Presents a "mixture-of-experts" (MOE) approach to develop customized electrocardiogram (EGG) beat classifier in an effort to further improve the performance of ECG processing and to offer individualized health care. A small customized classifier is developed based on brief, patient-specific ECG data. It is then combined with a global classifier, which is tuned to a large ECG database of many patients, to form a MOE classifier structure. Tested with MIT/BIH arrhythmia database, the authors observe significant performance enhancement using this approach.

A novel sensor network source localization method based on acoustic energy measurements is presented. This method makes use of the characteristics that the acoustic energy decays inversely with respect to the square of distance from the source. By comparing energy readings measured at surrounding acoustic sensors, the source location during that time interval can be accurately estimated as the intersection of multiple hyperspheres. Theoretical bounds on the number of sensors required to yield unique solution are derived. Extensive simulations have been conducted to characterize the performance of this method under various parameter perturbations and noise conditions. Potential advantages of this approach include low intersensor communication requirement, robustness with respect to parameter perturbations and measurement noise, and low-complexity implementation.

Generative adversarial networks (GANs) have achieved many excellent results in hyperspectral image (HSI) classification in recent years, as GANs can effectively solve the dilemma of limited training samples in HSI classification. However, due to the class imbalance problem of HSI data, GANs always associate minority-class samples with fake label. To address this issue, we first propose a semisupervised generative adversarial network incorporating a transformer, called HyperViTGAN. The proposed HyperViTGAN is designed with an external semisupervised classifier to avoid self-contradiction when the discriminator performs both classification and discrimination tasks. The generator and discriminator with skip connection are utilized to generate HSI patches by adversarial learning. The proposed HyperViTGAN captures semantic context and low-level textures to reduce the loss of critical information. In addition, the generalization ability of the HyperViTGAN is improved through the use of data augmentation. Experimental results on three well-known HSI datasets, Houston 2013, Indian Pines 2010, and Xuzhou, show that the proposed model achieves competitive HSI classification performance in comparison with the current state-of-the-art classification models.

We present a novel, robust estimation method to distinguish salient objects from complicated, dynamic backgrounds in videos. In this method, we propose a novel approach to model motion energy based on motion magnitude, motion orientation, gradient flow field, and spatial gradient of the video frame. Furthermore, an effective spatiotemporal objectness map is also proposed to estimate a compact object-like region in the current video frame leveraging both the objectness proposals and the saliency map of the previous frame. Then the current video frame is oversegmented into the granularity of superpixels using the simple linear iterative clustering algorithm. Each superpixel is designated as a node of a graph. The similarity between adjacent superpixels will be assigned as the weight of an edge that connects these two nodes. The feature values of motion energy and spatiotemporal objectness within each superpixel will be averaged respectively, and used to graphically cluster similar superpixels to form the detected salient object. Extensive experiments comparing this proposed new method against twelve existing salient object detection (SOD) methods have been performed using the benchmark datasets unconstrained video saliency detection and densely annotated video segmentation. Superior performance of this proposed SOD method has been observed through three well-known performance metrics: precision-recall curves, F-measure curves, and the mean absolute error.

A cross-layer design methodology that jointly optimizes the link-layer network flow rate and the physical layer transmission power wherein rate is developed for energy-constrained underwater wireless sensor networks. The objective of the design is to maximize the lifetime of the network operation for a given network topology and a bound of the total transmission time of all links. Initially, each node is set to transmit at the most energy-efficient transmission rate, which is derived using an underwater acoustic channel model. A constrained least square problem then is formulated which yields an analytical solution of the optimal network flow at each link that maximizes the network lifetime. If the timing constraint is not met, a new iteration will begin with the restriction on per-node energy consumption incrementally relaxed. Usually, the algorithm converges within a couple of iterations. Two design examples, a linear configured network, and a rhombus configured network are used to demonstrate the design procedures. Extensive simulations have been performed and superior performance (longer network lifetime) of this proposed design has been observed.

In this paper, a multi-image encryption scheme based on compressed sensing (CS) and deep learning in the optical gyrator domain is proposed. Firstly, multiple plaintext images are compressed by CS to obtain multiple measurements, and then the pixels of each measurement are scrambled by using a chaotic system. Secondly, the scrambled measurements are combined into a matrix and diffused by XOR operation with a chaotic matrix. Finally, the diffused matrix is encoded with a random phase and an optical gyrator transform to obtain a complex-valued matrix, and the amplitude of the complex-valued matrix is taken as the ciphertext. In decrypt, plaintext images are reconstructed from the CS measurements by a neural network, which achieves high reconstruction speed and quality compared with the traditional algorithm. Especially, the data amount of plaintext images can be compressed by up to 8 times while achieving high decryption quality. To our best knowledge, CS reconstruction algorithms based on deep learning is firstly used for image encryption. Moreover, the proposed scheme is highly robust against occlusion, noise, and chosen-plaintext attack.