Human activity recognition (HAR) based on sensor networks is an important research direction in the fields of pervasive computing and body area network. Existing researches often use statistical machine learning methods to manually extract and construct features of different motions. However, in the face of extremely fast-growing waveform data with no obvious laws, the traditional feature engineering methods are becoming more and more incapable. With the development of deep learning technology, we do not need to manually extract features and can improve the performance in complex human activity recognition problems. By migrating deep neural network experience in image recognition, we propose a deep learning model (InnoHAR) based on the combination of inception neural network and recurrent neural network. The model inputs the waveform data of multi-channel sensors end-to-end. Multi-dimensional features are extracted by inception-like modules by using various kernel-based convolution layers. Combined with GRU, modeling for time series features is realized, making full use of data characteristics to complete classification tasks. Through experimental verification on three most widely used public HAR datasets, our proposed method shows consistent superior performance and has good generalization performance, when compared with the state-of-the-art.

Millimeter-wave (mmWave)-based human motion sensing, such as activity recognition and skeleton tracking, enables many useful applications. However, it suffers from a scarcity issue of training datasets, which fundamentally limits a widespread adoption of this technology in practice, as collecting and labeling such datasets are difficult and expensive. This paper presents SynMotion, a new mmWave-based human motion sensing system. Its novelty lies in harvesting available vision-based human motion datasets, for knowing the coordinates of body skeletal points under different motions, to synthesize mmWave sensing signals that bounce off the human body, so that the synthesized signals could inherit labels (skeletal coordinates and the name of each motion) from vision-based datasets directly. SynMotion demonstrates the ability to generate such labeled synthesized data at high quality to address the training-data scarcity issue and enable two sensing services that can work with commercial radars, including 1) zero-shot activity recognition, where the classifier reads real mmWaves for recognition, but it is only trained on synthesized data; and 2) body skeleton tracking with few/zero-shot learning on real mmWaves. To design SynMotion, we address the challenges of both the inherent complication of mmWave synthesis and the micro-level differences compared to real mmWaves. Extensive experiments show that SynMotion outperforms the latest zero-shot mmWave-based activity recognition method. For skeleton tracking, SynMotion achieves comparable performance to the state-of-the-art mmWave-based method trained on the labeled mmWaves, and SynMotion can further outperform it for the unseen users.

Underground target detection plays an important role in maintaining underground space security and relies on efficient real-time detection methods. Electromagnetic detection methods including ground penetrating radar and electromagnetic induction have been implemented in detecting underground targets and have been shown to have bright prospects for the future. This paper presents a comprehensive review of current electromagnetic detection methods and recent advances in underground target detection. There are two main problems in electromagnetic underground target detection, namely inversion and classification. We summarize solutions to these problems based on electromagnetic methods as well as deep learning. In addition, this paper proposes a technical framework based on data acquisition and data processing to describe the application of electromagnetic methods. Finally, this paper concludes that although various methods have been proposed, there still exist several future research hotspots in the fields of soil water content, tree root system research, and underground unexploded ordnance. This paper can be a reference for researchers in underground target detection and related fields. Therefore, the research of underground target electromagnetic detection method is expected to increase significantly.

With scientific research in materials science becoming more data intensive and collaborative after the announcement of the Materials Genome Initiative, the need for modern data infrastructures that facilitate the sharing of materials data and analysis tools is compelling in the materials community. In this paper, we describe the challenges of developing such infrastructure and introduce an emerging architecture with high usability. We call this architecture the Materials Genome Engineering Databases (MGED). MGED provides cloud-hosted services with features to simplify the process of collecting datasets from diverse data providers, unify data representation forms with user-centered presentation data model, and accelerate data discovery with advanced search capabilities. MGED also provides a standard service management framework to enable finding and sharing of tools for analyzing and processing data. We describe MGED’s design, current status, and how MGED supports integrated management of shared data and services.

In Internet of Things, various things realize interconnection and intercommunication by the wireless sensor network (WSN). The time-division multiple-access (TDMA) slot scheduling algorithm for WSN is a crucial technology for improving network efficiency and prolonging network lifetime. Considering the industrial field of WSN, the problems of limited energy and changed topology restrict the development of the centralized resource allocation. Although some traditional distributed time slot allocation algorithms solve the network topology change to a certain extent, there are still some problems and challenges in terms of efficiency and energy consumption. In this paper, we propose a distributed TDMA slot scheduling algorithm based on energy and topology factor. By analyzing and summarizing the distributed randomized time slot assignment algorithm, it is found that energy and topology information of network nodes are very important for TDMA slot scheduling. Then the definition of energy-topology (E-T) factor is proposed, which is based on the influence of residual energy and topology on the time slot allocation. Moreover, the distributed TDMA slot scheduling algorithm based on E-T factor is proposed by taking the number of neighbors and residual energy of network nodes into consideration to reduce the execution time and energy consumption of algorithm. Finally, we simulate the proposed scheme to evaluate its performance. The experimental results indicate that the proposed algorithm can reduce the message complexity, time complexity, and energy consumption, and improve the efficiency of TDMA slot scheduling.