Input Mode Research Articles

COVID-19 Social Distancing Tracking and Monitoring System (SDMOS-19)

The Coronavirus disease (COVID-19), stemming from the SARS-CoV-2 virus, has garnered global concern as a virulent infectious ailment. Recognized as an epidemic by the World Health Organization (WHO), the persistently mutating virus sustains its transmission within communities. Individuals have been advised to uphold a safe interpersonal distance, notably around five feet, to mitigate its spread during social interactions. Addressing this imperative, an innovative automated social distancing detection system is conceived, leveraging the Convolutional Neural Network (CNN) algorithm. This system operates on two distinct input modes: static images and recorded videos recorded on closed-circuit television (CCTV). Remarkably, the proposed automated system adeptly quantifies and surveils the extent of social distancing among individuals in densely populated settings. A sophisticated framework accurately discerns social distancing compliance, delineating between hazardous and secure intervals via distinct red and green bounding box indicators. The culmination of this endeavor reveals an impressive 90% detection accuracy for both input modes. Notably, this proposed system holds substantial promise for implementation within sprawling premises such as expansive shopping malls or recreational parks. Seamlessly enforcing automated safety distance assessment expedites real-time insights to security departments and other relevant authorities. Consequently, the efficacy of citizens in upholding safe interpersonal distances can be promptly evaluated and, if necessary, corrective measures can be expeditiously instituted. This automated system ensures public health and safety maintenance, particularly in difficult circumstances.

Low-power, linear, and uniform bimodal resistive switching in proton conducting/insulating bilayer-based memristor

The emergence of artificial intelligence (AI) has recently necessitated the processing of big data. However, a separation between the memory and processing unit leads to significant time and power waste in conventional computing architecture. Therefore, memristors have been spotlighted due to their ability to store and process information at once with simple structures. However, conductive filament formation due to random ion movement induces stochastic resistive switching and nonlinear conductance modulation. In this study, we demonstrate a proton-electron coupled memristor controlled either by humidity or voltage using a tyrosine-rich peptide/Al2O3 bilayer to mitigate those bottlenecks. Introducing the proton insulating layer into a peptide memristor device significantly enhanced electrical performance in terms of linear weight update, uniform resistive switching, and low power consumption. The interlayered memristor device exhibited computing voltage reduction of 56%, two orders of magnitude increased switching window, and linearity and uniformity improvement by 36% and 85%, respectively. Consequently, the image recognition simulation showed a 20% accuracy improvement. These improvements are elucidated by filament confinement effect from the switching medium shift due to input mode. Therefore, these results not only provide a material-level strategy for high-performance memristors but also demonstrate bimodally controllable memristors for biomimetic electronics.

Local-Global Multi-Modal Distillation for Weakly-Supervised Temporal Video Grounding

This paper for the first time leverages multi-modal videos for weakly-supervised temporal video grounding. As labeling the video moment is labor-intensive and subjective, the weakly-supervised approaches have gained increasing attention in recent years. However, these approaches could inherently compromise performance due to inadequate supervision. Therefore, to tackle this challenge, we for the first time pay attention to exploiting complementary information extracted from multi-modal videos (e.g., RGB frames, optical flows), where richer supervision is naturally introduced in the weaklysupervised context. Our motivation is that by integrating different modalities of the videos, the model is learned from synergic supervision and thereby can attain superior generalization capability. However, addressing multiple modalities† would also inevitably introduce additional computational overhead, and might become inapplicable if a particular modality is inaccessible. To solve this issue, we adopt a novel route: building a multi-modal distillation algorithm to capitalize on the multi-modal knowledge as supervision for model training, while still being able to work with only the single modal input during inference. As such, we can utilize the benefits brought by the supplementary nature of multiple modalities, without compromising the applicability in practical scenarios. Specifically, we first propose a cross-modal mutual learning framework and train a sophisticated teacher model to learn collaboratively from the multi-modal videos. Then we identify two sorts of knowledge from the teacher model, i.e., temporal boundaries and semantic activation map. And we devise a local-global distillation algorithm to transfer this knowledge to a student model of single-modal input at both local and global levels. Extensive experiments on large-scale datasets demonstrate that our method achieves state-of-the-art performance with/without multi-modal inputs.

Ultra-compact SOI-based higher-order mode pass wavelength demultiplexer.

