Lossy image compression always faces a tradeoff between rate-distortion performance and compression/decompression speed. With the advent of neural image compression, hardware (GPU) becomes the new vertex in the tradeoff triangle. By resolving the high GPU dependency and improving the low speed of neural models, this paper proposes two non-GPU models that get the best of the three worlds. First, the CPU-friendly Independent Separable Down-Sampling (ISD) and Up-Sampling (ISU) modules are proposed to lighten the network while ensuring a large receptive field. Second, an asymmetric autoencoder architecture is adopted to boost the decoding speed. At last, the Inverse Quantization Residual (IQR) module is proposed to reduce the error caused by quantization. In terms of rate-distortion performance, our network surpasses the state-of-the-art real-time GPU neural compression work at medium and high bit rates. In terms of speed, our model's compression and decompression speeds surpass all other traditional compression methods except JPEG, using only CPUs. In terms of hardware, the proposed models are CPU friendly and perform stably well in a non-GPU environment. The code is publicly available at https://github.com/kengchikengchi/FasiNet.

Temporal action localization is an important task of computer vision. Though a variety of methods have been proposed, it still remains an open question how to predict the temporal boundaries of action segments precisely. Most works use segment-level classifiers to select video segments pre-determined by action proposal or dense sliding windows. However, in order to achieve more precise action boundaries, a temporal localization system should make dense predictions at a fine granularity. A newly proposed work exploits Convolutional-Deconvolutional-Convolutional (CDC) filters to upsample the predictions of 3D ConvNets, making it possible to perform per-frame action predictions and achieving promising performance in terms of temporal action localization. However, CDC network loses temporal information partially due to the temporal downsampling operation. In this paper, we propose an elegant and powerful Temporal Preservation Convolutional (TPC) Network that equips 3D ConvNets with TPC filters. TPC network can fully preserve temporal resolution and downsample the spatial resolution simultaneously, enabling frame-level granularity action localization with minimal loss of time information. TPC network can be trained in an end-to-end manner. Experiment results on public datasets show that TPC network achieves significant improvement in both per-frame action prediction and segment-level temporal action localization.

Human-machine systems required a deep understanding of human behaviors. Most existing research on action recognition has focused on discriminating between different actions, however, the quality of executing an action has received little attention thus far. In this paper, we study the quality assessment of driving behaviors and present WiQ, a system to assess the quality of actions based on radio signals. This system includes three key components, a deep neural network based learning engine to extract the quality information from the changes of signal strength, a gradient-based method to detect the signal boundary for an individual action, and an activity-based fusion policy to improve the recognition performance in a noisy environment. By using the quality information, WiQ can differentiate a triple body status with an accuracy of 97%, whereas for identification among 15 drivers, the average accuracy is 88%. Our results show that, via dedicated analysis of radio signals, a fine-grained action characterization can be achieved, which can facilitate a large variety of applications, such as smart driving assistants.

One of the most important topics in Human-Centered Computing (HCC) is to recognise human's activities. In this paper, the technology of wireless-based activity recognition is introduced. By using wireless signals, one can achieve Non-Line-Of-Sight (NLOS) recognition without carrying any devices. Also, it is easy to deploy a wireless-based recognition system due to the ubiquity of wireless communication systems. The basic idea is to detect different characteristics of signal propagation that correspond to the distinct human behaviors. As a result, action recognition is performed by analyzing the distinguishable features of signal propagation. This paper introduces the basic principles and applications of wireless-based activity recognition, and discusses the challenges and related performance metrics. Finally, open problems are discussed to point out the future research trends.

As the data processing demands have been increasing, different types of data processing systems are being developed. The new programming systems have different characteristics like types of data handled, processing technique and performance. However, multiple new systems have introduced difficulties for non-expert users like choosing the right system and usage methodology of the new systems. In order to relieve the burden of common users of conducting data processing tasks and taking relevant advantage of the systems features, we intend to integrate the popular programming systems and provide more efficient data processing services. In this paper, we propose to address the task scheduling problem for integrating multiple programming systems. We have designed a cluster based approach for task scheduling across multiple programming systems. This approach helps in minimizing the makespan of workflows and resource consumption. The simulation results show that the proposed approach can reduce the resource consumption significantly while achieving a low makespan for the workflows.

