SLT码能高效恢复信息的同时具有更小的译码开销。因此,我们将SLT码应用于Wyner降阶窃听信道模型进行研究,提出一种基于部分信息转移的SLT-LT联合码防窃听方案,信源利用接收者已知的部分信息对度分布进行调整,并对信源符号进行SLT-LT码级联编码。由于窃听信道是合法信道的降阶信道,因此外在的窃听者截获到的消息符号是合法接收者的降阶版本,在相同时间内,合法接收者能够收到更多消息符号,随着编解码过程的不断进行,合法接收者的优势不断累积,能够优先完成解码,而度分布的调整以及级联编码方案使得编码符号的平均度进一步增大,窃听者难以完成解码,进一步降低了窃听者译出率;之后,对所提方案的编解码性能以及安全性进行理论分析,并通过实验仿真进行验证,仿真结果表明,与其他防窃听LT方案相比,本文所提方案仅增加少量的译码开销但具有更好的安全性能。

Improved Belief Propagation Decoder for LDPC-CRC-Polar Codes with Bit-Freezing

Though belief propagation bit-flip(BPBF)decoding improves the error correction performance of polar codes,it uses the exhaustive flips method to achieve the error correction performance of CA-SCL decoding,thus resulting in high decoding complexity and latency.To alleviate this issue,we incorporate the LDPC-CRC-Polar coding scheme with BPBF and propose an improved belief propagation decoder for LDPC-CRC-Polar codes with bit-freezing(LDPCCRC-Polar codes BPBFz).The proposed LDPCCRC-Polar codes BPBFz employs the LDPC code to ensure the reliability of the flipping set,i.e.,critical set(CS),and dynamically update it.The modified CS is further utilized for the identification of error-prone bits.The proposed LDPC-CRC-Polar codes BPBFz obtains remarkable error correction performance and is comparable to that of the CA-SCL(L=16)decoder under medium-to-high signal-to-noise ratio(SNR)regions.It gains up to 1.2dB and 0.9dB at a fixed BLER=10-4compared with BP and BPBF(CS-1),respectively.In addition,the proposed LDPC-CRC-Polar codes BPBFz has lower decoding latency compared with CA-SCL and BPBF,i.e.,it is 15 times faster than CA-SCL(L=16)at high SNR regions.

On Some Properties of Graph of Prefix Code

We investigate decomposition of codes and finite languages. A prime decomposition is a decomposition of a code or languages into a concatenation of nontrivial prime codes or languages. A code is prime if it cannot be decomposed into at least two nontrivial codes as the same for the languages. In the paper, a linear time algorithm is designed, which finds the prime decomposition. If codes or finite languages are presented as given by its minimal deterministic automaton, then from the point of view of abstract algebra and graph theory, this automaton has special properties. The study was conducted using system for computational Discrete Algebra GAP. .

基于改进ResNet算法的太阳黑子分类方法研究

文中提出一种基于改进的ResNet的太阳黑子图像分类方法,该方法在ResNet的基础上引入了空洞卷积和残差连接等技术,增强了训练模型的特征提取能力以及感受野。在模型训练过程中,采用交叉熵损失函数和SGD优化器进行参数优化,以提高模型的准确性和泛化能力。其中深度可分离卷积被用于替代传统的卷积操作,以减少模型的参数量和计算量。最后,在太阳黑子图像分类的公共数据集上进行实验评估。实验结果表明,提出的基于改进的ResNet模型在太阳黑子图像分类任务上表现出较高的准确率和鲁棒性,相比于传统的ResNet模型,可以取得更好的分类效果。该方法为太阳黑子图像分类提供了一种新的思路和技术方案,对于太阳活动预测和环境监测等领域具有重要意义。

Concatenated Trellis Coding and CPM for DS/SSMA Communications in Rayleigh Fading Channel

本文在衰落信道下将级联网格编码调制（ＴＣＭ）和连续相位调制（ＣＰＭ）应用于直接序列扩频多址（ＤＳ／ＳＳＭＡ）系统中．作者分别采用了理论分析和数值模拟两种方法研究了级联ＴＣＭ／ＣＰＭＤＳ／ＳＳＭＡ系统的性能，结果表明在相同的条件下采用级联ＴＣＭ／ＣＰＭ编码调制方式的ＤＳ／ＳＳＭＡ系统的错误概率远小于采用单一ＴＣＭ或ＣＰＭ的ＤＳ／ＳＳＭＡ系统的错误概率．

