Obtaining a high level of situation awareness while maintaining optimal utilization of resources is becoming increasingly important, especially in the context of asymmetric warfare, where information superiority is crucial for maintaining the edge over the opponent. Obtaining an adequate level of situational information from an ISR system is dependent on sensor capabilities as well as the ability to cue the sensors appropriately based on the current information needs and the ability to utilize the collected data with suitable data processing methods. Applying the Data to Decision approach for managing the behavior of sensor systems facilitates optimal use of sensor assets while providing the required level of situational information. The approach presented in the paper combines the Data to Decision approach with the Fog Computing paradigm, where the computation is pushed to the edge of the network. This allows to take advantage of Big Data potentially generated by the sensor systems while keeping the resource requirements in terms of bandwidth manageable. We suggest a System of Systems approach for assembling the ISR system, where individual systems have a high level of autonomy and the computational resources to perform the necessary computation tasks. To facilitate a composition of a System of Systems of sensors for tactical applications the proactive middleware ProWare is applied. The work presented in the paper has been conducted as part of the European Defense Agency project IN4STARS, in the context of which an implementation of a sensor solution is being built, which facilitates on-line sensor cueing and collaboration between sensors by building upon the Fog Computing paradigm and utilizing the Data to Decision concepts.

This paper describes the Speech Technology Center (STC) system for the 5th CHiME challenge. This challenge considers the problem of distant multi-microphone conversational speech recognition in everyday home environments. Our efforts were focused on the single-array track, however, we participated in the multiple-array track as well. The system is in the ranking A of the challenge: acoustic models remain frame-level tied phonetic targets, lexicon and language model are not changed compared to the conventional ASR baseline. Our system employs a combination of 4 acoustic models based on convolutional and recurrent neural networks. Speaker adaptation with target speaker masks and multi-channel speaker-aware acoustic model with neural network beamforming are two major features of the system. Moreover, various techniques for improving acoustic models are applied, including array synchronization, data cleanup, alignment transfer, mixup, speed perturbation data augmentation, room simulation, and backstitch training. Our system scored 3rd in the single-array track with Word Error Rate (WER) of 55.5% and 4th in the multiple-array track with WER of 55.6% on the evaluation data, achieving a substantial improvement over the baseline system.

The application of wireless sensor networks (WSN) for the task of acoustic localization provides great opportunities for distributed cooperative tracking of sound sources in large areas. However WSNs are significantly more limited in terms of computational resources and power than typical computer systems. Therefore the methods applied for acoustic localization in WSN must be optimized for minimal resource consumption. This paper builds on the advances of Steered Response Power with Phase Transform (SRP-PHAT) optimization and proposes a further simplification in terms of additional minimization of the initial search volume. By using several linear microphone arrays we are able to estimate the initial region of sound source and reduce the number of computations by at least one order of magnitude. The results of several experiments on real signals confirm the achieved improvements.

This paper considers an autonomous ground Intelligence, Surveillance and Reconnaissance (ISR) system comprising of multiple distributed, wirelessly communicating smart sensors. The ISR system, in turn, is a part of a larger System of Systems (SoS) consisting of aerial, manned, etc. surveillance systems and information collection centers. The smart sensors of the ISR system perform environment monitoring using different modalities and exchange object detection and identification results to assess the situation and provide other SoS components with this information. In the paper we discuss using acoustic, magnetic and Passive Infrared (PIR) sensor information for target detection and identification. We also propose an approach of distributed acoustic source localization and a method of velocity estimation using PIR data. For sensor communication an asynchronous adhoc WSN configuration is proposed. The system is implemented on low power smart sensors utilizing Atmel ATmega128RFA1 processors with integrated 2.4GHz IEEE 802.15.4 compliant radio transceivers.

Computing on the edge of the Internet of things comprises among other tasks in-sensor signal processing and performing distributed data fusion and aggregation at network nodes. This poses a challenge to distributed sensor networks of low computing power devices that have to do complex fusion, aggregation and signal processing in situ. One of the difficulties lies in ensuring validity of data collected from heterogeneous sources. Ensuring data validity, for example, the temporal and spatial correctness of data, is crucial for correct in-network data fusion and aggregation. The article considers wireless sensor technology in military domain with the aim of improving situation awareness for military operations. Requirements for contemporary intelligence, surveillance and reconnaissance applications are explored and an experimental wireless sensor network, designed to enhance situation awareness to both in-the-field units and remote intelligence operatives, is described. The sensor nodes have the capability to perform in-sensor signal processing and distributed in-network data aggregation and fusion complying with edge computing paradigm. In-network data processing is supported by service-oriented middleware which facilitates run-time sensor discovery and tasking and ad hoc (re)configuration of the network links. The article describes two experiments demonstrating the ability of the wireless sensor network to meet intelligence, surveillance and reconnaissance requirements. The efficiency of distributed data fusion is evaluated and the importance and effect of establishing data validity is shown.

No abstract

No abstract

Situation awareness is an important aspect of ubiquitous computer systems, as these systems of systems are highly integrated with the physical world and for successful operation they must maintain high awareness of the environment. Acoustic information is one of the most popular modalities, by which the environment states are estimated. Multi-sensor approaches also provide the possibility for acoustic source localization. This paper considers an acoustic localization system of dual channel smart sensors interconnected through a Wireless Sensor Network (WSN). The low computational power of smart sensor devices requires distribution of localization tasks among WSN nodes. The Initial Search Region Reduction method is used in the WSN to meet this requirement. The system is implemented on smart dust motes utilizing Atmel ATmega128RFA1 processors with integrated 2.4GHz IEEE 802.15.4 compliant radio transceivers. The paper discusses complications, introduced by low power hardware, and reviews conditions of real-time operation.