Coordination among users is inevitable in wireless communication for efficient medium access. Even though the data rate of individual user increases significantly, the performance of wireless network does not grow up accordingly due to the high MAC coordination overhead. In this paper, we present VFA, namely virtual frame aggregation, to achieve high coordination efficiency by amortizing the overhead over multiple transmissions. VFA provides a novel way to construct a winner cluster and allow the winners to transmit without interruption. Specifically, in a multicarrier network, every contending node chooses a subcarrier and the nodes are ordered by the index of the chosen subcarrier. When there are some subcarriers chosen by two or more nodes, an additional slot is exploited to reorder the collided nodes. Finally, all ordered nodes form a cluster and the transmissions are issued sequentially and uninterruptedly. Simulation results show that usually two slots are enough to construct a sufficiently large winner cluster. Moreover, VFA achieves a notable throughput gain over IEEE 802.11 as high as 120% with better fairness under various scenarios.

Coordination among users is an indispensable part in wireless networks for efficient medium access. Alone with the rapid increase of transmission rate, however, coordination time becomes insufferable. We present AFD, namely asymmetric full duplex, to achieve high coordination efficiency at nearly zero overhead. In AFD, channel contention is performed simultaneously with data transmission. We propose a 3D pipeline contention scheme where the contention process is divided into several parallel stages and executed in a pipelined manner in a 3D domain specified by time, frequency and spatial antenna. To mitigate the interference between the data packet and the contention signal, we adopt a singleton PN sequence as a contention pilot. AFD provides a novel network-scale full duplex capability. The performance is evaluated by both simulations and measurements in a testbed. AFD outperforms IEEE 802.11 significantly, i.e., the Jain's fairness index is around 0.95 with a throughput gain up to 120%.

Successive interference cancellation (SIC) is an effective way of multipacket reception to combat interference in wireless networks. We focus on link scheduling in wireless networks with SIC, and propose a layered protocol model and a layered physical model to characterize the impact of SIC. In both the interference models, we show that several existing scheduling schemes achieve the same order of approximation ratios, independent of whether or not SIC is available. Moreover, the capacity order in a network with SIC is the same as that without SIC. We then examine the impact of SIC from first principles. In both chain and cell topologies, SIC does improve the throughput with a gain between 20 and 100 percent. However, unless SIC is properly characterized, any scheduling scheme cannot effectively utilize the new transmission opportunities. The results indicate the challenge of designing an SIC-aware scheduling scheme, and suggest that the approximation ratio is insufficient to measure the scheduling performance when SIC is available.

Successive interference cancellation (SIC) is an effective way of multipacket reception to combat interference. As conventional CSMA (Carrier Sense Multiple Access) is designed for single packet reception, it is unclear whether or not CSMA performs well to exploit the SIC capability. In this paper, we analyze the performance of a simple CSMA protocol in a network with SIC. For a given link, we derive the residing areas of an interfering node when simultaneous transmission is allowed and when the interference is harmful, respectively. We show that, though SIC provides many new transmission opportunities, CSMA cannot effectively exploit them. There is a fundamental tradeoff in a CSMA protocol between exploiting the transmission opportunities from SIC and capturing the harmful interference. In many cases, when CSMA achieves its best performance, almost all new transmission opportunities are not exploited. It is therefore very necessary to design a new distributed access protocol in wireless networks with SIC.

Successive interference cancellation (SIC) is an effective way of multipacket reception to combat interference. We consider the problem of maximizing the number of successful transmissions based on the physical model in wireless networks with SIC at the physical layer. We propose weighted simultaneity graph to characterize the sequential detection nature of SIC and the accumulative effect of multiple interfering signals. A context-aware metric, transmission price, is defined to measure the interference of a link set. As maximizing the number of supported links is NP-hard, a greedy scheme is proposed to efficiently construct a near-optimal maximal feasible set of links. We show that the approximation performance is bounded by the transmission price of the constructed link set. The performance of the proposed scheme is further verified by simulation.

