Adaptive Quantization Research Articles

Quantized-State-Feedback-Based Neural Control for a Class of Switched Nonlinear Systems With Unknown Control Directions

This paper investigates the problem of unknown virtual control directions in a state-quantized adaptive recursive control design for a class of arbitrarily switched uncertain pure-feedback nonlinear systems in a band-limited network. State quantization is considered for state feedback control in a band-limited network. The primary contribution of this study is to provide a quantized state feedback adaptive control strategy to address the unknown control direction and arbitrarily switched nonaffine nonlinearities. Herein, a coupling problem between Nussbaum functions and quantization errors caused by quantized state feedback control laws is considered in the Lyapunov-based design and stability analysis. A state-quantized adaptive recursive control scheme using the function approximation is constructed without a priori knowledge of the signs of the control gain functions, where the estimated parameters and Nussbaum-type functions are adaptively updated via quantized states. Theoretical lemmas are derived to show that the adaptive parameters and quantization errors of the closed-loop signals are bounded using the proposed control scheme. The boundedness of the closed-loop signals and the convergence of tracking error to a neighborhood of the origin are proved using the common Lyapunov function approach. Two simulation examples are shown to illustrate the effectiveness of the proposed theoretical result.

Physical-Layer Secret Key Generation in Power Line Communication Networks

Due to the inherent broadcast nature, power line communication (PLC) is vulnerable to eavesdropping. Traditionally, the secret keys are generated/distributed by pre-shared keys, adaptive tone mapping, or asymmetric encryption algorithm. All of these methods face problems such as high computational complexity, high cost, or less security. To solve the problem, a physical layer security (PLS)-based secret key generation (SKG) scheme aiming at the typical in-home broadband PLC network is proposed in this paper. However, the slowly time-varying characteristic and the serious noise interference in PLC channels will degrade the performance of the PLS-based SKG scheme. To break through the limitation of the randomness of PLC channels, two-way random signals are introduced. To make the SKG system more resistant to noise, the channel frequency response-based adaptive quantization (CFR-AQ) algorithm is designed, in which the amplitude threshold and the quantization level are adaptive to the noise power and the target key disagreement rate (KDR) to maximize the key generation rate (KGR). The performance is analyzed in two scenarios: single eavesdropper and multiple eavesdroppers. The simulation results demonstrate that the proposed scheme has a good overall performance in terms of KDR, KGR, and security with a lower complexity in the presence of single or multiple non-collusive eavesdroppers, and it can be regarded as an alternative solution to generate secret keys in PLC networks.

Bipartite Containment Control for Multiagent Systems with Adaptive Quantization Information

In this study, bipartite containment control for multiagent systems (MASs) with quantitative information is investigated. Communication topology is structurally balanced, and the follower’s trajectory is within the area surrounded by the leader by designing distributed error terms. The leader is an external system, and the matrix information of the leader cannot be accessed by followers. Based on an adaptive quantization information distributed observer, the authors introduce an output feedback protocol to study bipartite containment control. Finally, simulations demonstrate the effectiveness of proposed control algorithms.

Grayscale-inversion and rotation invariant image description using local ternary derivative pattern with dominant structure encoding

Local ternary pattern (LTP) is sensitive to inverse grayscale changes and its split encoding limits its descriptive ability. In this paper, we investigate these problems and present an enhanced LTP descriptor called local ternary derivative pattern with dominant structure encoding (DLTDP). Firstly, we propose a local ternary derivative pattern (LTDP) operator which encodes the neighboring relationship in the magnitude response space of Gaussian derivative filters. In LTDP, adaptive ternary quantization is performed on the magnitude responses of Gaussian derivative filters to achieve grayscale-inversion and rotation invariance. An extended non-split ternary encoding scheme is developed to obtain compact yet discriminative LTDP codes. Secondly, to complement the magnitude responses used in LTDP, we leverage a dominant structure measure to additionally encode each pixel in the original filter response space as well as the input image space. Finally, we integrate all the generated codes to construct joint histogram features as the DLTDP descriptor. Extensive experiments on four benchmark databases (i.e., Outex, CUReT, KTH-TIPS and DTD) demonstrate the superiority of our DLTDP descriptor over state-of-the-art LBP and LTP variants for texture classification under (linear or nonlinear) grayscale inversion and image rotation.

