Correct Decoding Research Articles

High Fidelity Makeup via 2D and 3D Identity Preservation Net

In this article, we address the challenging makeup transfer task, aiming to transfer makeup from a reference image to a source image while preserving facial geometry and background consistency. Existing deep neural network-based methods have shown promising results in aligning facial parts and transferring makeup textures. However, they often neglect the facial geometry of the source image, leading to two adverse effects: (1) alterations in geometrically relevant facial features, causing face flattening and loss of personality, and (2) difficulties in maintaining background consistency, as networks cannot clearly determine the face-background boundary. To jointly tackle these issues, we propose the High Fidelity Makeup via two-dimensional (2D) and 3D Identity Preservation Network (IP23-Net), to the best of our knowledge, a novel framework that leverages facial geometry information to generate more realistic results. Our method comprises a 3D Shape Identity Encoder, which extracts identity and 3D shape features. We incorporate a 3D face reconstruction model to ensure the three-dimensional effect of face makeup, thereby preserving the characters’ depth and natural appearance. To preserve background consistency, our Background Correction Decoder automatically predicts an adaptive mask for the source image, distinguishing the foreground and background. In addition to popular benchmarks, we introduce a new large-scale High Resolution Synthetic Makeup Dataset containing 335,230 diverse high-resolution face images to evaluate our method’s generalization ability. Experiments demonstrate that IP23-Net achieves high-fidelity makeup transfer while effectively preserving background consistency. The code will be made publicly available.

Применение голографического кодирования для повышения надежности передачи информации в зашумленных каналах связи

Two options for using previously developed holographic coding to increase the reliability of information transmission in a communication channel with an unmanned aerial vehicle are considered – digital channel coding and spectral coding. It is shown that the holographic code has high noise immunity and provides a probability of correct decoding of more than 0.9999 with a probability of error in the channel of 0.34. Spectral holographic coding ensures reliable communication in the presence of noise and intentional interference, the power of which is 10 dB higher than the signal power, which in turn ensures the secrecy of the communication channel. An additional advantage of spectral coding is increased resistance to intersymbol interference, which makes it possible to eliminate the introduction of guard intervals and increase spectral efficiency, and reduces the requirements for stability of the sampling frequency when performing FFT in the receiver demodulator.

Denoising and Baseline Correction Methods for Raman Spectroscopy Based on Convolutional Autoencoder: A Unified Solution.

Preprocessing plays a key role in Raman spectral analysis. However, classical preprocessing algorithms often have issues with reducing Raman peak intensities and changing the peak shape when processing spectra. This paper introduces a unified solution for preprocessing based on a convolutional autoencoder to enhance Raman spectroscopy data. One is a denoising algorithm that uses a convolutional denoising autoencoder (CDAE model), and the other is a baseline correction algorithm based on a convolutional autoencoder (CAE+ model). The CDAE model incorporates two additional convolutional layers in its bottleneck layer for enhanced noise reduction. The CAE+ model not only adds convolutional layers at the bottleneck but also includes a comparison function after the decoding for effective baseline correction. The proposed models were validated using both simulated spectra and experimental spectra measured with a Raman spectrometer system. Comparing their performance with that of traditional signal processing techniques, the results of the CDAE-CAE+ model show improvements in noise reduction and Raman peak preservation.

Collaborative FFE-assisted simplified soft-output MLSE with LDPC for high-speed IM-DD transmissions.

