Search

Musicians and audio engineers sculpt and transform their sounds by connecting multiple processors, forming an audio processing graph. However, most deep-learning methods overlook this real-world practice and assume fixed graph settings. To bridge this gap, we develop a system that reconstructs the entire graph from a given reference audio. We first generate a realistic graph-reference pair dataset and train a simple blind estimation system composed of a convolutional reference encoder and a transformer-based graph decoder. We apply our model to singing voice effects and drum mixing estimation tasks. Evaluation results show that our method can reconstruct complex signal routings, including multi-band processing and sidechaining., Comment: Accepted to ICASSP 2023

In the pursuit of artificial general intelligence (AGI), we tackle Abstraction and Reasoning Corpus (ARC) tasks using a novel two-pronged approach. We employ the Decision Transformer in an imitation learning paradigm to model human problem-solving, and introduce an object detection algorithm, the Push and Pull clustering method. This dual strategy enhances AI's ARC problem-solving skills and provides insights for AGI progression. Yet, our work reveals the need for advanced data collection tools, robust training datasets, and refined model structures. This study highlights potential improvements for Decision Transformers and propels future AGI research.

Feature fusion modules from encoder and self-attention module have been adopted in semantic segmentation. However, the computation of these modules is costly and has operational limitations in real-time environments. In addition, segmentation performance is limited in autonomous driving environments with a lot of contextual information perpendicular to the road surface, such as people, buildings, and general objects. In this paper, we propose an efficient feature fusion method, Feature Fusion with Different Norms (FFDN) that utilizes rich global context of multi-level scale and vertical pooling module before self-attention that preserves most contextual information while reducing the complexity of global context encoding in the vertical direction. By doing this, we could handle the properties of representation in global space and reduce additional computational cost. In addition, we analyze low performance in challenging cases including small and vertically featured objects. We achieve the mean Interaction of-union(mIoU) of 73.1 and the Frame Per Second(FPS) of 191, which are comparable results with state-of-the-arts on Cityscapes test datasets., Comment: 5 pages, 4 figures

Unmanned aerial vehicles (UAVs) are expected to be used extensively in the future for various applications, either as user equipment (UEs) connected to a cellular wireless network, or as an infrastructure extension of an existing wireless network to serve other UEs. Next generation wireless networks will consider the use of UAVs for joint communication and radar and/or as dedicated radars for various sensing applications. Increasing number of UAVs will naturally result in larger number of communication and/or radar links that may cause interference to nearby networks, exacerbated further by the higher likelihood of line-of-sight signal propagation from UAVs even to distant receivers. With all these, it is critical to study network coexistence of UAV-mounted base stations (BSs) and radar transceivers. In this paper, using stochastic geometry, we derive closed-form expressions to characterize the performance of coexisting UAV radar and communication networks for spectrum overlay multiple access (SOMA) and time-division multiple access (TDMA). We evaluate successful ranging probability (SRP) and the transmission capacity (TC) and compare the performance of TDMA and SOMA. Our results show that SOMA can outperform TDMA on both SRP and TC when the node density of active UAV-radars is larger than the node density of UAV-comms.

Photosystem I (PS I) has a symmetric structure with two highly similar branches of pigments at the center that are involved in electron transfer, but shows very different efficiency along the two branches. We have determined the structure of cyanobacterial PS I at room temperature (RT) using femtosecond X-ray pulses from an X-ray free electron laser (XFEL) that shows a clear expansion of the entire protein complex in the direction of the membrane plane, when compared to previous cryogenic structures. This trend was observed by complementary datasets taken at multiple XFEL beamlines. In the RT structure of PS I, we also observe conformational differences between the two branches in the reaction center around the secondary electron acceptors A1A and A1B. The π-stacked Phe residues are rotated with a more parallel orientation in the A-branch and an almost perpendicular confirmation in the B-branch, and the symmetry breaking PsaB-Trp673 is tilted and further away from A1A. These changes increase the asymmetry between the branches and may provide insights into the preferential directionality of electron transfer.

