This paper presents BatchQR, a mobile APP that can batch read densely arranged QR codes attached to caps of the tubes and vials in clinical and biological labs. BatchQR could work in two modes: photo mode and preview mode, which are respectively suitable for scenarios that favor one-time batch processing capacity and real-time tracking performance. For the photo mode, we first propose an IFFT based lightweight code detection mechanism, which can adaptively adjust operating parameters to identify densely arranged QR codes in practice. We propose an image refocus mechanism to deal with blurs/distorts that may appear in the photo, and a lightweight learning based classifier to filter out falsely detected QR codes. We further optimize BatchQR by enabling batch reading in the preview mode of the camera, which is more in line with common usage habits. To this end, we develop a QR code tracking algorithm based on Kalman filtering, which keeps track of each code image dynamically. We also design a parallel acceleration mechanism based on multi-core CPU and GPU, which significantly mitigates the end-to-end delay. Comprehensive experimental results show that BatchQR can read up to 160 Version 1-H QR codes in one shot, with 95 percent accuracy and 100-400ms delay, which is only 0.1 percent of the time consumed by the regular QR code reader in the same situation.

In this article, we investigate the performance bottleneck of existing deep learning (DL) systems and propose DLBooster to improve the running efficiency of deploying DL applications on GPU clusters. At its core, DLBooster leverages two-level optimizations to boost the end-to-end DL workflow. On the one hand, DLBooster selectively offloads some key decoding workloads to FPGAs to provide high-performance online data preprocessing services to the computing engine. On the other hand, DLBooster reorganizes the computational workloads of training neural networks with the backpropagation algorithm and schedules them according to their dependencies to improve the utilization of GPUs at runtime. Based on our experiments, we demonstrate that compared with baselines, DLBooster can improve the image processing throughput by 1.4× - 2.5× and reduce the processing latency by 1/3 in several real-world DL applications and datasets. Moreover, DLBooster consumes less than 1 CPU core to manage FPGA devices at runtime, which is at least 90 percent less than the baselines in some cases. DLBooster shows its potential to accelerate DL workflows in the cloud.

Named Data Network (NDN) is an alternative to host-centric networking exemplified by today's Internet. One key feature of NDN is in-network caching that reduces access delay and query overhead by caching popular contents at the source as well as at a few other nodes. Unfortunately, in-network caching suffers various privacy risks by different attacks, one of which is termed timing attack. This is an attack to infer whether a consumer has recently requested certain contents based on the time difference between the delivery time of those contents that are currently cached and those that are not cached. In order to prevent the privacy leakage and resist such kind of attacks, we propose a detection scheme by adopting Long Short-term Memory (LSTM) model. Based on the four input features of LSTM, cache hit ratio, average request interval, request frequency, and types of requested contents, we timely capture more important eigenvalues by dividing a constant time window size into a few small slices in order to detect timing attacks accurately. We have performed extensive simulations to compare our scheme with several other state-of-the-art schemes in classification accuracy, detection ratio, false alarm ratio, and F-measure. It has been shown that our scheme possesses a better performance in all cases studied.

In recent years, deep learning (DL) has prospered again due to improvements in both computing and learning theory. Emerging studies mostly focus on the acceleration of refining DL models but ignore data preprocessing issues. However, data preprocessing can significantly affect the overall performance of end-to-end DL workflows. Our studies on several image DL workloads show that existing preprocessing backends are quite inefficient: they either perform poorly in throughput (30% degradation) or burn too many (>10) CPU cores. Based on these observations, we propose DLBooster, a high-performance data preprocessing pipeline that selectively offloads key workloads to FPGAs, to fit the stringent demands on data preprocessing for cutting-edge DL applications. Our testbed experiments show that, compared with the existing baselines, DLBooster can achieve 1.35×~2.4× image processing throughput in several DL workloads, but consumes only 1/10 CPU cores. Besides, it also reduces the latency by 1/3 in online image inference.

This paper presents BatchQR, a mobile APP that can batch read the densely arranged QR codes attached to caps of the tubes and vials in clinical and biological labs. The basic idea of BatchQR is to detect each code in the image and then decode in an one-by-one manner. However, the unique characteristics of the QR code and the application scenario bring technical challenges: First, off-the-shelf lightweight object detection mechanisms are unable to distinguish those densely arranged codes that are highly similar to each other; second, the focus area of the camera is limited, which blurs or distorts parts of the image. To this end, we propose a lightweight code detection mechanism, which can adaptively adjust operating parameters to identify densely arranged QR codes in practice. We also propose a simple but effective image refocus mechanism, which takes an auto-focused image and multiple refocused ones, and then replaces the blurred or distorted code parts with the high-quality counterparts in the refocused images. Comprehensive experimental results show that BatchQR can read 160-180 Version 1-L QR codes in batch with 90%-95% accuracy in 10-14s, which is only 4% of the time consumed by the regular QR code reader in the same situation.

Facial paralysis makes patients lose their facial movements, which can incur eye damage even blindness since patients are incapable of blinking. We design and implement a pair of smart glasses iBlink to assist facial paralysis patients to blink. The basic idea is to monitor the normal side of the face with a camera and stimulate the paralyzed side, so that the blink of the both eyes become symmetric. Our contributions are: First, we propose an eye-blink detection mechanism based on support vector machine (SVM), which can detect asymmetric blinks of patients under various illumination conditions with an accuracy above 99 percent. Our eye-image library for training the model is published online for further related studies, which contains more than 30,000 eye images. Second, we design and implement an automatic stimulation circuits to generate electrical impulse for stimulating the patient's facial nerve branches, which can configure operational parameters in a self-adaptive manner for different patients. Third, we implement the entire iBlink system, which integrates the two functions above and a communication function module for tele-medicine applications. We conduct experiments in a hospital to obtain the design basis and verify effectiveness of our device.