In this paper, we propose a feed-forward equalizer (FFE)-assisted simplified soft-output MLSE (sMLSE) by collaborating the maximum likelihood sequence estimation (MLSE) with soft-decision low-density-parity-check (LDPC) decoding. The simplified sMLSE results in undetermined log-likelihood ratio (LLR) magnitudes when the reserved level is less than or equal to the half of modulation order. This severely degrades the performance of soft-decision forward error correction (SD-FEC) decoding. In the FFE-assisted simplified sMLSE, we use the LLRs calculated from pre-set FFE to replace these undetermined LLRs of simplified sMLSE. Thus, the proposed method eliminates the SD-FEC decoding performance degradation resulted from simplification. We conduct experiments to transmit 184-Gb/s PAM-4 or 255-Gb/s PAM-8 signal in IM-DD system at C-band to evaluate the performance of the proposed sMLSE. The results show that the proposed sMLSE can effectively compensate for the degradation of LLR quality. For 255-Gb/s PAM-8 signal transmissions, the FFE-assisted simplified sMLSE achieves almost the same SD-FEC decoding performance as the conventional sMLSE but with 85% complexity reduction.

Dynamic decoding and dual synthetic data for automatic correction of grammar in low-resource scenario.

Grammar error correction systems are pivotal in the field of natural language processing (NLP), with a primary focus on identifying and correcting the grammatical integrity of written text. This is crucial for both language learning and formal communication. Recently, neural machine translation (NMT) has emerged as a promising approach in high demand. However, this approach faces significant challenges, particularly the scarcity of training data and the complexity of grammar error correction (GEC), especially for low-resource languages such as Indonesian. To address these challenges, we propose InSpelPoS, a confusion method that combines two synthetic data generation methods: the Inverted Spellchecker and Patterns+POS. Furthermore, we introduce an adapted seq2seq framework equipped with a dynamic decoding method and state-of-the-art Transformer-based neural language models to enhance the accuracy and efficiency of GEC. The dynamic decoding method is capable of navigating the complexities of GEC and correcting a wide range of errors, including contextual and grammatical errors. The proposed model leverages the contextual information of words and sentences to generate a corrected output. To assess the effectiveness of our proposed framework, we conducted experiments using synthetic data and compared its performance with existing GEC systems. The results demonstrate a significant improvement in the accuracy of Indonesian GEC compared to existing methods.

Software for assessment and monitoring the decoding skills development of children from the elementary school: validity based on response process.

To continue the validation process of the Decoding Development Monitoring Protocol (PRADE) in software format in the validity evidence stage based on response processes. 250 individuals participated in this study, 125 individuals from private schools and 125 individuals from public schools. The assessment was carried out in person using the software that hosts the instrument's tasks, which are organized into decoding linguistically balanced words and non-words, respecting the decoding rules of Brazilian Portuguese. The software prepares an individual performance report for each participant, counting the decoding time for each stimulus, as well as the number of words decoded correctly. The data is organized considering the correct decoding time of the stimuli, decoding accuracy and percentage of correct answers. All data underwent statistical analysis using SPSS software. The data indicated an important effect of the length of words and non-words on public and private school students. Furthermore, it was possible to observe the evolution of decoding, depending on the school year, in all the variables studied. In both groups, a strong influence of non-words on student performance throughout Elementary School I was observed. The data indicate validity in the analysis of response processes, since it was possible to adequately characterize the performance of school children public and private throughout Elementary School I, characterizing each group, as well as their differences according to the advancement of schooling.

Discursive formulas of response in Modern Greek

The article is dedicated to the study of discursive formulas of response in Modern Greek. Discursive formulas of response, i.e. idioms which are used in speech as a reaction to what was previously said by the interlocutor and which reproduce some of the phonetic properties of the previous utterance, constitute one of the most language-specific layers of phraseology and pose a difficulty for a correct decoding and translation. The aim of the article is to describe the little-studied material of Modern Greek formulas of response, to determine their types and specific properties and to identify typical communicative situations in which they are used. The study is based on the instances of use of such structures in literary works, internet texts, movies and series in Modern Greek, along with data from dictionaries. Methods of structural, semantic, contextual and comparative analysis are used. The findings of the study provide evidence that in Modern Greek there exist structures corresponding to the definition of discursive formulas of response. Eight types of such structures can be distinguished based on the type of the initiating utterance: reactions to a question; reactions to an address; reactions to an expression of agreement or disagreement; reactions to an expression of opinion or an evaluation; reactions to a factual statement; reactions to a greeting, a farewell or a well-wish; reactions to an expression of gratitude; reactions to a directive. It is shown, however, that the key role in every case is played by the actual lexical make-up of the initiating utterances. From the authors’ point of view, what is particularly interesting for the comparative analysis of discursive formulas of response in different languages is the fact that some Modern Greek formulas of response have the form of a curse or of an ill-wish.