面向6G空天通信的极化编码技术:需求、现状与展望

6G空天通信复杂多样的业务、场景对信息的可靠传输及系统适应能力提出了新的挑战。作为第一种理论可达信道容量的高性能编码,极化码是未来6G数据传输的重要候选方案。简要介绍了6G空天通信系统信息传输的难点及对信道编码技术的需求。对包括极化码编码构造、级联极化码、高性能译码算法在内的高可靠极化码增强技术,以及以速率适配、极化编码调制、极化编码混合重传为代表的高韧性极化码自适应传输技术等现状进行了综述。最后结合6G关键技术,展望未来极化码可能的发展方向。

冷凝+催化氧化处理技术在罐区及装车栈台油气处理系统中的应用

兰州石化公司采用乙烯运行一部储运区域罐区及装车栈台,在实际运行过程中,会排放出苯、甲苯、二甲苯等对环境有害的有机物,排放的尾气具有浓度高、波动大的特点,目前采用上海东化环境工程有限公司的“冷凝+催化氧化”工艺技术,对三苯罐区和装车栈台的油气进行综合治理。根据现场运行情况,不断优化尾气处理设施的运行参数,对装置进行技术改造,使罐区及装车站台的尾气达标排放,排放指标满足国标GB 31571—2015《石油化学工业污染物排放标准》中表6规定的特别污染物排放的要求,苯≤4 mg/Nm^(3),甲苯≤15 mg/Nm^(3),异丙苯≤50 mg/Nm^(3),非甲烷总烃转化率最高能达到99%以上,在系统保证安全的前提下,环保达标排放。

Iterative Decoding of Parallel Concatenated Block Codes and Coset Based MAP Decoding Algorithm for F24 Code

A multi dimensional concatenation scheme for block codes is introduced, in which information symbols are interleaved and re encoded for more than once. It provides a convenient platform to design high performance codes with flexible interleaver size. Coset based MAP soft in/soft out decoding algorithms are presented for the F24 code. Simulation results show that the proposed coding scheme can achieve high coding gain with flexible interleaver length and very low decoding complexity.

SERIALLY CONCATENATED CONTINUOUS PHASE MODULATION WITH REDUCED ITERATIVE DEMODULATION AND DETECTION

A reduced state Soft Input Soft Output (SISO) a posteriori probability algorithm for Seri-ally Concatenated Continuous Phase Modulation (SCCPM) is proposed in this paper. Based on the Reduced State Sequence Detection (RSSD),it has more general form compared with other reduced state SISO algorithms. The proposed algorithm can greatly reduce the state number,thus leads to the computation complexity reduction. It also minimizes the degradation in Euclidean distance with decision feedback in the reduced state trellis. Analysis and simulation results show that the perform-ance degradation is little with proper reduction scheme.

Chemical Synthesis and Cloning of Overlapping Concatenated Human Atrial Natriuretic Factor Gene with Gene Engineering

Eight oligonucleotide fragments were designed with the aid of a computer and synthesizedaccording to the amino add sequcnce of human atrial natriuretic factor(ANF).By means of an-nealing and ligation,these fragments were assembled into an overlapping concatenator consisting oftwo ANF genes ligated by TGATG for termination and initiation of translation.Theconcatenator was omserted into plasmid pRC23 and the recobinant DNA was transformed into E.coli strain TAP106.Analysis by restriction enzyme mapping,hybridization and DNA sequenongshowed that the orientation and reading frame of the gene were correct.

Decoding Technique of Concatenated HadamardCodes and Its Performance

The decoding technique of concatenated Hadamard codes and its performance are studied. Efficient soft in soft out decoding algorithms based on the fast Hadamard transform are developed. Performance required by CDMA mobile or PCS speech services, e.g. , BER=10 -3 , can be achieved at Eb/No =0.9 dB using short interleaving length of 192 bits.

A Novel Concatenated Chaotic Communication System

A strategy for a novel concatenated chaotic communication system is presented. The transmitter system comprises chaotic turbo encoder and logistic CSK block in a serially concatenated form. Chaotic turbo code is capable of reducing bit error rate (BER) of the chaotic system in the AWGN channel. Through the chaotic turbo encoder, the coded sequence, which has quasi-chaotic properties, will be transmitted into the logistic CSK block. Having a very sensitive dependence on initial conditions of the map, the logistic CSK block can also be taken as the chaotic authentication method. The receiver, which has logistic demodulation block and chaotic decoder, is a linear asymptotic approximation to the inverse of the transmitter system. A chaotic iterative soft-decision decoding algorithm is also developed based on conventional maximum A posteriori decoding algorithm. At last, a two-step authentication method of this chaotic system is also presented.