We propose an ultra-compact mode filtering wavelength demultiplexer design with a footprint of 3µm×3µm. Our device can route input T E 1 mode signals at 1310nm and 1550nm to different output ports while simultaneously blocking fundamental transverse electric (T E 0) mode input. Our device is designed based on the topology optimization algorithm, which results in an ultra-compact footprint combining wavelength routing and mode filtering functions for the first time, to the best of our knowledge. Our final optimized devices demonstrated insertion losses of 1.26dB and 1.47dB for the C- and O-band output ports, respectively, with inter-port crosstalk as low as -21.25d B and -30.99d B. The extinction ratios between T E 1 mode and T E 0 mode are 24.02dB and 30.12dB at the 1310nm and 1550nm output ports. The combination of small footprint, broad transmission bandwidth, T E 1 to T E 0 mode selectively filtering, and C- and O-band T E 1 mode demultiplexing functions make this a uniquely versatile device that can play an important role in future high density mode-wavelength multiplexing systems.

Low-noise cryogenic microwave amplifier characterization with a calibrated noise source.

Parametric amplifiers have become a workhorse in superconducting quantum computing; however, research and development of these devices has been hampered by inconsistent and, sometimes, misleading noise performance characterization methodologies. The concepts behind noise characterization are deceptively simple, and there are many places where one can make mistakes, either in measurement or in interpretation and analysis. In this article, we cover the basics of noise performance characterization and the special problems it presents in parametric amplifiers with limited power handling capability. We illustrate the issues with three specific examples: a high-electron mobility transistor amplifier, a Josephson traveling-wave parametric amplifier, and a Josephson parametric amplifier. We emphasize the use of a 50-Ω shot noise tunnel junction (SNTJ) as a broadband noise source, demonstrating its utility for cryogenic amplifier amplifications. These practical examples highlight the role of loss as well as the additional parametric amplifier "idler" input mode.

A Refined Wind Power Forecasting Method with High Temporal Resolution Based on Light Convolutional Neural Network Architecture

With a large proportion of wind farms connected to the power grid, it brings more pressure on the stable operation of power systems in shorter time scales. Efficient and accurate scheduling, operation control and decision making require high time resolution power forecasting algorithms with higher accuracy and real-time performance. In this paper, we innovatively propose a high temporal resolution wind power forecasting method based on a light convolutional architecture—DC_LCNN. The method starts from the source data and novelly designs the dual-channel data input mode to provide different combinations of feature data for the model, thus improving the upper limit of the learning ability of the whole model. The dual-channel convolutional neural network (CNN) structure extracts different spatial and temporal constraints of the input features. The light global maximum pooling method replaces the flat operation combined with the fully connected (FC) forecasting method in the traditional CNN, extracts the most significant features of the global method, and directly performs data downscaling at the same time, which significantly improves the forecasting accuracy and efficiency of the model. In this paper, the experiments are carried out on the 1 s resolution data of the actual wind field, and the single-step forecasting task with 1 s ahead of time and the multi-step forecasting task with 1~10 s ahead of time are executed, respectively. Comparing the experimental results with the classical deep learning models in the current field, the proposed model shows absolute accuracy advantages on both forecasting tasks. This also shows that the light architecture design based on simple deep learning models is also a good solution in performing high time resolution wind power forecasting tasks.

Transmission across a ribbon containing a square [formula omitted] impurity

We study the spectrum and transmission coefficient of plane waves propagating along square ribbons of varying widths, containing a square-shaped, PT-symmetric impurity region. We start with a zero-width ribbon (1D chain) and place a PT symmetric dimer. The spectrum is computed numerically and the instability gain is computed as a function of the gain/loss dimer strength. The transmission coefficient is obtained in closed form and examined as a function of wavevector and the gain/loss parameter. Next, we study a ribbon in a narrow ladder configuration containing a square PT impurity. As before, we compute the instability gain numerically and the transmission coefficient in closed form for the two possible input modes. Finally, we repeat the calculations for a wider ladder ribbon containing a Lieb-like impurity in a PT configuration. For all cases and transmission channels, we obtain transmission divergences in wavevector-gain/loss parameter space, whose number increases with the width of the ribbon.

