Announcements
  • Mr. Kimin Kim's work on coherent detection of small and manoeuvring objects using simultaneous trajectory estimation is now submitted to IEEE Transactions on Aerospace and Electronic Systems. Preliminary results appeared in the IEEE Radar Conference 2017, in Seattle, WA USA.
  • Our work on pseudo-likelihoods for intractable estimation problems in multi-sensor state space models is under review for IEEE Transactions on Signal and Information Processing over Networks. This work has a similar perspective to but differs substantially from our previous work (open access here) in the multi-object state space model used, and, the pseudo-likelihood itself.
  • My UDRC Summer School 2017 slide+handouts on optimal and adaptive filtering of stochastic proceeses are here.
  • I visited the brilliant Sensor Informatics and Biomedical Technology Lab. directed by Prof. Simo Sarkka, at Aalto University, in the greater Helsinki region in Finland. Interesting research on probabilistic models and statistical processing methodology for spatio-temporal data and applications, and, great hospitality. Highly recommended for those interested in the field.
  • Our cooperative calibration algorithm for distributed fusion networks is now published in IEEE TSP and open access via the ieeexplore link here.
  • The MATLAB class @functs can be found at sourceforge and also reached via Matlab Central Link Exchange. I find it easy to build customized message-passing algorithms using this class.
  • Publications
    JOURNAL ARTICLES - UNDER REVIEW
    1. Murat Üney, Bernard Mulgrew, Daniel Clark, "Latent Parameter Estimation in Fusion Networks Using Separable Likelihoods," IEEE Transactions on Signal and Information Processing Over Networks, under review.
    Latent parameter estimation in multi-sensor state space models underpin many desirable capabilities such as sensor self-calibration in fusion networks. This is an intractable likelihood problem, in general, due to the inherent "association uncertainties". Here, we propose a pseudo-likelihood that scales with the number of sensors. This approximation is built upon a different state space model than that used in our TSP'16 article. Inference in this approximate model is still carried out using Belief Propagation. We provide theoretical results regarding the accuracy of this approximation and explicit formulae for sensor self-localisation in linear Gaussian multi-sensor multi-oject state space models.
    2. Kimin Kim, Murat Üney, Bernard Mulgrew, "Detection via simultaneous trajectory estimation and long time integration," IEEE Transactions on Aerospace and Electronic Systems, under review.
    This work addresses challenges in detecting small and manoeuvring objects with radars. For the case, a comparably long time window of measurements should be used which causes a combinatorial growth in the number of possible trajectories. This work sequentially estimates these trajectories using a novel Markov state space model. The measurements at each time step is the radar data cube received in a coherent processing interval. The most efficient statistical test on an arbitrarily long time window of measurements is treated as statistical inference on this model.
    JOURNAL ARTICLES - PUBLISHED
    3. Murat Üney, Bernard Mulgrew, Daniel Clark, "A cooperative approach to sensor localisation in distributed fusion networks," IEEE Transactions on Signal Processing, vol. 64, no. 5, pp. 1187-1199, March 2016 ( ieeexplore open access link).
    A pre-requisite for fusing local information in networks of sensors (e.g. radar networks) is registration/calibration. Algorithmic self-calibration using measurements from non-cooperative objects is a highly desired capability. This work addresses the computational complexity of this problem and offers a solution that leads to a unique capability. The likelihood function that assesses different calibration configurations is intractable to evaluate. This work introduces ``node-wise separable likelihoods'' as scalable surrogates for the original likelihood and theoretically proves performance guarantees using information theory.
    4. Emmanuel Delande, Murat Üney, Jeremie Houssineau, Daniel Clark, "Regional variance for multi-object filtering," IEEE Transactions on Signal Processing, vol.62, no.13, pp.3415-3428, July 2014 (open access) .
    Recent progress in multi-target tracking has led to the estimation of target populations such that the target states (e.g. position and velocity) are represented by a single object, a Random Finite Set (RFS). RFS filtering algorithms compute the number of targets in arbitrarily selected regions using the first-order moment. This work proposes the concept of regional variance which quantifies the level of confidence on target number estimates in these regions and introduces explicit formulae for its computation. This second-order statistic facilitates information-based decisions. Algorithms are provided for its computation for the probability hypothesis density (PHD) and the cardinalized probability hypothesis density (CPHD) filters.
    5. Murat Üney and Müjdat Çetin, "Optimization of Decentralized Random Field Estimation Networks under communication constraints through Monte Carlo methods," Digital Signal Processing, vol. 36, pp. 16-28, November 2014.
    This work extends our TSP'11 article on the design of in-network estimation strategies for poly-tree networks towards undirected topologies. This setting is relevant to random field estimation with distributed sensors that can perform nearest-neighbour communications over bandwidth-limited links.
    6. Murat Üney, Daniel E. Clark, Simon J. Julier, "Distributed Fusion of PHD Filters via Exponential Mixture Densities," IEEE Journal of Selected Topics in Signal Processing, special issue on multi-target tracking, vol. 7, no.3, pp. 521-531, June 2013.
    This work addresses the challenge of integrating information from geographically distributed and networked sensors. The proposed solution for the first time enabled fusion of high-level information at the population level of abstraction by extending the celebrated covariance intersection fusion algorithm of Uhlmann and Julier to RFS population distributions. Sensor nodes use RFS filters to jointly estimate the number of objects and their locations. These local results are merged by solving the variational problem underlying covariance intersection for multi-object distributions. This approach is demonstrated on real radar and camera streams in maritime surveillance with BAE Systems and UCL~(see our IMA'13 paper).
