United Kingdom
United Kingdom

Time filter

Source Type

Koike-Akino T.,MERL | Duan C.,MERL | Parsons K.,MERL | Kojima K.,MERL | And 3 more authors.
Optics Express | Year: 2012

Fiber nonlinearity has become a major limiting factor to realize ultra-high-speed optical communications. We propose a fractionally-spaced equalizer which exploits a trained high-order statistics to deal with datapattern dependent nonlinear impairments in fiber-optic communications. The computer simulation reveals that the proposed 3-Tap equalizer improves Q-factor by more than 2 dB for long-haul transmissions of 5,230 km distance and 40 Gbps data rate. We also demonstrate that the joint use of a digital backpropagation (DBP) and the proposed equalizer offers an additional 1-2 dB performance improvement due to the channel shortening gain. A performance in high-speed transmissions of 100 Gbps and beyond is evaluated as well. ©2012 Optical Society of America.

Vincent E.,French Institute for Research in Computer Science and Automation | Barker J.,University of Sheffield | Watanabe S.,MERL | Le Roux J.,MERL | And 3 more authors.
ICASSP, IEEE International Conference on Acoustics, Speech and Signal Processing - Proceedings | Year: 2013

Distant-microphone automatic speech recognition (ASR) remains a challenging goal in everyday environments involving multiple background sources and reverberation. This paper is intended to be a reference on the 2nd 'CHiME' Challenge, an initiative designed to analyze and evaluate the performance of ASR systems in a real-world domestic environment. Two separate tracks have been proposed: a small-vocabulary task with small speaker movements and a medium-vocabulary task without speaker movements. We discuss the rationale for the challenge and provide a detailed description of the datasets, tasks and baseline performance results for each track. © 2013 IEEE.

Jones M.,MItsubishi Electric | Geng Y.,Boston University | Nikovski D.,MERL | Hirata T.,MItsubishi Electric
IEEE Conference on Intelligent Transportation Systems, Proceedings, ITSC | Year: 2013

We study the problem of predicting travel times for links (road segments) using floating car data. We present four different methods for predicting travel times and discuss the differences in predicting on congested and uncongested roads. We show that estimates of the current travel time are mainly useful for prediction on links that get congested. Then we examine the problem of predicting link travel times when no recent probe car data is available for estimating current travel times. This is a serious problem that arises when using probe car data for prediction. Our solution, which we call geospatial inference, uses floating car data from nearby links to predict travel times on the desired link. We show that geospatial inference leads to improved travel time estimates for congested links compared to standard methods. © 2013 IEEE.

Benosman M.,MItsubishi Electric | Farahmand A.-M.,MERL
IFAC-PapersOnLine | Year: 2016

We study in this paper the problem of adaptive trajectory tracking control for a class of nonlinear systems with parametric uncertainties. We propose to use a modular adaptive approach, where we first design a robust nonlinear state feedback which renders the closed loop input-to-state stable (ISS). The input is considered to be the estimation error of the uncertain parameters, and the state is considered to be the closed-loop output tracking error. We augment this robust ISS controller with a model-free learning algorithm to estimate the model uncertainties. We implement this method with a Bayesian optimization-based method called Gaussian Process Upper Confidence Bound (GP-UCB). The combination of the ISS feedback and the learning algorithms gives a learning-based modular indirect adaptive controller. We test the efficiency of this approach on a two-link robot manipulator example, under noisy measurements conditions. © 2016

Benosman M.,MItsubishi Electric | Kramer B.,Massachusetts Institute of Technology | Boufounos P.T.,MERL | Grover P.,MERL
Proceedings of the American Control Conference | Year: 2016

We present results on stabilization for reduced order models (ROM) of partial differential equations using learning. Stabilization is achieved via closure models for ROMs, where we use a model-free extremum seeking (ES) dither-based algorithm to optimally learn the closure models' parameters. We first propose to auto-tune linear closure models using ES, and then extend the results to a closure model combining linear and nonlinear terms, for better stabilization performance. The coupled Burgers' equation is employed as a test-bed for the proposed tuning method. © 2016 American Automatic Control Council (AACC).

Bedri H.,Massachusetts Institute of Technology | Feigin M.,Massachusetts Institute of Technology | Boufounos P.T.,MERL | Raskar R.,Massachusetts Institute of Technology
Proceedings of the IEEE International Conference on Computer Vision | Year: 2016

SONAR imaging can detect reflecting objects in the dark and around corners, however many SONAR systems require large phased-arrays and immobile equipment. In order to enable sound imaging with a mobile device, one can move a microphone and speaker in the air to form a large synthetic aperture. We demonstrate resolution limited audio images using a moving microphone and speaker of a mannequin in free-space and a mannequin located around a corner. This paper also explores the 2D resolution limit due to aperture size as well as the time resolution limit due to bandwidth, and proposes Continuous Basis Pursuits (CBP) to super-resolve. © 2015 IEEE.

