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.
13 April 2020 in IEEE Transactions on Vehicular Technology
In the past several years, unmanned aerial vehicle (UAV) have been employed to provide enhanced coverage or relay service to mobile users in a scenario with limited or even no infrastructure since they can be deployed to almost everywhere and can be manipulated at anytime. This paper studies UAV as aerial base station (BS) enabled wireless communication system, where a UAV is dispatched to provide wireless communication service to a set of ground users with difference quality-of-experience (QoE) requirements. In real world, user requirements are randomly and unevenly distributed. In addition, UAV communication coverage and on-board energy are limited and system resources are also limited (e.g., transmission power, spectrum). In order to meet the QoE of all users with limited system resources and limited UAV energy, we jointly optimize user communication scheduling, UAV trajectory, transmit power and bandwidth allocation to maximize energy-efficiency and satisfy user QoE requirement. The formulated problem is mixes integer non-convex and non-concave so it is difficult to solve. In this paper, we solve the problem with two steps as follows. Firstly, we transform the objective function into a tractable form. Secondly, we divide the optimal problem into four sub-optimal problems, and then use a powerful iterative algorithm with the Dinkelbach and block coordinate descent to solve the optimal problem. That is to say, the user scheduling, UAV trajectory, transmission power and bandwidth allocation are alternately optimized in each iteration. Extensive simulation results present that our proposed method can obtain higher energy efficiency than that of baselines. Specifically, the energy efficiency obtained by our proposed method is 12.5% higher than the approach that only maximizes throughput.
13 January 2022 in IEEE Transactions on Multimedia
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.
Pedestrian trajectory prediction is a critical technology in the evolution of self-driving cars toward complete artificial intelligence. Over recent years, focusing on the trajectories of pedestrians to model their social interactions has surged with great interest in more accurate trajectory predictions. However, existing methods for modeling pedestrian social interactions rely on pre-defined rules, struggling to capture non-explicit social interactions. In this work, we propose a novel framework named DTGAN, which extends the application of Generative Adversarial Networks (GANs) to graph sequence data, with the primary objective of automatically capturing implicit social interactions and achieving precise predictions of pedestrian trajectory. DTGAN innovatively incorporates random weights within each graph to eliminate the need for pre-defined interaction rules. We further enhance the performance of DTGAN by exploring diverse task loss functions during adversarial training, which yields improvements of 16.7% and 39.3% on metrics ADE and FDE, respectively. The effectiveness and accuracy of our framework are verified on two public datasets. The experimental results show that our proposed DTGAN achieves superior performance and is well able to understand pedestrians' intentions.
29 November 2023 in IEEE Internet of Things Journal
Real-time data collected from users can help various applications provide services, but there is a risk that sensitive information will be leaked. Existing local differential privacy (LDP)-based approaches mainly perturb each data point, which severely affects the time-series patterns, leading to sensitive information being leaked out from multiple consecutive significant patterns. In this article, we focus on dynamic data streams under honest but curious servers and propose a privacy-preserving method called PP-LDP to protect data privacy while preserving data stream patterns and improving the utility of published data. To this end, our approach consists of three main parts. First, we sample the points that can represent the data stream patterns by improving the popular least squares segmented linear fit method. Then, we use an adaptive budget allocation method to perturb the sampling points and provide $w$ -event level privacy. Finally, we perform post-processing optimization of the data streams with Kalman filters to further improve the utility of the data streams. Extensive experimental results on realistic data sets show that our proposed scheme can not only protect the data streams’ privacy but also effectively preserve patterns and guarantee the utility of private data.