    7. Murat Üney and Müjdat Çetin, "Monte Carlo Optimization of Decentralized Estimation Networks Over Directed Acyclic Graphs Under Communication Constraints", IEEE Transactions on Signal Processing, pp. 5558-5576, vol. 59, no. 11, November 2011.
    Design of in-network processing strategies for sensor networks is challenging because the communication constraints should be taken into account. This work addresses the trade-off between the estimation accuracy (e.g. accuracy in estimating a temperature field in environmental monitoring) and the energy cost of communication. We show that the optimal estimation and communication strategy has a variational form with no closed form expressions. We develop a Monte Carlo algorithm to optimise the processing strategy numerically and demonstrate graceful degradation of estimation accuracy with the cost of communications. Preliminary results received the best runner-up award in the IEEE student paper competition at IEEE SIU'09.
    CONFERENCES
  • Kimin Kim, Murat Üney, Bernard Mulgrew "Simultaneous tracking and long time integration for detection in collaborative array radars," IEEE Radar Conference 2017, Seattle, May 2017.
  • Kimin Kim, Murat Üney, Bernard Mulgrew "Detection of manoeuvring low SNR objects in receiver arrays," SSPD Conference 2016, Edinburgh, September 2016.
  • Murat Üney, Bernard Mulgrew, Daniel Clark, "Distributed localisation of sensors with partially overlapping field-of-views in fusion networks," Fusion 2016, Heidelberg, July 2016.
  • Murat Üney, Bernard Mulgrew, Daniel Clark, "Distributed estimation of latent parameters in state space models using separable likelihoods," in the Proc. of the 41st IEEE ICASSP 2016 , March 2016 (long version w/proofs).
  • Murat Üney, Bernard Mulgrew, Daniel Clark, "Maximum likelihood signal parameter estimation via track before detect," in the Proc. of the SSPD 2015, Edinburgh UK, 2015.
  • Murat Üney, Bernard Mulgrew, Daniel Clark "Cooperative sensor localisation in distributed fusion networks by exploiting non-cooperative targets," IEEE Workshop on Statistical Signal Processing 2014, Gold Coast Australia, 2014.
  • Murat Üney, Bernard Mulgrew, Daniel Clark "Target aided online sensor localisation for bearing only clusters," in the Proc. of the SSPD 2014, Edinburgh UK, 2014.
  • J. Barr, Murat Üney, D. E. Clark, D. Miller, M. Porter, A. Gning and S. J. Julier, ``A multi-sensor inference and data fusion method for tracking small, manoeuvrable maritime craft in cluttered regions,'' the Proc. of the 3rd IMA Conference on Mathematics in Defence. IMA, Malvern, UK, October 2013.
  • Murat Üney, Daniel E. Clark, Simon J. Julier, "Distributed Sensor Registration based on Random Finite Set Representations", in the Proc. of the SSPD 2012. UDRC, London, UK, September 2012.
  • Murat Üney, Daniel E. Clark and Simon J. Julier, "Information Measures in Distributed Multitarget Tracking", Proc. of Fusion 2011, July 2011.
  • Murat Üney, Simon J. Julier, Daniel E. Clark, Branko Ristic, "Monte Carlo Realization of a Distributed Multi-Object Fusion Algorithm", in the Proc. of the SSPD 2010. UDRC, London, UK, September 2010.
  • Murat Üney and Müjdat Çetin, "An Efficient Monte Carlo Approach for Optimizing Decentralized Estimation Networks Constrained by Undirected Topologies", in the Proc. of the Workshop on Statistical Signal Processing (SSP) 2009. IEEE, Cardiff, Wales, UK, Aug. 2009.
  • Murat Üney and Müjdat Çetin, "An Efficient Monte Carlo Approach for Optimizing Communication Constrained Decentralized Estimation Networks", in the Proc. of The 17th EUSIPCO. EURASIP, Glasgow, Scotland, UK, Aug. 2009.
  • Murat Üney and Müjdat Çetin, "İletişim Kısıtları Altında Dağıtık Rasgele-Alan Kestirimi (Decentralized Random-Field Estimation Under Communication Constraints)", in the Proc. of The 17th Conference on Signal Processing, Communications, and their Applications (SIU 2009). IEEE, Antalya, Turkey, April 2009, (in Turkish, best runner-up in the IEEE student paper competition).
  • Murat Üney and Müjdat Çetin, "Akustik Algılayıcı Ağlarında Çarpan Çizgeleri Kullanarak Hedef Konumlandırma (Target Localization in Acoustic Sensor Networks Using Factor Graphs)", in the Proc. of The 16th Conference on Signal Processing, Communications, and their Applications (SIU 2008). IEEE, Aydın, Turkey, April 2008, (in Turkish).
  • Murat Üney and Müjdat Çetin, "Graphical Model-based Approaches to Target Tracking in Sensor Networks: An Overview of Some Recent Work and Challenges", in the Proc. of The International Symposium on Image and Signal Processing and Analysis (ISPA 2007). IEEE, İstanbul, Turkey, September 2007.
  • DISSERTATIONS and TECHNICAL REPORTS
  • Murat Üney and Müjdat Çetin, "Monte Carlo Optimization Approach for Decentralized Estimation Networks Under Communication Constraints", Technical Report, SU_FENS_2010/0007 Sabancı University, Nov. 2010.
  • Murat Üney, "Distributed Estimation Under Communication Constraints", Ph.d. Dissertation, August 2009.
  •  
    Home Research Publications Resume