An MV Hybrid Isolated DC Transformer for Wide Output Voltage Range Applications

DC transformer (DCT) is the key energy conversion equipment that converts medium voltage dc to low voltage dc. Aiming at the industrial loads requiring wide output voltage range and unidirectional power flow, a hybrid isolated DC transformer (HIT) as an alternative to the traditional MMC-DCT is proposed in this paper. It is composed by thyristors, diodes, and submodule (SM) based arms. The operation principle, circuit performance, and control strategy are introduced in detail. Particularly, a discontinuous input and continuous output currents (DICO) operation mode is further proposed to suppress submodule capacitor voltage ripple. Compared with the traditional MMC-DCT, HIT can save 50% SMs, 28% semiconductor cost and 46.3% total SM capacitance. The effectiveness and the theoretical analysis have been confirmed by both simulation and experimental results.

Refined fire detection and band selection method in hyperspectral remote sensing imagery based on sparse-VIT

Hyperspectral remote sensing has powerful ability to perceive information of land surface, and can be applied to the field of fire monitoring to achieve refined fire analysis. In this paper, we study the fire detection method based on hyperspectral remote sensing, design an end to end fire detection model based on sparse visual transformer, and propose a sensitive band selection method under the transformer framework. In order to suppress the interference of invalid bands in hyperspectral data, the sparse attention mechanism and the top-k selection mechanism are introduced in the model to zeroing the attention score of invalid bands. To achieve dimension reduction, a non-maximum attention suppression algorithm under the transformer framework is designed and band pruning is further introduced to eliminate invalid bands and redundant bands. The model adopts input mode of band-exclusive-token, which makes the pruning operation equivalent to band selection. We publish a hyperspectral fire detection dataset and the performance of our proposed model is verified on the dataset.

Compact SOI Dual-Mode (De)multiplexer Based on the Level Set Method

Mode (de)multiplexer is an essential device in integrated multimode photonic systems. Here, we present a dual-mode (de)multiplexer that separates two input modes, TE0 and TE1, into two output ports while converting TE1 to TE0 mode. Based on the adjoint and level set method, the device features a small footprint of 9.4 μm × 2.9 μm, and a minimum feature size over 200 nm is achieved, affirming stable and reliable fabrication. Through simulations, we observed insertion losses of less than 0.28 dB for TE0 mode and 0.35 dB for TE1 mode within the wavelength range of 1500–1600 nm, accompanied by crosstalk levels lower than −30 dB. In our experimental tests, we achieved insertion losses of less than 0.89 dB for TE0 mode and 0.44 dB for TE1 modes within the 1530 nm to 1570 nm range, with crosstalk maintained below −25 dB. Furthermore, we conducted an experimental verification of the differences between the standard device and the boundary dilation/erosion device, observing an insertion loss degradation by 0.61 dB within a deviation range of ±40 nm, which demonstrates the device’s robustness to the fabrication. The proposed devices exhibit exceptional performance and feature a compact structure, thus holding significant potential for the development of future multimode integrated photonic circuits.

Control of the total orbital angular momentum of light beams propagating through a turbulent medium.

The robustness of the orbital angular momentum (OAM) of light beams propagating in a turbulent medium, e.g., atmosphere, is critical for many applications such as OAM-based free-space optical communications and remote sensing. However, the total OAM of a beam interacting with the turbulent medium inevitably changes. Here, we demonstrate a practical algorithm to control the total OAM of a beam transmitted through a time-evolving, turbulent medium by dynamically modulating the weights of two coherently superimposed OAM modes, which served as the input beam. A cross-OAM matrix is introduced, and applied for checking whether the desired total OAM in the output plane can be achieved. Furthermore, analytical relations between the weights of two input modes and the output total OAM, as well as its modulation range, are established. As a numerical example, we study the behavior of total OAM of the two-mode beam after passing through a thermal convection occurring in an aqueous medium and suggest a possible application of our strategy.

Enabling Visual Object Detection With Object Sounds via Visual Modality Recalling Memory.

When humans hear the sound of an object, they recall associated visual information and integrate the sound with recalled visual modality to detect the object. In this article, we present a novel sound-based object detector that mimics this process. We design a visual modality recalling (VMR) memory to recall information of a visual modality based on an audio modal input (i.e., sound). To achieve this goal, we propose a VMR loss and an audio-visual association loss to guide the VMR memory to memorize visual modal information by establishing associations between audio and visual modalities. With the visual modal information recalled through the VMR memory along with the original audio input, we perform audio-visual integration. In this step, we introduce an integrated feature contrastive loss that allows the integrated feature to be embedded as if it were encoded using both audio and visual modal inputs. This guidance enables our sound-based object detector to effectively perform visual object detection even when only sound is provided. We believe that our work is a cornerstone study that offers a new perspective to conventional object detection studies that solely rely on the visual modality. Comprehensive experimental results demonstrate the effectiveness of the proposed method with the VMR memory.

