Search

Depth maps play an important role in the representation of 3D information. They are often simultaneously acquired with color images; however, their resolution is significantly lower than that of color images owing to hardware limitations. In this paper, we propose a novel approach to upsample depth maps by using geometric deformation instead of pixel value refinement, which is employed in a majority of existing methods. This approach, known as grid warping, displaces the position of blurred pixels around the edge towards the center of the edge. The displacement vector for warping is obtained from an analysis of the corresponding high-resolution color image. Furthermore, we propose an edge signal and displacement vector modeling for a more effective analysis. The experimental results show that the proposed method significantly improves the quantitative and visual performance, as compared to state-of-the-art methods. The source codes of the proposed method will be available at https://github.com/yym064/DeepGridWarp.

Piezoelectric energy harvesters (PEHs) are widely used to convert energy from a piezoelectric transducer into a stable DC form, which enables low-power IoT devices to have an unlimited operating life without using batteries. Under weak excitation conditions, however, the power-extraction efficiency of conventional PEHs is too low to provide power even to low-power IoT devices that requires low operation voltages less than 2 V. This paper proposes an asymmetric synchronous electric charge extraction (ASECE) scheme that improves the extraction efficiency of PEHs at low output voltages under weak excitation. The proposed ASECE is implemented using 0.18 μm CMOS technology. The figure-of-merits (FOMs) of the proposed ASECE while operating under 2 V of output voltage are 7.14 and 6.24 at weak and strong excitations, respectively. The maximum FOM for various different excitation levels is observed to be as high as 7.7. The proposed ASECE is superior to prior art with respect to FOM, by at least 1.15×, 1.63×, and 2.21× under 2 V, 1 V, and 0.5 V outputs, respectively, under strong excitation.

Direct AC LED drivers have advantages over conventional LED drivers in size and cost because of no need for bulky and expensive inductive components. However, none of conventional direct AC LED drivers researched so far support triode AC switch (TRIAC) compatibility at low percent flicker. This paper achieves both TRIAC dimmer compatibility and low percent flicker by correlating a phase-cut ratio to luminance and using always-on multi-arrays, respectively, without the need of a dedicated IC. The proposed LED driver consists of a phase-cut/DC converter, a switch controller, a current regulator, and three LED arrays. The experimental results based on commercial 15W LEDs show that the percent flicker of 18.6% @ 120 Hz is achieved with an efficiency of 84.7% and a power factor of 97.3% @ 120 V.

One of the most important and recurrent tasks in managing a store is to provide accurate information of products to customers by maintaining up-to-date price tags of every item. Till today, however, updating price tags is done manually in most markets, which is labor-intensive, time-consuming, and error-prone. Adopting wirelessly reconfigurable electronic price tags could be of significant benefit to this problem. A key requirement of such a system, then, is to establish a robust, reliable, and scalable wireless communication network while spending minimal energy on thousands of densely-deployed and battery-powered price tags. In this work, we present a comprehensive study of electronic shelf labeling (ESL) system including its design and real-world deployment experiences. While the overall design turns out to be relatively straightforward, our preliminary study using 1000 real commercial e-tags reveals technical challenges in reliable and low-power wireless communication with ultra-high node density in scatter-rich indoor environments. To address the problem, we adopt multi-radio and multi-channel diversity techniques which need to modify only the gateways but significantly improve overall system performance. Our evaluation from a real-world deployment of 550 tags at an actual convenience store shows that our proposal achieves 98.51% network connectivity, 590 ms average latency for price updates, and over 5 years of e-tag lifetime. ESL system is part of the upcoming smart market era, and we see this work as a practical industrial application case study of Internet of Things technologies.

Low-power wireless network for the emerging Internet of Things (IoT) should be reliable enough to satisfy the application requirements, and also energy-efficient for embedded devices to remain battery powered. Synchronized communication methods such as Time Slotted Channel Hopping (TSCH) have shown promising results for these purposes, achieving end-to-end reliability over 99% with low duty-cycles. However, they lack one thing: flexibility to support a wide variety of applications and services with unpredictable traffic load and routing topology due to “fixed” slotframe sizes. To this end, we propose TESLA, a traffic-aware elastic slotframe adjustment scheme for TSCH networks which enables each node to dynamically self-adjust its slotframe size at run time. TESLA aims to minimize its energy consumption without sacrificing reliable packet delivery by utilizing incoming traffic load to estimate channel contention level experienced by each neighbor. We extensively evaluate the effectiveness of TESLA on large-scale 110-node and 79-node testbeds, demonstrating that it achieves up to 70.2% energy saving compared to Orchestra (the de facto TSCH scheduling mechanism) while maintaining 99% reliability.