S-Band WDM Transmission Using PPLN-Based Wavelength Converters and 400-Gb/s C-Band Real-Time Transceivers

Multi-band WDM transmission beyond the C+L-band is a promising technology for achieving larger capacity transmission by a limited number of installed fibers. In addition to the C- and L-band, we can expect use the S-band as the next band. Although the development of optical components for new bands, particularly transceivers, entails resource dispersion, which is one of the barriers to the realization of multi-band systems, wavelength conversion by transparent all-optical signal processing enables new wavelength band transmission using existing components. Therefore, we proposed a transmission system including a new wavelength band such as the S-band and made it possible to use a transceiver for the existing band by performing the whole-band wavelength conversion without using a transceiver for the new band. As a preliminary verification to demonstrate multi-band WDM transmission including S-band, we investigated the application of a novel wavelength converter between C-band and S-band, which consists of periodically poled lithium niobate waveguide, to the proposed system. We first characterized the conversion efficiency and noise figure of the wavelength converter and estimated the transmission performance of the system through the wavelength converter. Using the evaluated wavelength converters and test signals of 64 channels arranged in the C-band at 75-GHz intervals, we constructed an experimental setup for S-band transmission through an 80-km standard single-mode fiber. We then demonstrated error-free transmission of real-time 400-Gb/s DP-16QAM signals after forward error correction decoding. From the experimental results, it was clarified that thewavelength converter which realizes the uniformlossless conversion covering the whole C-band effectively achieves the S-band WDM transmission, and it was verified that the capacity improvement of the multi-band WDM system including the S-band can be expected by applying it in combination with the C+L-band WDM system.

Semi-supervised learning and bidirectional decoding for effective grammar correction in low-resource scenarios

The correction of grammatical errors in natural language processing is a crucial task as it aims to enhance the accuracy and intelligibility of written language. However, developing a grammatical error correction (GEC) framework for low-resource languages presents significant challenges due to the lack of available training data. This article proposes a novel GEC framework for low-resource languages, using Arabic as a case study. To generate more training data, we propose a semi-supervised confusion method called the equal distribution of synthetic errors (EDSE), which generates a wide range of parallel training data. Additionally, this article addresses two limitations of the classical seq2seq GEC model, which are unbalanced outputs due to the unidirectional decoder and exposure bias during inference. To overcome these limitations, we apply a knowledge distillation technique from neural machine translation. This method utilizes two decoders, a forward decoder right-to-left and a backward decoder left-to-right, and measures their agreement using Kullback-Leibler divergence as a regularization term. The experimental results on two benchmarks demonstrate that our proposed framework outperforms the Transformer baseline and two widely used bidirectional decoding techniques, namely asynchronous and synchronous bidirectional decoding. Furthermore, the proposed framework reported the highest F1 score, and generating synthetic data using the equal distribution technique for syntactic errors resulted in a significant improvement in performance. These findings demonstrate the effectiveness of the proposed framework for improving grammatical error correction for low-resource languages, particularly for the Arabic language.

Accelerating the Verification of Forward Error Correction Decoders by PCIe FPGA Cards

Pre-silicon forward error correction (FEC) decoding hardware is typically designed using hardware description languages (HDL). Its verification is a hard task due to its intrinsic tendency to correct errors. The generation and injection of millions of random inputs as well as the cross-checking of the corresponding outputs is highly recommended. Using HDL simulations for such work leads to prohibitive execution times. This letter proposes a verification strategy in which the software testbed is executed on a multi-core host and the hardware under verification is prototyped on a PCIe accelerator card. Data are transferred in big blocks of codewords over a high-bandwidth PCIe channel and applied to the decoder using a pipeline management to maximize the use of computational resources and to minimize the verification time. The decoder is replicated with parallel access to DDRs. OpenMP is used to leverage the parallel capabilities of the host and OpenCL, together with Xilinx Runtime Library (XRT), to manage the PCIe FPGA card execution. The results show an important speed-up with respect to HDL simulation and to other prototyping approaches.