Video coding optimization in AVS2

Chinese second generation of the Audio Video Coding Standard, known as the AVS2, competing with HEVC/H.265 and AV1, has become a well-known video compression standard. Many unique tools have been developed and incorporated in AVS2. This paper proposes several encoder optimization schemes with respect to rate-distortion optimization (RDO). Specifically, by exploiting the close relationship between the total number of referenced pictures and the hierarchy of reference structure, a reference structure determined parameter (RSDP) algorithm reallocates Lagrange multiplier. Moreover, an adaptive quantization parameter offset (AQPO) algorithm is provided at both picture-level and Groups of Pictures (GoP)-level to adjust quantization parameters adaptively on top of the preassigned ones. Both RSDP algorithm and AQPO algorithm have already been adopted by AVS2 and incorporated in its reference software. Experimental results show that in the low-delay, RSDP and AQPO achieved savings of 1.35% and 1.26% on BDBR, respectively.

Study of compression statistics and prediction of rate-distortion curves for video texture

Encoding textural content remains a challenge for current standardised video codecs. It is therefore beneficial to understand video textures in terms of both their spatio-temporal characteristics and their encoding statistics in order to optimise encoding performance. In this paper, we analyse the spatio-temporal features and statistics of video textures, explore the rate-quality performance of different texture types and investigate models to mathematically describe them. For all considered theoretical models, we employ machine-learning regression to predict the rate-quality curves based solely on selected spatio-temporal features extracted from uncompressed content. All experiments were performed on homogeneous video textures to ensure validity of the observations. The results of the regression indicate that using an exponential model we can more accurately predict the expected rate-quality curve (with a mean Bjøntegaard Delta rate of .46% over the considered dataset), while maintaining a low relative complexity. This is expected to be adopted by in the loop processes for faster encoding decisions such as rate–distortion optimisation, adaptive quantisation, partitioning, etc.

Information-based distributed extended Kalman filter with dynamic quantization via communication channels

This paper proposes a new information-based distributed extended Kalman filter algorithm under dynamic quantization. Our quantization framework has the advantage of utilizing online updated quantizer’s parameters. A practical adjustment strategy is derived to ensure the availability of adaptive quantizer’s parameters for the encoder and decoder. It is proved that estimation error of the proposed algorithm is exponentially bounded in mean square under some assumptions. A numerical example concerning target tracking is presented to demonstrate the validity of the main results, in which a complex network model is used to simulate the sensor network.

Efficient Multi-modal Hashing with Online Query Adaption for Multimedia Retrieval

Multi-modal hashing supports efficient multimedia retrieval well. However, existing methods still suffer from two problems: (1) Fixed multi-modal fusion. They collaborate the multi-modal features with fixed weights for hash learning, which cannot adaptively capture the variations of online streaming multimedia contents. (2) Binary optimization challenge. To generate binary hash codes, existing methods adopt either two-step relaxed optimization that causes significant quantization errors or direct discrete optimization that consumes considerable computation and storage cost. To address these problems, we first propose a Supervised Multi-modal Hashing with Online Query-adaption method. A self-weighted fusion strategy is designed to adaptively preserve the multi-modal features into hash codes by exploiting their complementarity. Besides, the hash codes are efficiently learned with the supervision of pair-wise semantic labels to enhance their discriminative capability while avoiding the challenging symmetric similarity matrix factorization. Further, we propose an efficient Unsupervised Multi-modal Hashing with Online Query-adaption method with an adaptive multi-modal quantization strategy. The hash codes are directly learned without the reliance on the specific objective formulations. Finally, in both methods, we design a parameter-free online hashing module to adaptively capture query variations at the online retrieval stage. Experiments validate the superiority of our proposed methods.