The purpose of this study was to investigate differences in evaluations of election broadcasting according to the audience’s language awareness. We first summarized the implications of previous research on the characteristics of electoral broadcast language and then conducted experiments and FGI studies with language experts and ordinary people. Using an election broadcasting language evaluation scale, ordinary people and language experts rated several five-minute election videos; the ratings of “publicness,” “fairness,” and “soundness” showed differences between the two groups. In general, experts’ evaluations involved the genre characteristics of election broadcasting more than ordinary people’s did. Based on these results, we propose evaluation items taking into account the genre characteristics of election broadcasting and call for an evaluation scale to be developed not only for deliberation, but also for advanced education or as guidelines.

For American Physical Society March Meeting 2020, Session on Detectors, Sensors, and Transducers APS March Meeting 2020 MEMS acoustic sensors were developed based on the mechanically coupled auditory structure of the Ormia Ochracea fly in order to enhance underwater directional sound-sensing technologies. MEMS directional acoustic sensors consists of two wings connected by a bridge in the middle. The entire mechanical structure is connected to a substrate using two torsional legs. The mechanical vibrations under sound excitation is transduced to an electrical signal using interdigitated comb finger capacitors attached to the edges of the wings. This presentation covers the design, fabrication, and characterization of Ormia-based MEMS directional acoustic sensor operated underwater as an inertial sensor. The sensors were designed using FE modeling tools and fabricated using commercially available MEMSCap SOIMUMPS processes. Characterization was performed in air and underwater, showing the predicted frequency and directional responses. For underwater operation, the sensors were housed in a near-neutral-buoyant, hermetically sealed enclosure. Results indicate that the MEMS acoustic sensor's microphone characteristics are preserved when operated as accelerometers, and they have a great potential to be used for underwater applications in a neutral-buoyant configuration. GIMS Office of Naval Research GIMS

The aim of this study is to analyze the adverse effects of statements making personal attacks in broadcasting debates, to understand the problems involved, and to propose measures to redress the same. For this purpose, based on the preceding studies on the personal attack phenomenon in broadcasting debates, the study reviewed the target and contents of personal attack in order to prepare an analysis frame. For the analysis of personal attack statements in broadcasting debates, the study analyzed debate programs of three main broadcasting companies for 20 times. As a result of the study, it was found that the attacks mainly focused on the expertise and bias of debaters. Reliability, validity, and suitability on the evidence of speaking to the other parts formed the majority of cases. There were also instances of such attacks involving affiliated groups of debaters. Based on the results, the study proposed improvement guidelines for language expression in broadcasting debates for the purpose of improvement of personal attack phenomenon in broadcasting debates.

The magnetically levitated system, so called maglev system, has been researched and developed with the purpose of vacuum compatibility in the semiconductor industry. In the maglev system, the back electromotive force is inevitably generated when the system moves. The back electromotive force causes force/moment disturbances. Especially, the moment disturbances have negative effect on controlling the rotational motions (θ x , θ y , θ z ) whose control bandwidth is low. Therefore, the back electromotive force causes rotational motion errors. The rotational motion errors should be suppressed since they prevent high speed motion of the maglev system due to the rotational motion allowance of sensors. The rotational motion errors are suppressed by compensating the back electromotive force. In this paper, the back electromotive force, the cause of the rotational motion errors, is mathematically found in terms of the mover velocity and element of force-current matrix. A maglev system without the compensation was simulated and the rotational motion errors due to back electromotive force were found. Then, a feedforward compensated system using a mathematically modeled back electromotive force was simulated. It was verified that the feedforward compensation method with the modeled equation could be useful for suppressing rotational motion errors.

