Compression Techniques Research Articles

Multisensor‐Driven Lightweight Networks for Intelligent Sports Training System: Design and Research

ABSTRACTEdge computing and multisensor integration have revolutionized sports training technologies, yet many existing systems remain computationally intensive and complex for everyday use. This study introduces an efficient and user‐friendly intelligent sports training system that leverages edge‐based multisensor data processing with lightweight neural networks. The system utilizes low‐latency activity and real‐time pose estimation data, powered by a modified MobileNetV2 model, to generate personalized training plans through reinforcement learning. Tests show that the system matches the accuracy of more complex models while significantly reducing computational needs. Compression techniques further enhance efficiency with minimal accuracy loss. User studies revealed notable improvements in fitness levels and adherence compared to traditional methods. The system adapts to real‐time user performance, offering feedback and dynamically adjusting plans, with low energy consumption across mobile devices. This research shows that our developed system, which integrates multisensors and artificial intelligence, can make personalized sports training more accessible.

Opportunities and Challenges in Deep Compressed Sensing Techniques for Multichannel ECG Data Compression

Opportunities and Challenges in Deep Compressed Sensing Techniques for Multichannel ECG Data Compression

Optimization and Performance Analysis of Real-time Speech Translation Systems Based on Mobile Technology

As globalization deepens and mobile technology rapidly advances, the demand for crosslinguistic communication has been steadily increasing, making real-time speech translation systems a research focus. However, given the limited computational capacity and storage space of mobile devices, optimizing system performance while maintaining translation quality has become a critical challenge. Current optimization approaches for real-time speech translation systems primarily focus on improvements to model architectures and hardware acceleration, often neglecting a systematic study of model compression. This is particularly evident when handling real-time data, where achieving both high efficiency and translation accuracy remains difficult. To address these challenges, a model compression method based on the connectionist temporal classification (CTC) criterion was proposed, along with an in-depth study of parameter compression tailored for mobile applications. The research focuses on two key areas: first, model compression techniques based on the CTC criterion were explored to enhance the efficiency of real-time speech translation; second, parameter compression methods were investigated to significantly reduce resource consumption in mobile applications while preserving translation quality. The aim of this study is to improve the performance and user experience of real-time speech translation systems on mobile devices.

Topologically-protected four-wave mixing enhanced by tailoring topological edge states

Laser pulse compression serves as a pivotal technique in nonlinear optics and light-matter interactions. Traditional methods, including Q-switching, mode-locking techniques, and chirped pulse amplification, are employed to generate high-intensity fields. However, these methods pose challenges when applied to topological lasers. In this study, we explore the pulse compression technique for topological edge states (TESs). Our findings reveal a significant correlation between the pulse width of TESs and their group velocity. Though factors such as nonlinear TES dispersion, disorders, and edge bending also affect pulse width, their impacts are comparatively less pronounced compared to that of group velocity. Through the customization of TES’s group velocity, we demonstrate a remarkable enhancement of four-wave mixing, showcasing it as a prime example of a nonlinear processing technique. Importantly, our approach promises seamless integration into topological laser systems, ensuring optimal performance with zero energy transfer loss and introducing a flexible dimension and robustness to topological nonlinear optical devices.

A Survey on Deep Neural Network Pruning: Taxonomy, Comparison, Analysis, and Recommendations.