On a Strategy of Mobile Networks Synthesis

Effective design of mobile communication network includes optimization of two bounded together processes: the network base stations placement and the channel assignment. In real environments the well-known cellular concept fails due to not uniformly spaced traffic and not isotropic wave propagation. We find a rather universal method for synthesis of a close to optimal network structure. The proposed design approach is based on the idea of adaptive vector quantization for a map with random traffic. As a result, the service zone of the network becomes discrete and is being transformed to a map with not equal cells and approximately equal number of requests in every one. This fact allows to apply the simplest periodic reuse pattern. The algorithm finds a point with minimal average Euclidean distance from all its requests. This point defines optimal placement of the corresponding base station. The approach guarantees maximum coverage

Research on Remote Sensing Image Matching with Special Texture Background

The purpose of image registration is to find the symmetry between the reference image and the image to be registered. In order to improve the registration effect of unmanned aerial vehicle (UAV) remote sensing imagery with a special texture background, this paper proposes an improved scale-invariant feature transform (SIFT) algorithm by combining image color and exposure information based on adaptive quantization strategy (AQCE-SIFT). By using the color and exposure information of the image, this method can enhance the contrast between the textures of the image with a special texture background, which allows easier feature extraction. The algorithm descriptor was constructed through an adaptive quantization strategy, so that remote sensing images with large geometric distortion or affine changes have a higher correct matching rate during registration. The experimental results showed that the AQCE-SIFT algorithm proposed in this paper was more reasonable in the distribution of the extracted feature points compared with the traditional SIFT algorithm. In the case of 0 degree, 30 degree, and 60 degree image geometric distortion, when the remote sensing image had a texture scarcity region, the number of matching points increased by 21.3%, 45.5%, and 28.6%, respectively and the correct matching rate increased by 0%, 6.0%, and 52.4%, respectively. When the remote sensing image had a large number of similar repetitive regions of texture, the number of matching points increased by 30.4%, 30.9%, and −11.1%, respectively and the correct matching rate increased by 1.2%, 0.8%, and 20.8% respectively. When processing remote sensing images with special texture backgrounds, the AQCE-SIFT algorithm also has more advantages than the existing common algorithms such as color SIFT (CSIFT), gradient location and orientation histogram (GLOH), and speeded-up robust features (SURF) in searching for the symmetry of features between images.

Adaptive Video Watermarking against Scaling Attacks Based on Quantization Index Modulation

Video watermarking plays a vital role in protecting the video copyright. The quantization-based methods are widely used in the existing watermarking algorithms, owing to their low computational complexity and completely blind extraction. However, most of them work poorly in resisting scaling attacks, by which the quantization value may fall outside the original quantization interval. For addressing this issue, an adaptive quantization index modulation method is proposed. The property that is associated with the ratio of the DC coefficient before and after scaling the video resolution motivates us to select the DC coefficient as the quantization value and set the size of the quantization interval by the video resolution to maintain the synchronization between them before and after scaling. Moreover, a strategy taking advantage of the high decoding reliability of the QRCode is proposed to terminate the extraction in advance, and both the embedding and the extracting process are performed in the spatial domain, which all contribute to further enhance the execution efficiency. The experimental results show that our algorithm outperforms the state-of-the-art method in terms of imperceptibility, robustness, and computational cost.

SpaceDML: Enabling Distributed Machine Learning in Space Information Networks

Space information networks (SINs) have become a rapidly growing global infrastructure service. Massive volumes of high-resolution images and videos captured by low-orbit satellites and unmanned aerial vehicles have provided a rich training data source for machine learning applications. However, SIN devices' limited communication and computation resources make it challenging to perform machine learning efficiently with a swarm of SIN devices. In this article, we propose Spacedml, a distributed machine learning system for SIN platforms that applies dynamic model compression techniques to adapt distributed machine learning training to SINs' limited bandwidth and unstable connectivity. Spaced-ml has two key algorithm:s adaptive loss-aware quantization, which compresses models without sacrificing their quality, and partial weight averaging, which selectively averages active clients' partial model updates. These algorithms jointly improve communication efficiency and enhance the scalability of distributed machine learning with SIN devices. We evaluate Spacedml by training a LeNet-S model on the MNIST dataset. The experimental results show that Spacedml can increase model accuracy by 2-3 percent and reduce communication bandwidth consumption by up to 60 percent compared to the baseline algorithm.