We consider the greedy scheduling based on the physical model in wireless networks with successive interference cancellation (SIC). There are two major stages in a scheduling scheme, link selection (to decide which link is scheduled next) and time slot selection (to deciding which slot is allocated to a given link). Most available schemes take a first-fit policy in the latter and strive to achieve good performance by careful selection of link ordering with respect to interference. Due to the accumulation effect and sequential detection nature of SIC, however, it is difficult to evaluate the interference of a link. As a result, many existing scheduling schemes become less efficient. In this paper, we take a new look on the problem and focus to the time slot selection stage. We define tolerance margin to measure the saturation of a link set and present two heuristic policies: one is to schedule a link to a slot such that the resulting set of links has a maximum tolerance margin; the other is to choose a slot such that the increase of tolerance margin is minimum. Simulation results show that the performance of the proposed schemes is better than the first-fit policy and is close to the optimal solution.

Successive interference cancellation (SIC) is an effective way of multipacket reception (MPR) to combat interference in wireless networks. To understand the potential MPR advantages, we study link scheduling in an ad hoc network with SIC at the physical layer. The fact that the links detected sequentially by SIC are correlated at the receiver poses key technical challenges. We characterize the link dependence and propose simultaneity graph (SG) to capture the effect of SIC. Then interference number is defined to measure the interference of a link. We show that scheduling over SG is NP-hard and the maximum interference number bounds the performance of maximal greedy schemes. An independent set based greedy scheme is explored to efficiently construct a maximal feasible schedule. Moreover, with careful selection of link ordering, we present a scheduling scheme that improves the bound. The performance is evaluated by both simulations and measurements in testbed. The throughput gain is on average 40% and up to 120% over IEEE 802.11. The complexity of SG is comparable with that of conflict graph, especially when the network size is not large.

Successive interference cancellation (SIC) is an effective way of multipacket reception to combat interference. We study link scheduling under SINR (Signal to Interference Noise Ratio) model in ad hoc networks with SIC at physical layer. The facts that interference is accumulated and the links decoded sequentially by SIC are correlated pose key technical challenges. We propose conflict set graph (CSG) to characterize the interference and define interference degree to measure the interference of a link. As scheduling over CSG is NP-hard, independent set based greedy scheme is explored to efficiently construct maximal feasible schedule. The performance is evaluated by simulations. As compared to the simple greedy method, the throughput gain is on average 30% and up to 60%.

Successive interference cancellation (SIC) is an effective technique of multipacket reception to combat interference. As not all collision are resolvable, careful transmission coordination is required. We study link scheduling in wireless networks with SIC at the physical layer. A new model, simultaneity graph (SG), is proposed to characterize the link correlation introduced by SIC. Then two new scheduling schemes are presented: 1) a slot-oriented scheme which assigns a maximal feasible link set to a time slot and 2) a link-oriented scheme which assigns each link a sufficient number of slots. The performance is evaluated by simulations and the results demonstrate that the throughput gain is on average 50% and up to 110% over IEEE 802.11. The complexity of SG is only a bit higher than that of the available widely-used models (e.g., conflict graph).

Given the multi-rate option in most 802.11 standards, rate adaptation based on dynamic channel condition is crucial to the system performance. This paper presents an efficient rate adaptation mechanism at MAC layer, called ERA, to improve the transport layer throughput. Such scheme raises the transmission rate aggressively, e.g., upon only a few consecutive successful transmissions. In opposite, we reduce the transmission rate more cautiously only when the station encounters severe consecutive frame losses or relatively high loss ratio at MAC layer. ERA also attempts to protect the performance under steady condition without losing the ability to respond to the channel variation quickly. We evaluate ERA using NS-2 simulations and find that it performs better than or at least similarly with many other proposals in various scenarios. ERA is a sender-based method and requires minor modification at sender side only, thus completely standard-compliant and very easy to be implemented and deployed.

Link bandwidth utilization is obviously critical for numerous network management tasks. Using the flow-conservation law, we could reduce the number of activated monitor agents. The problem of efficiently monitoring link-bandwidth based on flow-conservation law could be reduced to weak vertex cover problem, which is NP-hard. In this paper, we demonstrate an approximation preserving reduction from the vertex cover problem to weak vertex cover problem. Due to this reduction, it follows that it is very difficult to get an approximation algorithm with approximation ratio lower than 2 for weak vertex cover problem. Using the primal-dual method, we give an approximation algorithm with approximation ratio 2 to solve the problem. The effectiveness of our monitoring algorithm is validated by simulations evaluation over a wide range of network topologies. We also demonstrate the problem of weak vertex cover with blackout vertices could be reduce to weak vertex cover problem. Hence we could use the approximation algorithms for weak vertex cover problem to solve the problem of weak vertex cover with blackout vertices.