Modern deep neural networks, particularly recent large language models, come with massive model sizes that require significant computational and storage resources. To enable the deployment of modern models on resource-constrained environments and to accelerate inference time, researchers have increasingly explored pruning techniques as a popular research direction in neural network compression. More than three thousand pruning papers have been published from 2020 to 2024. However, there is a dearth of up-to-date comprehensive review papers on pruning. To address this issue, in this survey, we provide a comprehensive review of existing research works on deep neural network pruning in a taxonomy of 1) universal/specific speedup, 2) when to prune, 3) how to prune, and 4) fusion of pruning and other compression techniques. We then provide a thorough comparative analysis of eight pairs of contrast settings for pruning (e.g., unstructured/structured, one-shot/iterative, data-free/data-driven, initialized/pre-trained weights, etc.) and explore several emerging topics, including pruning for large language models, vision transformers, diffusion models, and large multimodal models, post-training pruning, and different levels of supervision for pruning to shed light on the commonalities and differences of existing methods and lay the foundation for further method development. Finally, we provide some valuable recommendations on selecting pruning methods and prospect several promising research directions for neural network pruning. To facilitate future research on deep neural network pruning, we summarize broad pruning applications (e.g., adversarial robustness, natural language understanding, etc.) and build a curated collection of datasets, networks, and evaluations on different applications. We maintain a repository on https://github.com/hrcheng1066/awesome-pruning that serves as a comprehensive resource for neural network pruning papers and corresponding open-source codes. We will keep updating this repository to include the latest advancements in the field.

New impulse-based test method for early-age elastic modulus measurement in cementitious materials

Accurate monitoring of the elastic modulus evolution during the early-age fluid-solid transition in cementitious materials is crucial for understanding their mechanical behavior under early stresses. While existing methods, such as EMM-ARM, have proven effective, they are often associated with high costs or complex setups. This study presents the development of a low-cost, easy-to-implement system that employs impulse excitation applied to a composite beam, introducing the EMM-IRM method for the effective assessment of the elastic modulus of cement paste from the earliest stages of hydration. The use of impulse excitation is particularly advantageous in low-cost systems with accelerometers of lower sensitivity, as it enhances signal clarity and reduces interference from background noise, thereby improving measurement reliability. The system was validated through experimental comparisons with conventional EMM-ARM, cyclic compression, and traditional impulse excitation techniques, using Portland cement paste as the test material. The results show that EMM-IRM consistently estimates the elastic modulus, with a maximum deviation of 6.33% from other methods. Moreover, despite its cost-effectiveness, EMM-IRM demonstrates a higher signal-to-noise ratio compared to conventional EMM-ARM. This approach offers a practical solution for early-age characterization of cementitious materials, balancing cost, simplicity, and measurement accuracy.

Towards Lightweight Super-Resolution With Dual Regression Learning.

Deep neural networks have exhibited remarkable performance in image super-resolution (SR) tasks by learning a mapping from low-resolution (LR) images to high-resolution (HR) images. However, the SR problem is typically an ill-posed problem and existing methods would come with several limitations. First, the possible mapping space of SR can be extremely large since there may exist many different HR images that can be super-resolved from the same LR image. As a result, it is hard to directly learn a promising SR mapping from such a large space. Second, it is often inevitable to develop very large models with extremely high computational cost to yield promising SR performance. In practice, one can use model compression techniques to obtain compact models by reducing model redundancy. Nevertheless, it is hard for existing model compression methods to accurately identify the redundant components due to the extremely large SR mapping space. To alleviate the first challenge, we propose a dual regression learning scheme to reduce the space of possible SR mappings. Specifically, in addition to the mapping from LR to HR images, we learn an additional dual regression mapping to estimate the downsampling kernel and reconstruct LR images. In this way, the dual mapping acts as a constraint to reduce the space of possible mappings. To address the second challenge, we propose a dual regression compression (DRC) method to reduce model redundancy in both layer-level and channel-level based on channel pruning. Specifically, we first develop a channel number search method that minimizes the dual regression loss to determine the redundancy of each layer. Given the searched channel numbers, we further exploit the dual regression manner to evaluate the importance of channels and prune the redundant ones. Extensive experiments show the effectiveness of our method in obtaining accurate and efficient SR models.