Реализация метода адаптивного квантования в измерительных генераторах с цифровым управлением

Measuring generators with digital control, in particular power calibrators, used to calibrate electricity meters, contain a digital-to-analog converter (DAC) that converts codes of the generated signal into voltage. Signal codes are stored in the generator memory. A truncation discreteness error (quantization noise) arises caused by sampling (quantization) in time and by the level of signal samples in the DAC. A relative value of the quantization noise depends on the amplitude of the generated signal (relative to the reference voltage of the DAC): the larger the amplitude, the more significant bits of the DAC are involved in the conversion process, and the less the relative value of the noise. In generators, where the amplitude of the output signal changes over a wide range (high dynamic range) by changing the digital samples of the signal, the quantization noise at low signal amplitudes can become unacceptably large. This situation occurs in power calibrators where the output current changes hundreds of times since the error of the verified electricity meter is normalized in a wide range of current flowing through it. A new algorithm for generating samples of a sinusoidal signal in measuring generators with digital control called adaptive quantization is proposed. Adaptive quantization can significantly improve one of the selected signal parameters (the so-called optimality criterion), for example, reduce the error in reproduction of the first harmonic, or reduce the value of higher harmonic components. In addition, the proposed algorithm reduces the dependence of the selected parameter on the sampling frequency and on the number of DAC bits used, which makes it possible to expand the dynamic range of the generator (in the current channel) without using additional amplifiers with programmable gain (PGA). Studies carried out using computer simulation have confirmed the efficiency of the adaptive quantization algorithm.

Adaptive quantisation synchronisation control for bilateral teleoperation systems with communication delays

In this paper, the quantisation synchronisation control problem of bilateral teleoperation systems with communication delays is investigated. The inaccurate modelling problem including the system uncertainties and the unknown disturbances are compensated by using the radial basis function networks (RBFN). By introducing the hysteretic quantiser, the control signals are quantised to a set of discrete values to ease the accuracy requirement for actual physical systems and the requirement for communication rate. The position synchronisation of joint variables between the master and slave manipulators is achieved by using the designed control signals and adaptive parameters. Simulation results verify that the synchronisation tracking errors gradually converge to a small neighbourhood around zero.

Distribution Adaptive INT8 Quantization for Training CNNs

Researches have demonstrated that low bit-width (e.g., INT8) quantization can be employed to accelerate the inference process. It makes the gradient quantization very promising since the backward propagation requires approximately twice more computation than forward one. Due to the variability and uncertainty of gradient distribution, a lot of methods have been proposed to attain training stability. However, most of them ignore the channel-wise gradient distributions and the impact of gradients with different magnitudes, resulting in the degradation of final accuracy. In this paper, we propose a novel INT8 quantization training framework for convolutional neural network to address the above issues. Specifically, we adopt Gradient Vectorized Quantization to quantize the gradient, based on the observation that layer-wise gradients contain multiple distributions along the channel dimension. Then, Magnitude-aware Clipping Strategy is introduced by taking the magnitudes of gradients into consideration when minimizing the quantization error, and we present a theoretical derivation to solve the quantization parameters of different distributions. Experimental results on broad range of computer vision tasks, such as image classification, object detection and video classification, demonstrate that the proposed Distribution Adaptive INT8 Quantization training method has achieved almost lossless training accuracy for different backbones, including ResNet, MobileNetV2, InceptionV3, VGG and AlexNet, which is superior to the state-of-the-art techniques. Moreover, we further implement the INT8 kernel that can accelerate the training iteration more than 200% under the latest Turing architecture, i.e., our method excels on both training accuracy and speed.

Algorithm based on 2‐bit adaptive delta modulation and fractional linear prediction for Gaussian source coding

A novel 2-bit adaptive delta modulation (ADM) algorithm is presented based on uniform scalar quantization and fractional linear prediction (FLP) for encoding the signals modelled by a Gaussian probability density function. The study focusses on two major areas: realization of a 2-bit adaptive quantizer based on Q-function approximation that significantly facilitates quantizer design; and implementation of a recently introduced FLP approach with the memory of two samples, which replaces the first-order linear prediction used in standard ADM algorithms and enables improved performance without increasing transmission costs. It furthermore represents the first implementation of FLP in signal encoding, therefore confirming its applicability in a real signal-processing scenario. Based on the performance analysis conducted on a real speech signal, the proposed ADM algorithm with FLP is demonstrated to outperform other 2-bit ADM baselines by a large margin for the gain in signal-to-noise ratio achieved over a wide dynamic range of input signals. The results of this research indicate that ADM with adaptive quantization based on Q-function approximation and adaptive FLP represents a promising solution for encoding/compression of correlated time-varying signals following the Gaussian distribution.