Frame recovery using multiple images in rolling shutter based systems

AbstractAn algorithm for recovering transmitted static identifiers (IDs) in rolling shutter based Optical Camera Communication systems is proposed by the authors, considering a system comprised of a camera and a circular light source. The goal is to allow the correct decoding when the ID frame is only partially detected in the image. A baseline algorithm as reference for the frame recovery success rate (FRSR) and a reconstruction algorithm based on the idea of capturing multiple frame fragments and reassembling them is proposed in order to recover the transmitted ID not entirely seen on a single image. It was proven, by simulation, that the maximum distance at which the IDs recovery can be guaranteed is increased by 2.5 fold with the proposed algorithm, for 6‐bit, 8‐bit and 10‐bit codewords. An experimental validation algorithm was also proposed, using image processing techniques to extract the bitstreams and test the ID recovery process. The proposed algorithm improves the FRSR for a given distance, even in the presence of considerable bit errors in the bitstreams extracted from the images.

Abnormal gait partitioning and real-time recognition of gait phases in children with cerebral palsy

ObjectiveSince the gait of children with cerebral palsy (CP) lacks distinct foot event characteristics, it’s difficult to accurately distinguish the key gait events, especially in the stance phase. To solve the problem of gait phase detection for children with CP, a unique gait classification system is designed to detect the walking maneuvers during a gait cycle. This system can detect all gait phases of CP patients, and send the kinematic information to a rule-based algorithm through a series of sensors installed in the legs to identify the salient features of the signal. MethodsIn the light of the quasi-stiffness gait phase transformation as the classification guide, two sets of motion signal feature sets are extracted from the leg feature signals of 10 CP children, and two gait stage recognition algorithms are proposed to divide the six gait sub-phases. ResultsThe proposed algorithms of peak-zero-based angular velocity and thresholding-based angular realize the effective recognition of gait phases, especially for the stance phase. ConclusionThe proposed methods have the potential to obtain correct decoding of motion information and recognize gait phases. Significance: This study provides a new idea to solve the problem of difficult identification of gait phase for children with CP, which can be applied to a behavior-driven lower limb exoskeleton robot system for assisting walking and gait correction.

A Novel ML-Based Symbol Detection Pipeline for Molecular Communication

Molecular Communication (MC) is the process of sending information by the use of particles instead of electromagnetic (EM) waves. This change in paradigm allows the use of MC in areas where EM transmission is undesirable. These include underground, underwater and even intra-body communications. While this novel paradigm promises new areas for communication, one of the major setbacks is its relatively low throughput caused by the propagation speed. This can be improved by decreasing the symbol duration; however, this can be a detriment to the correct decoding of symbols. This paper proposes a novel symbol detection pipeline to increase the possible throughput without increasing the error rate of the communication. This is based on a machine-learning algorithm for classification tasks using an L-point discrete time moving average filter and a wide range of features. Extensive simulations with long sequences at different signal-to-noise ratio (SNR) values were performed to determine how well the proposed method detects symbols. The results show that our method can detect symbols received when On-Off Keying (OOK) modulations are used with a 10 dB gain, even when transmissions with untrained SNR values occur.