Recently proposed Wi-Fi indoor localization systems utilizing channel state information (CSI) derived from the received signal achieve admirable accuracy. This paper presents an information-theoretical analysis in the received waveform level to explore the fundamental limits of the approach. In particular, we perform frequency domain Cramer-Rao bound (CRB) analysis for location estimation with CSI. Our analysis resolves the high-rank Fisher information matrix challenge, and establishes intrinsic connection between parameters to be estimated for localization and the received waveform information observable with the CSI retrieving toolkit. We also analyze the influence of the asynchronization between the transmitter and the receiver to the performance bound of the CSI approach. Moreover, we shed light on the insight into the design of the CSI localization systems. In particular, we show that the CSI approach presents varying performance for localizing targets in different distances and directions with respect to the CSI retrieving anchor point (AP), and the geometric distribution of AP antennas could fundamentally influence the localization performance. Comprehensive experimental results are demonstrated to validate our theoretical analysis.

Recent study presents a fundamental limit of the RSS fingerprinting based indoor localization. In this paper, we theoretically show that the temporal correlation of the RSS can further improve accuracy of the fingerprinting localization. In particular, we construct a theoretical framework to evaluate how the temporal correlation of the RSS can influence reliability of location estimation, which is based on a newly proposed radio propagation model considering the time-varying property of signals from Wi-Fi APs. The framework is then applied to analyze localization in the one-dimensional physical space, which reveals the fundamental reason why localization performance can be improved by leveraging temporal correlation of the RSS. We extend our analysis to high-dimensional scenarios and mathematically depict the boundaries in the RSS sample space, which distinguish one physical location from another. Moreover, we develop an algorithm to utilize temporal correlation of the RSS to improve the location estimation accuracy, where the process for choosing key design parameters are provided through experiments. Experiment results show that the localization reliability and accuracy can be improved by up to 13 and 30 percent with appropriate leveraging the RSS temporal correlation information.

Facial paralysis is a disease caused by nerve damage, which can make patients lose facial movements. Facial paralysis patients usually have muscles on one side of the face noticeably droop, which seriously impacts the person's quality of life as shown in Fig. [skull]. Worse still, the eye on the affected side is unable to blink and will become dry and infected by debries, which can incur eye damage even blindness. To the best of scientists' knowledge, the paralysis is due to the pressure incurred by infection in the tunnel containing main trunk of facial nerves, where the tunnel is inside of the people's head termed as the Facial canal. In this demo, we present iBlink [1], a novel system to help paralysis patients to blink. Paralysis usually occurs in just one side of the face, and clinical trials show that electrical stimulation could trigger blink. Based on such observations, the basic idea of iBlink is to monitor the normal side of the face with a camera and stimulate the paralysed side, so that eye-movements of the both sides become symmetric.

Facial paralysis makes patients lose their facial movements, which can incur eye damage even blindness since patients are incapable of blinking. The paralysis usually occurs on just one side of the face, and clinical trials show that electrical stimulation could trigger blink. Based on such observations, we design and implement a pair of smart glasses iBlink to assist facial paralysis patients to blink. The basic idea is to monitor the normal side of the face with a camera and stimulate the paralysed side, so that the blink of the both eyes become symmetric. To the best of our knowledge, this is the first piece of wearable device for facial paralysis therapy. Our contributions are: First, we propose an eye-blink detection mechanism based on deep convolutional neural network (CNN), which can detect asymmetric blinks of patients under various illumination conditions with an accuracy above 99%. Our eye-image library for training CNN models is published online for further related studies, which contains more than $30,000$ eye images. Second, we design and implement an automatic stimulation circuits to generate electrical impulse for stimulating the patient's facial nerve branches, which can configure operational parameters in a self-adaptive manner for different patients. Third, we implement the entire iBlink system, which integrates the two functions above and a communication function module for tele-medicine applications. Moreover, we conduct clinical trials in a hospital, in order to obtain the design basis and verify effectiveness of our device.

Indoor localization for quadrotors has attracted much attention recently. While efforts have been made to perform location estimation of quadrotors leveraging dedicated indoor infrastructures, the low-cost and commonly used RSS fingerprinting based approach utilizing existing Wi-Fi APs has yet to be applied. The challenge is that the high-speed flight reduces the RSS measuring opportunities for fingerprints comparison; moreover, the 3D space fingerprints collection incurs more overhead than in the traditional 2D case. In this paper, we present HiQuadLoc, a RSS fingerprinting based indoor localization system for quadrotors. We propose a series of mechanisms including path estimation, path fitting, and location prediction to deal with the negative influence incurred by the high-speed flight; moreover, we develop a 4D RSS interpolation algorithm to reduce the site survey overhead, where 3D is for the indoor physical space and 1D is for the RSS sample space. Experimental results demonstrate that HiQuadLoc reduces the average location error by more than 50 percent compared with simply applying the RSS fingerprinting based approach for 2D localization, and the overhead of RSS training data collection is reduced by more than 80 percent.