Design and Implementation of a High-Performance VLSI Architecture for Canonical Huffman Encoding

A key component of contemporary computer systems, data compression makes it possible for information to be stored and transmitted efficiently. A popular technique for lossless data compression, Huffman coding can drastically cut down on data size without sacrificing intensity. But conventional Huffman encoding techniques can have scalability and speed issues, especially when used in hardware. A high-throughput Very Large-Scale Integration (VLSI) design for a Canonical Huffman Encoder is shown in this study. To accomplish quick and effective encoding, the suggested approach makes use of parallel processing and improved algorithms. The system's use of the canonical Huffman algorithm lowers computational complexity and streamlines hardware implementation without sacrificing compression performance. The architecture incorporates real-time operation-optimized modules for frequency analysis, code generation, and symbol encoding. Results from simulations show that the suggested design achieves significantly higher throughput compared to conventional approaches, making it suitable for applications in data storage, multimedia, and communication systems. This study contributes to advancing VLSI designs for high-performance hardware Key Words: Data Compression, Huffman Coding, Lossless Compression, Very Large-Scale Integration (VLSI), Canonical Huffman Encoder, High-Throughput Encoding, Parallel Processing, Efficient Encoding, Hardware Implementation, Computational Complexity, Code Generation, Symbol Encoding, Frequency Analysis, Real-Time Operation, Throughput Optimization, Multimedia Systems, Communication Systems, Data Storage, Hardware Architecture, VLSI Design, Performance Optimization, Simulations, Compression Performance, Hardware Design, Throughput Enhancement, System-Level Optimization, Power Efficiency, Resource Management, Data Transfer Efficiency, Low-Cost Implementation.

Development and evaluation of pulsatile drug delivery system of Telmisartan

The aim of the present study was the formulation of pulsatile tablets of Telmisartan by press coating technique that could release the drug time controlled manner. The core tablets and press coated pulsatile tablet of Telmisartan was prepared by direct compression technique. The core tablets were formulated using super disintegrating agent cross carmilose sodium and pulsatile tablets were by using polymer HPMC and Ethyl cellulose in various concentrations. Compatibility studies indicate that there is no interaction between the excipient and the drug. All tablets formulations were subjected to evaluation parameter. All the parameters were found within the limit. Pulsatile tablets formulation F3, F4 and F5 showed sufficient lag time of 4hr, 3hr and 4 hr respectively. The optimized formulation of F3 was consider as the best formulation with respect to drug content, hardness, lag time and In vitro drug release. Formulation F3 showed optimum lag time of 4 hour and release maximum drug in 8 hour. The developed formulation was found to be stable during the stability studies of 3 month indicating good stability of the tablets.

Auxetic Closed Cell Nylon Foams Produced by Vacuum and Mechanical Compression (VMC)

AbstractThis research presents a method to convert a conventional Nylon closed cell foam, with a density of 48 kg m−3, into auxetic metamaterials with densities ranging between 68 and 138 kg m−3. The samples are produced by applying vacuum and mechanical compression techniques. The study reports on the effect of processing parameters such as vacuum time, temperature, and mechanical pressure. Under optimized conditions, the resulting auxetic foams exhibit a tensile Poisson's ratio as low as −0.86, while the minimum compressive Poisson's ratio for the same sample reached −0.16. Based on morphological analyses via scanning electron microscopy (SEM), structural changes induced by the treatment are determined to relate with tensile and compressive properties, including modulus and strength, as well as the analysis of the stress level for different strains. The characterizations also include differential scanning calorimetry (DSC) and dynamic mechanical analysis (DMA) for the Nylon foam before and after conversion. These findings underscore the potential applications of these auxetic foams in sports and military protective gears, as well as energy dissipation systems.

Adaptive Compression Techniques for Lightweight Object Detection in Edge Devices

Adaptive Compression Techniques for Lightweight Object Detection in Edge Devices

A Survey on Model Compression for Large Language Models

Abstract Large Language Models (LLMs) have transformed natural language processing tasks successfully. Yet, their large size and high computational needs pose challenges for practical use, especially in resource-limited settings. Model compression has emerged as a key research area to address these challenges. This paper presents a survey of model compression techniques for LLMs. We cover methods like quantization, pruning, and knowledge distillation, highlighting recent advancements. We also discuss benchmarking strategies and evaluation metrics crucial for assessing compressed LLMs. This survey offers valuable insights for researchers and practitioners, aiming to enhance efficiency and real-world applicability of LLMs while laying a foundation for future advancements.