본 연구에서는 수사구조이론(RST: Rhetorical Structure Theory)을 활용하여 학습자의 텍스트 의미 구성 능력을 진단할 수 있다고 보고, 텍스트 생산자의 의도성이 강한 설득 텍스트를 대상으로 학습자의 의미 구성 능력을 진단할 수 있는 도구의 형식을 탐색하고자 하였다. 이를 위해 진단 도구 개발을 위한 기준과 그에 따른 네 가지 도구의 형식을 설정한 뒤, 초, 중, 고등학교 학습자를 대상으로 네 가지 진단 도구 형식을 적용해 봄으로써 그 타당성을 확인하였다. 그 결과, `진단의 타당성` 측면에서는 네 가지 진단 도구의 형식 모두 학습자들의 중심 내용 설정 능력, 목적과 상황을 고려한 내용 생성 및 조직 능력, 텍스트의 전체 체제와 구조 조망 능력을 진단하는 데 무리가 없는 것으로 확인되었다. 다만, 주어진 내용을 자유롭게 변형할 수 있는 C, D 유형의 결과가 B 유형의 반응과 큰 차이가 없었던 것으로 나타났다. 다음으로 `시행의 용이성` 측면에서는 원고지로 의미 구성 결과를 옮기는 방식을 포함한 C 유형과 D 유형이 A, B 유형에 비해 적합성이 떨어지는 것으로 확인되었다. 즉, 의미 구성 능력을 진단하는 도구를 개발하는 데에 있어서 글쓰기 과정을 포함하는 것이 진단에 직접적인 영향을 미치지 않을 수 있음을 확인하였다. 마지막으로 `자료의 적합성` 측면에서는 A, C 유형이 B, D 유형에 비해 상대적으로 자료 가공이 용이한 것으로 판단되었다. 이상의 결과를 종합해 볼 때, A, B 유형이 C, D 유형에 비해 진단 도구로서 우위를 확보할 것으로 최종 판단하였다.

Today various IoT platforms exist, but most of them only consider IPv6, without other types of network including private network. They support, therefore, the private network to use virtual private network or through the hardware gateway. To solve this limitation, we propose a IoT platform that provides IoT Services on private networks. Also we verify this platform though constructing a testbed.Keywords:Internet of Things, Proxy, IoT Platform, Smart Building, Private Network 용이한 사설망 지원을 위한 IoT 플랫폼 구현 및 서비스 운영 방법 이 경 훈 † ⋅김 두 영 †† ⋅송 세 헌 ††† ⋅이 상 일 †††† ⋅박 재 현 †††† 요 약 현재 출시된 다양한 사물인터넷(IoT) 플랫폼은 대부분 IPv6의 네트워크 환경을 고려하여 개발되었지만, 아직까지 IPv4에서 IPv6로의 전환율은 저조하다. 따라서 IPv4기반의 네트워크에서 운용되는 IoT 플랫폼은 IP부족현상을 해결하기 위해 기구축된 사설망을 지원해야 한다. 기존 IoT 플랫폼들은 전문가에 의해 가상 사설망을 이용하거나 하드웨어 스위치/AP(Access Point)를 이용하여 사설망을 지원하고 있다. 본 연구에서는 기존 IoT 플랫폼들이 사설망 환경을 지원하기 위해 필요한 전문가 활용 또는 하드웨어 도입에 따른 추가 비용 문제를 해결하고, 비전문가도 사설망 환경에서 IoT 서비스를 제공할 수 있는 IoT 플랫폼을 개발하였다. 또한 IoT 플랫폼을 테스트베드에 적용하여 IoT 플랫폼의 기능 검증을 수행하였다.키워드:사물인터넷, 프록시, IoT 플랫폼, 스마트빌딩, 사설망

