This paper presents a self-supervised feature learning method for hyperspectral image classification. Our method tries to construct two different views of the raw hyperspectral image through a cross-representation learning method. And then to learn semantically consistent representation over the created views by contrastive learning method. Specifically, four cross-channel-prediction based augmentation methods are naturally designed to utilize the high dimension characteristic of hyperspectral data for the view construction. And the better representative features are learned by maximizing mutual information and minimizing conditional entropy across different views from our contrastive network. This 'Cross-View-Predicton' style is straightforward and gets the state-of-the-art performance of unsupervised classification with a simple SVM classifier.
Flow classification is a key technique in network encryption machine, which matches 5-tuples of IP packet against a predefined rule-set to decide the subsequent processing, such as secret transmission\clear transmission\discard. Currently, there is no flow classification algorithm that can support both 10Gbps line-rate range matching and rule set real-time modification on the CPU. We solve this problem through FPGA and CPU cooperation. We perform range matching on the first packet of the data flow on the FPGA, and create a flow table entry on the CPU side. Then perform the flow table hash matching on the subsequent packets on the CPU, and design the synchronous updating scheme of the rule set and the flow table. The mechanism can significantly reduce the computational and communication overhead.
Sign language is a bridge for communication be-tween people with normal hearing and impaired hearing. How-ever, only a few people with normal hearing have an understanding of sign language. Thus, Sign Language Recognition (SLR) has attracted significant interest from both industry and academia. Prior traditional methods have certain limitations such as light condition and extra hardware. To address these issues, this study aims to recognize sign languages using acoustic signals on a smartphone. We design an SLR system called SignGest, which captures user's sign language gestures with built-in microphones. Afterwards, we build a Convolutional Neural Network(CNN) model to extract features of different gestures to distinguish them. Furthermore, we use a Deep Convolutional Generative Adversarial Network(DCGAN) to generate abundant training data that look no different from the real samples. We handle various challenges including effective gesture segmentation and training data collection. Finally, SignGest is implemented on a server and an Android smartphone with built-in microphones and speakers without any extra hardware or infrastructure equipment. When users perform the sign language gestures before the smartphone, SignGest can recognize each of them and print the result. Through extensive experiments, our results show that SignGest can achieve robust and satisfactory performance.
The widespread application of specialized embedded devices and their ever-growing sensing capabilities lead to an overload of raw data reaching the Cloud via the Internet of Things. Despite this massive amount of available data, data-driven applications still produce a limited impact on the real world. We argue that this limitation arises from the complexity the developers still face when engineering and deploying solutions on a plurality of embedded devices. Instead of focusing on constructing valuable information from raw data and designing algorithms to exploit it, the developers still have to figure out on their own: how to partition the computation between Edge, Fog, and Cloud; how to move partial results around in highly distributed solutions; how to create abstract interfaces to expose information-rich endpoints. This paper introduces the Virtual Sensor Domain-Specific Language (DSL), which builds the keystone for designing and developing powerful Computer-Aided Design tools and Integrated Development Environments to support the developers in focusing on information manipulation. The Virtual Sensor DSL is built on top of C++ to ease its adoption by the embedded development community. It introduces a reactive approach to data collection, technology mapping, and workload distribution. It allows the creation of adaptable Virtual Sensors, streaming processed information coming from the integration of multiple raw data streams, and it comes with integrated simulation capabilities, supporting prototyping. In this paper, we describe the entities the language offers, how the developers can exploit them, and analyze a real-world use case that benefitted from the Virtual Sensor DSL.
In this paper, we consider an integrated terrestrial-aerial access network where multiple access points (APs) such as cellular base stations (BSs), unmanned aerial vehicles (UAVs), and high altitude platforms (HAPs) coordinate to serve various ground users. Advanced forms of typical data services for 5G and beyond such as enhanced mobile broadband (eMBB), massive machine-type communications (mMTC), and ultra-reliable low-latency communications (URLLC), are provided via network slicing (NS) to cope with the heterogeneity brought by the terrestrial-aerial integration. To satisfy various requirements of users, network slices consist of different virtual network functions (VNFs), each of which consumes different amounts of compute resource at APs. To achieve a tradeoff between data rate and consumed compute resource of VNFs, we aim to maximize the compute resource utilization efficiency by jointly optimizing the VNF deployment, power and spectrum resource allocation. The mutual relation between VNF deployment and user scheduling is modelled, based on which an iterative algorithm is proposed to solve the problem. Simulation results show that the proposed algorithm achieves a better tradeoff than benchmark ones. Influence of the VNF deployment and size on the system performance is revealed as well.