AI Accelerator with Ultralightweight Time-Period CNN-Based Model for Arrhythmia Classification.

This work proposes a classification system for arrhythmias, aiming to enhance the efficiency of the diagnostic process for cardiologists. The proposed algorithm includes a naive preprocessing procedure for electrocardiography (ECG) data applicable to various ECG databases. Additionally, this work proposes an ultralightweight model for arrhythmia classification based on a convolutional neural network and incorporating R-peak interval features to represent long-term rhythm information, thereby improving the model's classification performance. The proposed model is trained and tested by using the MIT-BIH and NCKU-CBIC databases in accordance with the classification standards of the Association for the Advancement of Medical Instrumentation (AAMI), achieving high accuracies of 98.32% and 97.1%. This work applies the arrhythmia classification algorithm to a web-based system, thus providing a graphical interface. The cloud-based execution of automated artificial intelligence (AI) classification allows cardiologists and patients to view ECG wave conditions instantly, thereby remarkably enhancing the quality of medical examination. This work also designs a customized integrated circuit for the hardware implementation of an AI accelerator. The accelerator utilizes a parallelized processing element array architecture to perform convolution and fully connected layer operations. It introduces proposed hybrid stationary techniques, combining input and weight stationary modes to increase data reuse drastically and reduce hardware execution cycles and power consumption, ultimately achieving high-performance computing. This accelerator is implemented in the form of a chip by using the TSMC 180 nm CMOS process. It exhibits a power consumption of 122 μW, a classification latency of 6.8 ms, and an energy efficiency of 0.83 μJ/classification.

Integration of AC/DC With Multi Input Mode Selection Converter for Fast Charging of Battery in Electric Vehicles

Integration of AC/DC With Multi Input Mode Selection Converter for Fast Charging of Battery in Electric Vehicles

Effects of the Type of Audio‐visual Input and Listenability on Young L2 Learners' Listening Comprehension

AbstractNumerous studies have compared the effect of input mode (video‐based vs. audio‐only) on the listening comprehension of adult second language (L2) learners. However, there is a notable absence of research regarding how various types of audio‐visual input in a listening test affect young learners' listening comprehension. To address this gap, this study examined the effects of audio‐visual input type and text listenability on the listening comprehension performance of 50 English language learners (ELLs) in Grades 3–5 in the U.S. In this study, participants completed a computer‐delivered English listening test comprising three test forms, each containing six passages. These passages included a combination of three video passages and three audio‐only passages. The video passages were carefully designed to balance two key variables in a counter‐balancing manner: the listenability of the text, assessed using Lexile Audio Measure, and the nature of visual elements, which encompassed speaker‐only visuals, visual‐only elements, and a combination of speaker‐and‐visuals. This study found that students performed equally well on both video‐based and audio‐only passages, and the type of visuals accompanying the content did not have a statistically significant impact on students' listening comprehension scores. However, the listenability of the text significantly influenced students' listening scores, especially in the context of audio‐only passages. Furthermore, responses to a post‐test questionnaire revealed that students had a positive perception of all three types of audio‐visual input. This study contributes to our understanding of how audio‐visual input types and text listenability affect the listening comprehension performance of Grades 3–5 ELLs.

The effectiveness of note taking through exposure to L2 input: A meta-analysis

AbstractThere has been increasing interest in the effects of note taking in second language (L2) research. However, no meta-analysis has been conducted to examine the relationship between note taking and learning through exposure to L2 input. We retrieved 28 effect sizes from 21 studies (N = 1992) to explore the overall effects of note taking as well as to examine the extent to which the effectiveness of note taking is likely to vary as a function of a set of potential moderators (i.e., learner variables, treatment variables, note-taking features, learning target, and measurement type). Results revealed that note taking had a small to medium positive overall effect on learning through exposure to L2 input (g = 0.56, 95% CI: 0.24–0.88). Subsequent moderator analyses revealed that variability in the size of note-taking effects across studies was explained by learner variables (context, region, orthographic scripts, institutional level), treatment variables (mode of input, material type), note-taking features (note-taking behavior, number of note-taking sessions, provision and type of note-taking strategy instruction, total length of instruction, opportunity to review notes), learning target, and measurement type. Based on the obtained findings, teachers are recommended to incorporate note taking in L2 classrooms. Pedagogical suggestions and directions for future research are also provided.