Extracorporeal membrane oxygenation (ECMO) is a life support system that circulates the blood through an oxygenating system to temporarily (days to months) support heart or lung function during cardiopulmonary failure until organ recovery or replacement. Currently, the need for high levels of systemic anticoagulation and the risk for bleeding are main drawbacks of ECMO that can be addressed with a redesigned ECMO system. Our lab has developed an approach using microelectromechanical systems (MEMS) fabrication techniques to create novel gas exchange membranes consisting of a rigid silicon micropore membrane (SμM) support structure bonded to a thin film of gas-permeable polydimethylsiloxane (PDMS). This study details the fabrication process to create silicon membranes with highly uniform micropores that have a high level of pattern fidelity. The oxygen transport across these membranes was tested in a simple water-based bench-top set-up as well in a porcine in vivo model. It was determined that the mass transfer coefficient for the system using SµM-PDMS membranes was 3.03 ± 0.42 mL O2 min-1 m-2 cm Hg-1 with pure water and 1.71 ± 1.03 mL O2 min-1 m-2 cm Hg-1 with blood. An analytic model to predict gas transport was developed using data from the bench-top experiments and validated with in vivo testing. This was a proof of concept study showing adequate oxygen transport across a parallel plate SµM-PDMS membrane when used as a membrane oxygenator. This work establishes the tools and the equipoise to develop future generations of silicon micropore membrane oxygenators.

Diffusion-based bioartificial pancreas (BAP) devices are limited by poor islet viability and functionality due to inadequate mass transfer resulting in islet hypoxia and delayed glucose-insulin kinetics. While intravascular ultrafiltration-based BAP devices possess enhanced glucose-insulin kinetics, the polymer membranes used in these devices provide inadequate ultrafiltrate flow rates and result in excessive thrombosis. Here, we report the silicon nanopore membrane (SNM), which exhibits a greater hydraulic permeability and a superior pore size selectivity compared to polymer membranes for use in BAP applications. Specifically, we demonstrate that the SNM-based intravascular BAP with ~10 and ~40 nm pore sized membranes support high islet viability (>60%) and functionality (85%) at clinically relevant islet density (5,700 IE/cm2), c-peptide concentration of 144 pM in the outflow ultrafiltrate, and hemocompatibility under convection. These promising findings offer insights on the development of next generation of full-scale intravascular devices to treat T1D patients in the future.

The technique of semidefinite programming (SDP) relaxation can be used to obtain a nontrivial bound on the optimal value of a nonconvex quadratically constrained quadratic program (QCQP). We explore concave quadratic inequalities that hold for any vector in the integer lattice ${\bf Z}^n$, and show that adding these inequalities to a mixed-integer nonconvex QCQP can improve the SDP-based bound on the optimal value. This scheme is tested using several numerical problem instances of the max-cut problem and the integer least squares problem., Comment: 24 pages, 1 figure

Publication in the conference proceedings of EUSIPCO, Lisbon, Portugal, 2014

Emerging IT technologies, specially Internet communication and web-based applications are adopted to the modern distributed control systems. This paper defines a web-based application domain for a distributed control system connected by Internet. XML(eXtensible Markup Language) is used for representing a control system and control devices. These IT technologies makes a distributed control system flexible and scalable than existing distributed systems with the standard RMI protocols such as CORBA or DCOM.

Coherent diffraction imaging (CDI) using synchrotron radiation, X-ray free electron lasers (X-FELs), high harmonic generation, soft X-ray lasers, and optical lasers has found broad applications across several disciplines. An active research direction in CDI is to determine the structure of single particles with intense, femtosecond X-FEL pulses based on diffraction-before-destruction scheme. However, single-shot 3D structure determination has not been experimentally realized yet. Here we report the first experimental demonstration of single-shot 3D structure determination of individual nanocrystals using ~10 femtosecond X-FEL pulses. Coherent diffraction patterns are collected from high-index-faceted nanocrystals, each struck by a single X-FEL pulse. Taking advantage of the symmetry of the nanocrystal, we reconstruct the 3D structure of each nanocrystal from a single-shot diffraction pattern at ~5.5 nm resolution. As symmetry exists in many nanocrystals and virus particles, this method can be applied to 3D structure studies of such particles at nanometer resolution on femtosecond time scales., Comment: 18 pages, 4 figures

Treball desenvolupat dins el marc del programa 'European Project Semester'. This document describes the full development process of a wooden geodesic dome. In this paper, all the components of the dome will be discussed. During the design of the whole dome, the team used sustainability as the first parameter to base their decisions. The result of this development is a complete solution for a geodesic dome structure with an automatic door and window.