Spatio-spectral mapping of a microbottle laser

The spatio-spectral mapping of a whispering gallery mode microbottle laser has been presented. The dye-doped polymer coated microbottle resonator (MBR) of diameter 237μm, fabricated by soften and compress technique was locally excited by a narrow pump-stripe and the lasing modes having a quality factor of 1.5×104 with 0.33 nm free spectral range were observed. The localized and delocalized lasing modes were investigated throughout the resonator axis. Unlike the tapered fiber excitation geometry, the lasing modes have been read-out by scanning a collection fiber in the imaging plane allowing the spectral mapping over the entire resonator length. When pumped at the neck of the MBR, the spatial correlations of the spatio-spectral maps reveal the excitation of modes in the unpumped regions of the resonator.

Optimizing Top-K query processing on GPU using data compression and pre-filtering

In large-scale data analysis, efficient Top-K query processing is critical for numerous applications in science, industry, and society. Traditional approaches often involve substantial data transfer and computational overhead, making it difficult to meet the scalability and efficiency demands of modern datasets. This paper proposes a GPU-accelerated Top-K query processing method that integrates data compression and pre-filtering techniques to address these challenges. By partitioning and compressing data on the host side, it alleviates common PCIe bottlenecks in heterogeneous computing environments. A metadata-driven pre-filtering technique further reduces the data volume processed on the GPU, significantly improving query performance, particularly when handling anti-correlated datasets. Experimental results demonstrate that this method markedly reduces data transfer and processing time, confirming its effectiveness in enhancing the efficiency and scalability of Top-K query processing compared to existing methods.

Evaluation of Mechanical and Thermomechanical Water Vapor Compression Techniques for Enabling High Temperature Lift Hydration-Based Chemical Heat Pumps

Abstract Achieving high temperature lifts (>200K) via a chemical heat pump based on salt hydration/dehydration reactions requires the transport of water vapor from low to high pressure. Alternative compression approaches require condensing of low-pressure water vapor, pumping of liquid water, and subsequent evaporation when the low-side pressure corresponds to sub-ambient water saturation temperatures. Thus, this study compares four steam compression methods for use within a chemical heat pump system based on a reversible calcium oxide hydration/dehydration reaction with a temperature lift from 350∘C heat to >600∘C. Purely mechanical and thermochemical/mechanical compression technologies are considered. A parametric study of maximum allowable temperature, the isentropic efficiency of mechanical compressors, the effectiveness of heat exchangers, and the assumed allowable heat exchanger pressure drop is conducted to determine the mechanical and thermal energy consumed per kilogram of compressed steam. The system complexity in terms of the number of main system components, maximum pressure ratio, and maximum allowable temperature is estimated. Model results show an absorption-based steam compressor has the highest exergetic efficiency for the required chemical heat pump conditions. This system configuration was then experimentally demonstrated to illustrate the impact of system performance on component effectiveness.

Efficient Video Compression Using Afterimage Representation.

Recent advancements in large-scale video data have highlighted the growing need for efficient data compression techniques to enhance video processing performance. In this paper, we propose an afterimage-based video compression method that significantly reduces video data volume while maintaining analytical performance. The proposed approach utilizes optical flow to adaptively select the number of keyframes based on scene complexity, optimizing compression efficiency. Additionally, object movement masks extracted from keyframes are accumulated over time using alpha blending to generate the final afterimage. Experiments on the UCF-Crime dataset demonstrated that the proposed method achieved a 95.97% compression ratio. In binary classification experiments on normal/abnormal behaviors, the compressed videos maintained performance comparable to the original videos, while in multi-class classification, they outperformed the originals. Notably, classification experiments focused exclusively on abnormal behaviors exhibited a significant 4.25% improvement in performance. Moreover, further experiments showed that large language models (LLMs) can interpret the temporal context of original videos from single afterimages. These findings confirm that the proposed afterimage-based compression technique effectively preserves spatiotemporal information while significantly reducing data size.

No-Reference Objective Quality Metrics for 3D Point Clouds: A Review.

