The growing commercial interest in indoor location-based services (ILBS) has spurred recent development of many positioning techniques. Due to the absence Global Positioning System (GPS) signal, other signals have been proposed for usage. Among them, Wi-Fi (802.11) emerges as a promising one due pervasive deployment wireless LANs (WLANs). In particular, fingerprinting attracting much attention recently because it does not require line-of-sight measurement access points (APs) and achieves...

To design fast neural networks, many works have been focusing on reducing the number of floating-point operations (FLOPs). We observe that such reduction in FLOPs, however, does not necessarily lead to a similar level re-duction latency. This mainly stems from inefficiently low per second (FLOPS). achieve faster we revisit popular operators and demonstrate FLOPS is due frequent memory access operators, especially depthwise con-volution. hence propose novel partial convolution (PConv)...

Provisioning random access functions in peer-to-peer on-demand video streaming is challenging, due to not only the asynchronous user interactivity but also unpredictability of group dynamics. In this paper, we propose VMesh, a distributed video-on-demand (VoD) scheme which efficiently supports seeking functionality. videos are divided into segments and stored at peers' local storage manner. An overlay mesh built upon peers support forward/backward seek, pause restart during playback. Our...

Wi-Fi fingerprinting has been extensively studied for indoor localization due to its deployability under pervasive WLAN. As the signals from access points (APs) may change to, example, AP movement or power adjustment, traditional approach is conduct site survey regularly in order maintain accuracy, which costly and time-consuming. Here, we study how accurately locate a target automatically update fingerprints presence of altered (or simply, “altered APs”). We propose Localization with...

Time series data, including univariate and multivariate ones, are characterized by unique composition complex multi-scale temporal variations. They often require special consideration of decomposition modeling to analyze. Existing deep learning methods on this best fit time only, have not sufficiently considered sub-series completeness. To address these challenges, we propose MSD-Mixer, a M ulti- S cale D ecomposition MLP- Mixer , which learns explicitly decompose represent the input in its...

Trilateration has been widely and successfully employed to locate outdoor mobile devices due its accuracy. However, it cannot be directly applied for indoor localization issues such as non-line-of-sight measurement multipath fading. Though fingerprinting overcomes these issues, accuracy is often hampered by signal noise the choice of similarity metric between vectors. We propose INTRI, a novel, simple effective technique combining strengths trilateration fingerprinting.

While deep learning demonstrates its strong ability to handle independent and identically distributed (IID) data, it often suffers from out-of-distribution (OoD) generalization, where the test data come another distribution (w.r.t. training one). Designing a general OoD generalization framework for wide range of applications is challenging, mainly due different kinds shifts in real world, such as shift across domains or extrapolation correlation. Most previous approaches can only solve one...

To design fast neural networks, many works have been focusing on reducing the number of floating-point operations (FLOPs). We observe that such reduction in FLOPs, however, does not necessarily lead to a similar level latency. This mainly stems from inefficiently low per second (FLOPS). achieve faster we revisit popular operators and demonstrate FLOPS is due frequent memory access operators, especially depthwise convolution. hence propose novel partial convolution (PConv) extracts spatial...

The delivery of quality video service often requires high bandwidth with low delay or cost in network transmission. Current routing protocols such as those used the Internet are mainly based on single-path approach (e.g., shortest-path routing). This cannot meet end-to-end requirement when is streamed over bandwidth-limited networks. In order to overcome this limitation, we propose multipath routing, where takes multiple paths reach its destination(s), thereby increasing aggregate...

In an on-demand video system, the repository generally has limited streaming capacities and may be far from users. order to achieve higher user capacity lower network transmission cost, distributed servers architecture can used, in which multiple local are placed close pools and, according their demands, dynamically cache contents streamed repository. We study a number of caching schemes as applied depending on whether is able multicast movie or not, exchange cached among themselves not. Our...

In traditional mobile streaming networks such as 3G cellular networks, all users pull streams from a server. Such model leads to high cost and problem in system scalability. this paper, we propose investigate fully distributed, scalable, cost-effective protocol distribute multimedia content mobiles peer-to-peer manner. Our protocol, termed Collaborative Streaming among Mobiles (COSMOS), makes use of broadcasting data sharing achieve performance (in terms delay, fairness, stream continuity,...