Numerical simulation of rebar corrosion in track bed of shield tunnel under stray current

Track bed is the important support structure at the bottom of shield tunnel, during metro operation, the leakage of stray current will lead to corrosion of rebar in the track bed, which will affect the safe operation of metro. In this paper, the corrosion calculation model of track bed rebar under stray current is established, and the effects of stray current input mode and concrete resistivity on the corrosion rate of track bed rebar are studied. The result shows the corrosion rate of track bed rebar decreases nonlinearly with the increase of concrete resistivity, and the reduction gradient of corrosion rate in bilateral input mode is greater than that in unilateral input mode. With the decrease of concrete resistance, the corrosion area will cover the left half in unilateral input mode and cover both ends of track bed rebar in bilateral input mode. When the input point voltage increases by 40%, the corrosion current density in the middle of track bed will increase by about 61%.

GMRLNet: A graph-based manifold regularization learning framework for placental insufficiency diagnosis on incomplete multimodal ultrasound data.

Multimodal analysis of placental ultrasound (US) and microflow imaging (MFI) could greatly aid in the early diagnosis and interventional treatment of placental insufficiency (PI), ensuring a normal pregnancy. Existing multimodal analysis methods have weaknesses in multimodal feature representation and modal knowledge definitions and fail on incomplete datasets with unpaired multimodal samples. To address these challenges and efficiently leverage the incomplete multimodal dataset for accurate PI diagnosis, we propose a novel graph-based manifold regularization learning (MRL) framework named GMRLNet. It takes US and MFI images as input and exploits their modality-shared and modality-specific information for optimal multimodal feature representation. Specifically, a graph convolutional-based shared and specific transfer network (GSSTN) is designed to explore intra-modal feature associations, thus decoupling each modal input into interpretable shared and specific spaces. For unimodal knowledge definitions, graph-based manifold knowledge is introduced to describe the sample-level feature representation, local inter-sample relations, and global data distribution of each modality. Then, an MRL paradigm is designed for inter-modal manifold knowledge transfer to obtain effective cross-modal feature representations. Furthermore, MRL transfers the knowledge between both paired and unpaired data for robust learning on incomplete datasets. Experiments were conducted on two clinical datasets to validate the PI classification performance and generalization of GMRLNet. State-of-the-art comparisons show the higher accuracy of GMRLNet on incomplete datasets. Our method achieves 0.913 AUC and 0.904 balanced accuracy (bACC) for paired US and MFI images, as well as 0.906 AUC and 0.888 bACC for unimodal US images, illustrating its application potential in PI CAD systems.

Review of English L2 Vocabulary Learning and Teaching: Concepts, Principles, and Pedagogy

Vocabulary knowledge is crucial for comprehending input and generating output. To date, vocabulary learning has been extensively researched regarding incidental vocabulary learning, the effect of various input modes and input enhancement techniques, yet more studies are needed to explore and discuss pedagogical implications. The monograph under review, authored by Lawrence J. Zwier and Frank Boers, bridges the gap between research and practice. The monograph elaborates on L2 vocabulary acquisition from theory development (Chapters 1 & 2), empirical studies (Chapters 3 & 4), curricular design (Chapters 5 to 7), and classroom practice (Chapters 8 to 11) respectively.

Hum-ble Beginnings: Developing Touch- and Proximity-Input-Based Interfaces for Zoo-Housed Giraffes’ Audio Enrichment

Though computer systems have entered widespread use for animals' enrichment in zoos, no interactive computer systems suited to giraffes have yet been developed. Hence, which input modes or audio stimuli giraffes might best utilise remains unknown. To address this issue and probe development of such systems alongside the animals themselves and zookeepers, researchers gathered requirements from the keepers and from prototyping with giraffes, then created two interfaces -- one touch-based and one proximity-based -- that play giraffe-humming audio or white noise when activated. Over two months of observation, giraffes utilised the proximity-based system more frequently than the touch-based one but in shorter episodes. Secondly, the study highlighted the significance of considering user-specific needs in computer systems' development: the lack of preference shown for any specific audio type indicates that the audio stimuli chosen were inappropriate for these giraffes. In addition, the paper articulates several lessons that can be drawn from human--computer interaction when one develops systems for animals and, in turn, what the findings presented mean for humans.