“Collaborative Signal Processing for Efficient Wireless Sensor Networks”, 09/2006 – 08/2010
Funded by: European Commission, FP7, Marie Curie Transfer of Knowledge (ToK-DEV).
Partners: FORTH-ICS (Coordinator: P. Tsakalides), University of Valencia, University of Southern California (USA).
Funding: €1,256,423 (FORTH-ICS: € 1,256,423)
SPL Contact Person: Prof. Panagiotis Tsakalides.
Summary: Wireless sensor networks (WSN) have been an emerging technology whose goal is to monitor the physical world by means of a densely distributed network of wireless sensor nodes. It will soon become feasible to deploy massive amounts of inexpensive devices to observe large ground surfaces, underwater regions, and areas in the atmosphere. These devices, integrated with miniature power supply, multiple modality sensors, on-board processors, and radio communication modules, will be capable of forming a large-scale ad hoc WSN. If judiciously and successfully deployed, they will be able to provide, to the benefit of society and several scientific communities (e.g. environmental science), unprecedented opportunities for instrumenting and controlling the environment, our cities, both our microcosm and macrocosm.
Realising the potential of large, distributed sensor networks requires major advances in the theory, fundamental understanding and practice of distributed data processing, self-organised communications, and information fusion in highly uncertain scenarios using sensing/communications nodes that are severely constrained in power, computation, and communication capabilities.
The ASPIRE project is pushing the boundaries of wireless sensor networks. By approaching traditional, entrenched problems from innovative new angles, the project is breaking through the ‘wireless sensor node’ glass ceiling. During the period 2006-2010, the ASPIRE Group has been active and productive in original research in 5 major axes, namely: (i) Distributed signal classification for wireless sensor networks; (ii) Compressive sensing (CS) and its applications; (iii) Multichannel audio coding and transmission; (iv) Non-Gaussian modeling and multiscale Bayesian processing for various signal modalities; and (v) Wireless network traffic modeling and localization in WSNs.
The role of SPL:
Overall management of the project, including dissemination and exploitation activities.
Conduct fundamental and applied research on statistical signal processing, compressed sensing theory and audio coding.
- Extend and explore the concept of Compressed Sensing to emerging application areas related to distributed signal processing and learning.
Griffin, Anthony; Hirvonen, Toni; Tzagkarakis, Christos; Mouchtaris, Athanasios; Tsakalides, Panagiotis, Single-Channel and Multi-Channel Sinusoidal Audio Coding Using Compressed Sensing, Audio, Speech, and Language Processing, IEEE Transactions on, 19 (5), pp. 1382-1395, 2011, ISSN: 1558-7916.
Tzagkarakis, Christos; Mouchtaris, Athanasios; Tsakalides, Panagiotis, A Multichannel Sinusoidal Model Applied to Spot Microphone Signals for Immersive Audio, Audio, Speech, and Language Processing, IEEE Transactions on, 17 (8), pp. 1483-1497, 2009, ISSN: 1558-7916.
Tzagkarakis, George; Papadopouli, Maria; Tsakalides, Panagiotis, Trend Forecasting Based on Singular Spectrum Analysis of Traffic Workload in a Large-scale Wireless LAN, Perform. Eval., 66 (3-5), pp. 173–190, 2009, ISSN: 0166-5316.
Tzagkarakis, George; Beferull-Lozano, Baltasar; Tsakalides, Panagiotis, Rotation-Invariant Texture Retrieval via Signature Alignment Based on Steerable Sub-Gaussian Modeling, Image Processing, IEEE Transactions on, 17 (7), pp. 1212-1225, 2008, ISSN: 1057-7149.
Mouchtaris, Athanasios; Karadimou, Kiki; Tsakalides, Panagiotis, Multiresolution Source/Filter Model for Low Bitrate Coding of Spot Microphone Signals, EURASIP J. Audio Speech Music Process., 2008 , pp. 1:1–1:16, 2008, ISSN: 1687-4714.
- Applied the concept of Support Vector Machines for addressing the problem of distributed classification in WSN, and yielded classes of (a) incremental and (b)gossip-based distributed consensus algorithms for training the classifier.
- Transfer the theoretical framework fo Compressed Sensing to emerging WSN applications related to: (a) data estimation and information retrieval, based on heavy-tailed (alpha-stable) multivariate distributions, (b) distributed signal reconstruction, by the means of introducing a Bayesian matching pursuit method.
- Apply the framework of Compress Sensing to process the harmonic part of a sinusoidally-modeled audio signal to encode audio streams with high-quality at low bitrates.
- Designed and tested coding schemes based on multichannel audio-specific models, namely: (a) the source / filter model and (b) the sinusoidal model, thereby allowing flexible manipulation and high-quality, low-bitrate encoding.
- Accurate characterization of the sparsity of the wavelet coefficients by using alpha-stable statistical modeling, synthesis of novel Bayesian algorithms based on the more accurate models, and validation to various applications domains, including: (a) ranging from image processing (retrieval, fusion, and watermarking); (b) SAR image denoising and autofocus; (c) underwater acoustic signal classification; (d) biomedical (ultrasound, microarray, miRNA) signal enhancement and classification.
- Proposed novel localization techniques based on multivariate (Gaussian and non-G models) of the signal strength measurements collected from several access points (APs) at different locations, and considered compressed sensing (CS) for signal reconstruction and channel estimation in OFDM-based high-rate ultra wideband (UWB) communication systems.