Peer-to-peer (P2P) technology has emerged as a promising scalable solution for live streaming to large group. In this paper, we address the design of overlay which achieves low source-to-peer delay, is robust user churn, accommodates asymmetric and diverse uplink bandwidth, continuously improves based on existing pool. A natural choice use mesh, where each peer served by multiple parents. Since delay in mesh depends its longest path through parents, study how optimize such while meeting...

Large-scale multimedia streaming over the Internet requires an enormous amount of server and network resources. Traditional client-server approaches allocate a dedicated stream from for each client request, which is expensive doesn't scale well. By using end hosts' huge bandwidth computational capacity, peer-to-peer technologies shed new light on media applications' development. Yet, locating supplying peers content delivery path maintenance are two major challenges in this area

Fusing fingerprints with mutual distance information potentially improves indoor localization accuracy. Such may be spatial (e.g., via inter-node measurement) or temporal dead reckoning). Previous approaches on fusion often require exact measurement, assume the knowledge of distribution, apply narrowly to some specific sensing technology scenario. Due random signal fluctuation, wireless are inherently noisy and cannot exactly measured. We hence propose Wi-Dist, a highly accurate framework...

In fingerprint-based indoor localization, keeping fingerprints current is important for localization accuracy. Because access point (AP) signals can change over time or suddenly, the traditional approach to survey site regularly and frequently, which laborious costly. The authors propose Chameleon, a novel survey-free localize users maintain an updated fingerprint database despite signal changes. clients are both location "queriers" unconscious "surveyors" (implicit crowdsourcing)....

In indoor localization based on Wi-Fi fingerprinting, a target sends its received signal strength indicator (RSSI) of access points (APs) to server estimate position. Traditionally, the estimates position by matching RSSI with fingerprints stored in database. Due noise fingerprint collection and measurement, this often results geographically disperse set reference (RPs), leading unsatisfactory estimation accuracy. To mitigate problem, we propose novel, efficient, highly accurate scheme...

As convolution has empowered many smart applications, dynamic further equips it with the ability to adapt diverse inputs. However, static and convolutions are either layout-agnostic or computation-heavy, making inappropriate for layout-specific e.g., face recognition medical image segmentation. We observe that these applications naturally exhibit characteristics of large intra-image (spatial) variance small cross-image variance. This observation motivates our efficient translation variant...

We study the forecasting problem for traffic with dynamic, possibly periodical, and joint spatial-temporal dependency between regions. Given aggregated inflow outflow of regions in a city from time slots 0 to <inline-formula><tex-math notation="LaTeX">$t - 1$</tex-math></inline-formula> , we predict at notation="LaTeX">$t$</tex-math></inline-formula> any region. Prior arts area often considered spatial temporal dependencies decoupled manner, or were rather computationally intensive training...

WiFi received signal strength (RSS) environment evolves over time due to the movement of access points (APs), AP power adjustment, installation and removal APs, etc. We study how effectively update an existing database fingerprints, defined as RSS values APs at designated locations, using a batch newly collected unlabelled (possibly crowdsourced) signals. Prior art either estimates locations new signals without updating fingerprints or filters out sufficiently embracing their features. To...

In a cellular system, if there are too many data users in cell, which we refer to as hot may suffer large delay, and system's quality of service (QoS) degrade. A dynamic channel-allocation scheme assigns more channels those cells under centralized control (CC). mobile-assisted forwarding (MADF), add an ad-hoc overlay the fixed infrastructure special channels-called channels-are used connect its surrounding cold without going through cell's base station. The forwarding-channel management MADF...

In this paper we propose an edge-directed error concealment (EDEC) algorithm, to recover lost slices in video sequences encoded by flexible macroblock ordering. First, the strong edges a corrupted frame are estimated based on neighboring frames and received area of current frame. Next, regions along these recovered using both spatial temporal pixels. Finally, remaining parts estimated. Simulation results show that compared existing boundary matching algorithm [1] exemplar-based inpainting...