On visual masking estimation for adaptive quantization using steerable filters

A fast and accurate assessment of visual masking effects is desirable while encoding in order to utilize such effects to improve the quality of compressed videos through an adaptive quantization (AQ) scheme. Here, we propose a method of estimating the contrast masking threshold on natural scene patches, using texture cues imparted by steerable filter responses. We then employ the estimated thresholds to perform AQ for AV1 encoding. Our experimental results establish that the proposed method is able to outperform existing visual masking models in terms of estimation performance while being relatively computationally inexpensive than these models, and is also able to improve the variance based AQ algorithm that is currently deployed in the SVT-AV1 codec. Using the multi-scale structural similarity index measure (MS-SSIM) as the quality model, our approach achieves an average BD-rate of -1.82% using the uniform quantization scheme as anchor as compared to 5.83% obtained with the variance based method. We note that the proposed approach produces less visible compression artifacts than the variance based AQ approach at lower bitrates, while maintaining similar encoding complexity.

Bilateral coordination quantisation control for master-slave flexible manipulators based on PDE dynamic model

This paper investigates a tracking adaptive control issue of the master-slave flexible system with input signal quantisation. Considering the distributed parameter characteristics of the flexible manipulator, the master-slave teleoperation system is a dynamic model described by partial differential equations. In this work, an adaptive boundary bilateral coordination quantisation control scheme is developed such that input signal quantisation can be achieved by applying random quantiser. Based on such control strategy, all signals of the closed-loop system are bounded and Lyapunov method is used to prove uniform ultimate boundedness of the control system. The proposed control law can not only effectively eliminate the deflection of the flexible link but also realise angle coordination. The numerical simulation results are provided for illustrating the control effectiveness.

Adaptive Sparse Quantization Kernel Least Mean Square Algorithm for Online Prediction of Chaotic Time Series

Kernel least mean square (KLMS) algorithm is a popular method for time series online prediction. It has the advantages of good robustness, low computational complexity, model simplicity and online learning ability. Unfortunately, as input data grows, the dictionary size increases and the computational complexity raises significantly. In addition, how to improve the adaptability in time-varying environments with noise is also one of the main challenges. Therefore, we propose an improved KLMS algorithm from sparse perspective in response to the above problems, called adaptive sparse quantization kernel least mean square (ASQ-KLMS) algorithm. In the new model, sequential outlier criterion for sparsification and weights adaptive adjustment are combined with coherence criterion and quantization to form ASQ-KLMS algorithm. Firstly, it makes full use of effective information and ignores the interference of abnormal information to obtain a more accurate and compact dictionary. Then, a good balance between algorithm efficiency and accuracy can be achieved by controlling the choice of parameters. In addition, it can adaptively adjust weights in time-varying environment. At last, the Lorenz chaotic time series, the ENSO chaotic time series and the Beijing $$\text {P}{{\text {M}}_{2.5}}$$ chaotic time series are used to prove the reliability of the ASQ-KLMS algorithm.

Efficient Post-Processing for Physical-Layer Secure Key Distribution in Fiber

Post-processing is an essential step towards the practical deployment of physical-layer secure key distribution (SKD) in classical fiber channels. However, the comparison and combination of different post-processing methods are not explored yet. Here, we fully evaluate the post-processing protocol and analyze various combinations of quantization and information reconciliation (IR), and present the efficient solutions associated with the key entropy source between the legal parties. In addition, the SHA3-256 hash function-based privacy amplification is exploited to further enhance the robust key secrecy and randomness. The results show that, the proposed protocol significantly increases the achievable key generation rate (KGR) up to 130-205% using the combination of multi-bit adaptive quantizer and IR with syndrome-decoding LDPC, compared with the previous post-processing using the same original key waveforms extracted from physical-layer SKD.