Adaptive Repetition Strategies in IEEE 802.11bd V2X Networks

Vehicle-to-everything (V2X) communications have very strict throughput and latency requirements, even in scenarios with high mobility. IEEE 802.11bd is being developed as a WiFi amendment to improve V2X performance, allowing up to three repetitions per packet, along with other features. Message repetitions increase time diversity and enable maximum ratio combining at the receiver, thus improving the probability of correct decoding. This paper investigates the IEEE 802.11bd packet repetition feature. First, we analyze how the increased channel load due to repetitions may in some cases result in a higher collision rate leading to lower network performance. Then, we propose two strategies for exploiting the IEEE 802.11bd repetition feature. The proposed strategies use the channel busy ratio to adapt the number of transmissions to the channel load and are validated through network-level simulations, which account for both the acquisition and decoding processes. Results show that the proposed strategies improve network performance under variable traffic conditions and enable fair access to the channel.

EEG Sinyallerinden Çıkarılan Poincare Ölçümlerini Kullanarak Çok Sınıflı Motor Hayali Görevlerin Sınıflandırılması

Motor Imaginary (MI) electroencephalography (EEG) signals are generated with the recording of brain activities when a participant imagines a movement without physically performing it. The correct decoding of MI signals have been became an important task due to the application of these signals in the rehabilitation process of paralyzed patients in recent studies. However, the decoding of the these signals is still an evolving challenge in the design of a brain-computer interface (BCI) system. In this study, a machine learning based approach using Poincare measurements from non-linear measurements of MI EEG signals is proposed for classification of four-class MI tasks. The m-lagged Poincare plots were used to extract non-linear features and m is set to be values from 1 to 10. The performances of feature vectors which are extracted from 10 lag values and feature vector which is the combinations of these vectors were investigated separately in experimental evaluation section. The 24 different typical classification algorithms were tested in differentiating MI tasks using 5-fold cross-validation. Each of the these algorithms tested 10 times to analyzed the repeatability of the experimental results. The highest classifier performance of 47.08% among these 11 feature vectors was achieved over the combination feature vector that includes all lag values features using Quadratic Support Vector Machine (SVM). According to average accuracy value of 24 classifiers in 11 feature vector, the most discriminative feature set is 9th vector that consists of features extracted when lag value defined as 9. As a result, the innovative aspect of this study is the application of Poincare plots, one of the nonlinear feature extraction methods, in motor imaginary task classification.

Deep Learning Based Successive Interference Cancellation for the Non-Orthogonal Downlink

Non-orthogonal communications are expected to play a key role in future wireless systems. In downlink transmissions, the data symbols are broadcast from a base station to different users, which are superimposed with different power to facilitate high-integrity detection using successive interference cancellation (SIC). However, SIC requires accurate knowledge of both the channel model and channel state information (CSI), which may be difficult to acquire. We propose a deep learning-aided SIC detector termed SICNet, which replaces the interference cancellation blocks of SIC by deep neural networks (DNNs). Explicitly, SICNet jointly trains its internal DNN-aided blocks for inferring the soft information representing the interfering symbols in a data-driven fashion, rather than using hard-decision decoders as in classical SIC. As a result, SICNet reliably detects the superimposed symbols in the downlink of non-orthogonal systems without requiring any prior knowledge of the channel model, while being less sensitive to CSI uncertainty than its model-based counterpart. SICNet is also robust to changes in the number of users and to their power allocation. Furthermore, SICNet learns to produce accurate soft outputs, which facilitates improved soft-input error correction decoding compared to model-based SIC. Finally, we propose an online training method for SICNet under block fading, which exploits the channel decoding for accurately recovering online data labels for retraining, hence, allowing it to smoothly track the fading envelope without requiring dedicated pilots. Our numerical results show that SICNet approaches the performance of classical SIC under perfect CSI, while outperforming it under realistic CSI uncertainty.

