Incorporating deep neural networks in image compressive sensing (CS) receives intensive attentions in multimedia technology and applications recently. As deep network approaches learn the inverse mapping directly from the CS measurements, the reconstruction speed is significantly faster than the conventional CS algorithms. However, for existing network-based approaches, a CS sampling procedure has to map a separate network model. This may potentially degrade the performance of image CS with block-wise sampling because of blocking artifacts, especially when multiple sampling rates are assigned to different blocks within an image. In this paper, we develop a multi-channel deep network for block-based image CS by exploiting inter-block correlation with performance significantly exceeding the current state-of-the-art methods. The significant performance improvement is attributed to block-wise approximation but full-image removal of blocking artifacts. Specifically, with our multi-channel structure, the image blocks with a variety of sampling rates can be reconstructed in a single model. The initially reconstructed blocks are then capable of being reassembled into a full image to improve the recovered images by unrolling a hand-designed block-based CS recovery algorithm. Experimental results demonstrate that the proposed method outperforms the state-of-the-art CS methods by a large margin in terms of objective metrics and subjective visual image quality. Our source codes are available at https://github.com/siwangzhou/DeepBCS.

Taxi demand prediction plays a significant role in assisting the pre-allocation of taxi resources to avoid mismatches between demand and service, particularly in the era of the sharing economy and autonomous driving. However, most studies have only tried to figure out the complex spatial-temporal pattern of taxi demand from historical taxi demand series, neglecting the intrinsic influences of regional functions, and failing to effectively capture the dynamic long-term periodicity. In this paper, we make two important observations: (1) taxi demand pattern varies significantly between different functional regions; and (2) taxi demand follows a dynamic daily and weekly pattern. To address these two issues, we adopt Points of Interest (POIs) to identify regional functions, and propose a novel BERT-based Deep Spatial-Temporal Network (BDSTN) to model the complex spatial-temporal relations from heterogeneous local and global features. In BDSTN, a Spatiotemporal Pattern Matching module is introduced to capture the complex spatiotemporal pattern of taxi demand while considering its dynamic temporal periodicity, and a Functional Similarity Embedding module is adopted to learn the functional similarity among all regions via POIs. To the best of our knowledge, this is the first work to use BERT-based architecture to learn taxi demand patterns, and is also the first to take functional similarity represented by POIs into consideration. Our experimental results on real-world traffic datasets in New York City demonstrate that the effectiveness of the proposed method outperforms the state-of-the-art methods, and that the efficiency of our proposed model is higher than other deep learning methods.

A number of image compressive sensing (CS) algorithms were proposed in the past two decades, aiming at yielding recovered images with the best possible visual effect. However, it is quite difficult to further improve the image quality for human eyes. For example, in the low-rate sampling scenarios, CS algorithms always suffer degraded performance and can only recover less visually appealing images. We notice that what human beings concern with is the visual quality of an image, while machine users care much more about its latent metrics, such as recognition accuracy, rather than the subjective visual effect. Inspired by this point, we develop a machine recognition-oriented image CS with an adversarial learning strategy. Some adversarial models are investigated to make the recognition accuracy as an additional optimization goal of the CS reconstruction network. Through end-to-end training, CS reconstruction network automatically learns an image recognition pattern, and produce recovered images owning extra recognition metric, which makes them become more suited for machine users. Experimental results indicate that the images recovered with the proposed adversarial learning strategy can be recognized with significantly higher accuracy compared to that with the existing CS algorithms.

Channel state information based activity recognition has gathered immense attention over recent years. Many existing works achieved desirable performance in various applications, including healthcare, security and Internet of Things, with different machine learning algorithms. However, they usually fail to consider the availability of enough samples to be trained. Besides, many applications only focus on the scenario where only single subject presents. To address these challenges, in this paper, we propose a three-phase system Wi-multi that targets at recognizing multiple human activities in a wireless environment. Different system phases are applied according to the size of available collected samples. Specifically, distance-based classification using Dynamic Time Warping is applied when there are few samples in the profile. Then, Support Vector Machine is employed when representative features can be extracted from training samples. Lastly, recurrent neural networks is exploited when a large number of samples are available. Extensive experiments results show that Wi-multi achieves an accuracy of 96.1% on average. It is also able to achieve a desirable tradeoff between accuracy and efficiency in different phases.

Past years have witnessed the rapid increasing number of smart devices and objects deployed in the urban environment. Leveraging helpful data generated by hundreds of millions of smart objects, a large number of services in the Internet of Things (IoT) are devised and developed to improve our urban life quality. However, uploading the unprecedented volume of sensing data from IoT sensors to the cloud directly can lead to huge unnecessary consumption and hurt the quality of IoT services. This work leverages the fog architecture to devise a task-driven data offloading algorithm in urban IoT services. Specifically, a three-layer urban IoT service architecture is proposed, and the Task-driven Data Offloading (TDO) process is formulated as a combination optimization problem taking task deadlines and abilities of fog devices into consideration. Then, we prove the TDO problem is NP-hard, and the algorithm is devised to solve it with a careful designed utility function. Also, we propose algorithm to improve the algorithm considering the overlaps of tasks. Finally, we demonstrate the significant performance of the proposed algorithms with extensive evaluations based on real-world dataset.

Low-density parity-check (LDPC) codes exhibit excellent error correcting capability. However, small trapping sets in the Tanner graph are harmful to the iterative decoding algorithm. In this letter, we present a method of constructing (3, n) girth-eight quasi-cyclic LDPC codes with low error floor by removing the small trapping sets from the Tanner graph. To address this issue, we analyze the relationship between eight-cycles and small trapping sets of Tanner graphs based on fully connected base graphs without parallel edges. We find that if some eight-cycles are not found in the Tanner graphs, any elementary trapping set in the range of a ≤ 8 and b ≤ 3 is removed naturally. We also derive a lower bound on the permutation size for the construction of such codes. The experimental simulation shows a favorable error rate performance with lower error floor over additive white Gaussian noise channels.