Sharma A.,University of Maryland College Park | Tuzel O.,MERL | Jacobs D.W.,University of Maryland College Park
Proceedings of the IEEE Computer Society Conference on Computer Vision and Pattern Recognition | Year: 2015

This paper proposes a learning-based approach to scene parsing inspired by the deep Recursive Context Propagation Network (RCPN). RCPN is a deep feed-forward neural network that utilizes the contextual information from the entire image, through bottom-up followed by top-down context propagation via random binary parse trees. This improves the feature representation of every super-pixel in the image for better classification into semantic categories. We analyze RCPN and propose two novel contributions to further improve the model. We first analyze the learning of RCPN parameters and discover the presence of bypass error paths in the computation graph of RCPN that can hinder contextual propagation. We propose to tackle this problem by including the classification loss of the internal nodes of the random parse trees in the original RCPN loss function. Secondly, we use an MRF on the parse tree nodes to model the hierarchical dependency present in the output. Both modifications provide performance boosts over the original RCPN and the new system achieves state-of-the-art performance on Stanford Background, SIFT-Flow and Daimler urban datasets. © 2015 IEEE.

Barker J.,University of Sheffield | Marxer R.,University of Sheffield | Vincent E.,French Institute for Research in Computer Science and Automation | Watanabe S.,MERL
2015 IEEE Workshop on Automatic Speech Recognition and Understanding, ASRU 2015 - Proceedings | Year: 2015

The CHiME challenge series aims to advance far field speech recognition technology by promoting research at the interface of signal processing and automatic speech recognition. This paper presents the design and outcomes of the 3rd CHiME Challenge, which targets the performance of automatic speech recognition in a real-world, commercially-motivated scenario: a person talking to a tablet device that has been fitted with a six-channel microphone array. The paper describes the data collection, the task definition and the baseline systems for data simulation, enhancement and recognition. The paper then presents an overview of the 26 systems that were submitted to the challenge focusing on the strategies that proved to be most successful relative to the MVDR array processing and DNN acoustic modeling reference system. Challenge findings related to the role of simulated data in system training and evaluation are discussed. © 2015 IEEE.

Koike-Akino T.,MERL | Molisch A.F.,University of Southern California | Duan C.,MERL | Tao Z.,MERL | Orlik P.,MERL
IEEE Transactions on Communications | Year: 2011

Block transmission with cyclic prefix is a promising technique to realize high-speed data rates in frequency-selective fading channels. Many popular linear precoding schemes, including orthogonal frequency-division multiplexing (OFDM), single-carrier (SC) block transmission, and time-reversal (TR), can be interpreted as such a block transmission. This paper presents a unified performance analysis that shows how the optimal precoding strategy depends on the optimization criterion such as capacity, mean-square error, and secrecy. We analyze three variants of TR methods (based on maximum-ratio combining, equal-gain combining and selective combining) and two-types of pre-equalization methods (zero-forcing and minimum mean-square error). As one application of our framework, we derive optimal precoding (i.e., OFDM with optimal power and phase control) in the presence of interference limitation for distributed antenna systems; we find that without power/phase control, OFDM does not have any capacity advantage over SC transmissions. When comparing SC and TR, we verify that for single-antenna systems in the high SNR regimes, SC has a capacity advantage; however, TR performs better in the low SNR regime. For distributed multiple-antenna systems, TR always provides higher capacity, and the capacity of TR can approach that of optimal precoders with a large number of distributed antennas. Furthermore, we make an analysis of secrecy capacity which shows how high-rate messages can be transmitted towards an intended user without being decoded by the other users from the viewpoint of information-theoretic security. We demonstrate that TR precoding can be the best candidate among the non-optimal precoders for achieving high secrecy capacity, while the optimal precoder offers a significant gain over those non-optimal precoders. © 2011 IEEE.

Sturm P.,French Institute for Research in Computer Science and Automation | Ramalingam S.,MERL | Tardif J.-P.,Carnegie Mellon University | Gasparini S.,French Institute for Research in Computer Science and Automation | Barreto J.,University of Coimbra
Foundations and Trends in Computer Graphics and Vision | Year: 2010

This survey is mainly motivated by the increased availability and use of panoramic image acquisition devices, in computer vision and various of its applications. Different technologies and different computational models thereof exist and algorithms and theoretical studies for geometric computer vision ("structure-from-motion") are often re-developed without highlighting common underlying principles. One of the goals of this survey is to give an overview of image acquisition methods used in computer vision and especially, of the vast number of camera models that have been proposed and investigated over the years, where we try to point out similarities between different models. Results on epipolar and multi-view geometry for different camera models are reviewed as well as various calibration and self-calibration approaches, with an emphasis on non-perspective cameras.We finally describe what we consider are fundamental building blocks for geometric computer vision or structure-from-motion: epipolar geometry, pose and motion estimation, 3D scene modeling, and bundle adjustment. The main goal here is to highlight the main principles of these, which are independent of specific camera models.

Loading MERL collaborators
Loading MERL collaborators