Hiding Message Using a Cycle Generative Adversarial Network

Training an image steganography is an unsupervised problem, because it is impossible to obtain an ideal supervised steganographic image corresponding to the cover image and secret message. Inspired by the success of cycle generative adversarial networks in unsupervised tasks such as style transfer, this article proposes to use a cycle generative adversarial network to solve the problem of unsupervised image steganography. Specifically, this article jointly trains five networks, i.e., a steganographic network, an inverse steganographic network, a hidden message reconstruction network, and two discriminative networks, which together constitute a hidden message cycle generative adversarial network (HCGAN). Compared with the recent image steganography based on generative adversative network, HCGAN provides more accurate supervised information, which makes the training process of HCGAN converge faster and the performance of the trained image steganography network is better. In addition, this article introduces an image steganographic network based on residual learning and shows that residual learning can effectively improve the performance of steganography. Furthermore, to the best of our knowledge, we are the first to propose an inverse steganographic network for eliminating steganographic message from steganographic images, which can be used to avoid steganographic message being discovered or acquired by a third party. The experimental results show that compared with the steganography based on generative adversarial network, the proposed HCGAN has a higher correct decoding rate, better visual quality of steganographic image, and higher secrecy.

Efficient Computation of Viterbi Decoder Reliability With an Application to Variable-Length Coding

This paper compares the accuracy and complexity of Raghavan and Baum’s Reliability Output Viterbi Algorithm (ROVA), Polyanskiy’s accumulated information density (AID), and Fricke and Hoeher’s lower complexity approximation of ROVA. It turns out that AID is far less accurate than ROVA in practice. This paper proposes codeword information density (CID), which modifies AID to improve its accuracy and leads to a lower-complexity implementation of ROVA. The paper includes an analytical expression for the random variable describing the correct decoding probability computed by ROVA and uses this expression to characterize how the probabilities of correct decoding, undetected error, and negative acknowledgement behave as a function of the selected threshold for reliable decoding. This paper examines both the complexity and the simulation time of ROVA, CID, AID, and the Fricke and Hoeher approximation to ROVA. This paper also derives an expression for the union bound on the frame error rate for zero-terminated trellis codes with punctured symbols and uses it to optimize the order of symbol transmission in an incremental retransmission scheme. This paper concludes by comparing the performance of an incremental retransmission scheme using ROVA as a stopping condition to one that uses a CRC as a stopping condition.

Improved encoding structure and decoding algorithms for spinal codes

To improve the error correction performance, an innovative encoding structure with tail-biting for spinal codes is designed. Furthermore, an adaptive forward stack decoding (A-FSD) algorithm with lower complexity for spinal codes is proposed. In the A-FSD algorithm, a flexible threshold parameter is set by a variable channel state to narrow the scale of nodes accessed. On this basis, a new decoding method of AFSD with early termination (AFSD-ET) is further proposed. The AFSD-ET decoder not only has the ability of dynamically modifying the number of stored nodes, but also adopts the early termination criterion to curtail complexity. The complexity and related parameters are verified through a series of simulations. The simulation results show that the proposed spinal codes with tail-biting and the AFSD-ET decoding algorithms can reduce the complexity and improve the decoding rate without sacrificing correct decoding performance.

Improving Decodability of Polar Codes by Adding Noise

This paper presents an online perturbed and directed neural-evolutionary (Online-PDNE) decoding algorithm for polar codes, in which the perturbation noise and online directed neuro-evolutionary noise sequences are sequentially added to the received sequence for re-decoding if the standard polar decoding fails. The new decoding algorithm converts uncorrectable received sequences into error-correcting regions of their decoding space for correct decoding by adding specific noises. To reduce the decoding complexity and delay, the PDNE decoding algorithm and sole neural-evolutionary (SNE) decoding algorithm for polar codes are further proposed, which provide a considerable tradeoff between the decoding performance and complexity by acquiring the neural-evolutionary noise in an offline manner. Numerical results suggest that our proposed decoding algorithms outperform the other conventional decoding algorithms. At high signal-to-noise ratio (SNR) region, the Online-PDNE decoding algorithm improves bit error rate (BER) performance by more than four orders of magnitude compared with the conventional simplified successive cancellation (SSC) decoding algorithm. Furthermore, in the mid-high SNR region, the average normalized complexity of the proposed algorithm is almost the same as that of the SSC decoding algorithm, while preserving the decoding performance gain.