Three-dimensional (3D) applications lead the digital transition toward more immersive and interactive multimedia technologies. Point clouds (PCs) are a fundamental element in capturing and rendering 3D digital environments, but they present significant challenges due to the large amount of data typically needed to represent them. Although PC compression techniques can reduce the size of PCs, they introduce degradations that can negatively impact the PC's quality and therefore the object representation's accuracy. This trade-off between data size and PC quality highlights the critical importance of PC quality assessment (PCQA) techniques. In this article, we review the state-of-the-art no-reference (NR) objective quality metrics for PCs, which can accurately estimate the quality of generated and compressed PCs solely based on feature information extracted from the distorted PC. These characteristics make NR PCQA metrics particularly suitable in real-world application scenarios where the original PC data are unavailable for comparison, such as in streaming applications.

Unified extractive-abstractive summarization: a hybrid approach utilizing BERT and transformer models for enhanced document summarization

With the exponential proliferation of digital documents, there arises a pressing need for automated document summarization (ADS). Summarization, a compression technique, condenses a source document into concise sentences that encapsulate its salient information for summary generation. A primary challenge lies in crafting a dependable summary, contingent upon both extracted features and human-established parameters. This article introduces an intelligent methodology that seamlessly integrates extractive and abstractive techniques to ensure heightened relevance between the input document and its summary. Initially, input sentences undergo transformation into representations utilizing BERT, subsequently transposed into a symmetric matrix based on their similarity. Semantically congruent sentences are then extracted from this matrix to construct an extractive summary. The transformer model integrates an objective function highly symmetric and invariant under unitary transformation for language generation. This model refines the extracted informative sentences and generates an abstractive summary akin to manually crafted summaries. Employing this hybrid summarization technique on the CNN/DailyMail dataset and DUC2004, we evaluate its efficacy using ROUGE metrics. Results demonstrate the superiority of our proposed technique over conventional summarization methods.

Enabling On-Device Inference of Large Language Models : Challenges, Techniques, and Applications

This comprehensive article explores the cutting-edge techniques and challenges associated with on-device inference of Large Language Models (LLMs), a transformative approach that brings advanced AI capabilities directly to mobile and edge devices. The article delves into the intricate balance between the computational demands of LLMs and the resource constraints of mobile hardware, presenting a detailed analysis of various strategies to overcome these limitations. Key areas of focus include model compression techniques such as pruning and knowledge distillation, quantization methods, and the development of efficient model architectures. The article also examines the role of specialized hardware accelerators, including Neural Processing Units (NPUs), FPGAs, and ASICs, in enhancing on-device performance. Additionally, the article addresses critical aspects of memory management and optimization strategies crucial for efficient LLM deployment. Through a rigorous evaluation of performance metrics, the article offers insights into the trade-offs between model size, inference speed, and accuracy. It further explores diverse applications and use cases, from real-time language translation to privacy-preserving text analysis, highlighting the transformative potential of on-device LLM inference. The article concludes with an examination of ongoing challenges and future research directions, including improving energy efficiency, enhancing model adaptability, and addressing privacy and security concerns. This comprehensive article provides researchers, developers, and industry professionals with a thorough understanding of the current state and future prospects of on-device LLM inference, underlining its significance in shaping the next generation of AI-powered mobile and IoT applications.

On the Robustness of Compressed Models with Class Imbalance

Deep learning (DL) models have been deployed in various platforms, including resource-constrained environments such as edge computing, smartphones, and personal devices. Such deployment requires models to have smaller sizes and memory footprints. To this end, many model compression techniques proposed in the literature successfully reduce model sizes and maintain comparable accuracy. However, the robustness of compressed DL models against class imbalance, a natural phenomenon in real-life datasets, is still under-explored. We present a comprehensive experimental study of the performance and robustness of compressed DL models when trained on class-imbalanced datasets. We investigate the robustness of compressed DL models using three popular compression techniques (pruning, quantization, and knowledge distillation) with class-imbalanced variants of the CIFAR-10 dataset and show that compressed DL models are not robust against class imbalance in training datasets. We also show that different compression techniques have varying degrees of impact on the robustness of compressed DL models.