Research Fellows

Dr.-Ing. Nilesh Madhu
Experienced Researcher
Division of Experimental Otorhinolaryngology
O.& N2, Herestraat 49 bus 721
Katholieke Universiteit Leuven
Leuven, Belgium
 nilesh.madhu@rub.de
 http://www.kuleuven.be/exporl
phone:+32 16 330474

Research topics within AUDIS

  • Multi-channel speech and audio processing for hearing instruments (Ongoing)
  • Processing and enhancement of EEG signals for the detection of auditory steady state responses (ASSRs) (Ongoing)
  • Black-box benchmarking of feedback control systems in hearing aids (Completed)

Biography

  • Doktor-Ingenieur, Faculty of Electrical Engineering and Information Sciences, Ruhr-Universität Bochum, Germany
  • Master of Science (Communications Engineering), Technische Universität München, Germany
  • Bachelor of Engineering (Electrical Engineering), University of Bombay, India

Research interests

  • Multi-channel audio processing
  • Acoustic source localisation and enhancement
  • Noise reduction
  • Signal processing for hearing instruments
  • EEG signal processing and enhancement
  • Beamformer design
  • Blind and semi-blind signal separation

Publications

Dissertation

 "Acoustic Source Localization: Algorithms, Applications and Extensions to Source Separation", Der Andere Verlag GmbH, Tönning, Germany, ISBN 9783899599695, 2010.

Summary
Microphone array technology is a valued tool in many acoustics based applications, some of which are already visible in our day-to-day life. These include the use of microphone arrays for improving human-machine interfaces, video-conferencing, noise-reduction in hands-free systems in automobiles, etc. The key issues in these applications are the localization and enhancement of the desired source and the suppression of unwanted interference, and these issues form the focus of this book.
We first introduce a localization taxonomy and provide an exhaustive overview of the contemporary state-of-the-art algorithms for source localization, drawn from different application areas deriving, in the process, the relations between these approaches. We then present design considerations for localization algorithms and highlight the need to take into account all the application specific characteristics and constraints.
We next address the use of source localization in the problem of active speaker separation and enhancement. After presenting an overview of the state-of-the-art, we develop a versatile localization based framework for source separation that allows us to realize a variety of robust algorithms for target speaker enhancement.
As a good localization estimate or separation performance is contingent upon a functional microphone array, we examine, in the last part, algorithms for determining the health of the array. The goal is to perform in situ self-tests and remove, if required, degraded microphones from further processing. In addition to detecting degradation, the algorithms may also be used to perform online gain calibration of the healthy microphones of an array - thereby reducing the effects of gain-mismatch in practical applications.

Book chapters

Journals

  • N. Madhu, R. Martin, "A versatile framework for speaker separation using a model-based speaker localisation approach", IEEE Trans. Audio, Speech and Language Processing, 2010 accepted
  • N. Madhu, R. Martin, "Low-complexity, robust algorithm for sensor anomaly detection and self-calibration of microphone arrays", IET Signal Processing, In press.
  • N. Madhu, "Note on Measures for Spectral Flatness", Electronics Letters, Issue 23, Volume 45, pages 1195 -- 1196, 2009.

Conference proceedings

  • S. Gergen, N. Madhu, R. Martin, "Performance characterization of linear arrays with respect to robust MVDR beamforming," Proc. International Workshop on Acoustic Echo and Noise Cancellation (IWAENC), Tel-Aviv, Israel, 2010.
  • S. Gergen, N. Madhu, R. Martin, "A new performance criterion for microphone array geometries and filter-and-sum beamforming," Proc. 9th ITG Conference on Speech Communication, Bochum, Germany, 2010.
  • N. Madhu, J. Wouters, "Localisation-based, situation-adaptive mask generation for source separation," Proc. 4th International Symposium on Communications, Control, and Signal Processing (ISCCSP), Limassol, Cyprus, 2010.
  • N. Madhu, R. Martin, "SCORE: A low complexity, robust algorithm for the detection of corrupt sensors and self-calibration of microphone arrays," Proc. 11th International Workshop on Acoustic Echo and Noise Cancellation (IWAENC), Seattle, USA, 2008.
  • N. Madhu, R. Martin, "A scalable framework for multiple speaker localization and tracking," Proc. 11th International Workshop on Acoustic Echo and Noise Cancellation (IWAENC), Seattle, USA, 2008.
  • N. Madhu, C. Breithaupt, R. Martin, "Temporal smoothing of spectral masks in the cepstral domain for speech separation," Proc. International Conference on Acoustics, Speech and Signal Processing (ICASSP), Las Vegas, USA, 2008.
  • N. Madhu, I. Tashev, A. Acero, "An EM-based probabilistic approach for acoustic echo suppression," Proc. International Conference on Acoustics, Speech and Signal Processing (ICASSP), Las Vegas, USA, 2008.
  • N. Madhu, A. Gückel, R. Martin, "Combined beamforming and frequency domain ICA for source separation," Proc. 10th International Workshop on Acoustic Echo and Noise Cancellation (IWAENC), Paris, France, 2006.
  • N. Madhu, R. Martin, "Robust speaker localization through adaptive weighted pair TDOA (AWEPAT) estimation," Proc. INTERSPEECH, Lisboa, Portugal, 2005.
  • D. Obradovic, N. Madhu, A. Szabo, C. S. Wong, "Independent component analysis for semi-blind signal separation in MIMO mobile frequency selective communication channels," Proc. International Joint Conference on Neural Networks (IJCNN), Budapest, Hungary, 2004.
  • C. S. Wong, D. Obradovic, N. Madhu, "Independent component analysis (ICA) for blind equalization of frequency selective channels," Proc. International Workshop on Neural Networks for Signal Processing (NNSP), Toulouse, France, 2003.

Abstracts

  • N. Madhu, "Data-driven mask generation for binaural source separation," Proc. International Symposium on Auditory and Audiological Research (ISAAR), Helsingor, Denmark, 2009.
  • N. Madhu, D. Oswald, R. Martin, "Speaker localization - Novel algorithm and practical aspects," Proc. 19th Conference on Speech Signal Processing (ESSV), Frankfurt am Main, Germany, 2008.
  • N. Madhu, R. Martin, H.-W. Rehn, A. Fischer, "Localization of brake squeal," Proc. 32nd Annual Conference of the German Acoustical Society (DAGA), Braunschweig, Germany, 2006.
  • C. Breithaupt, N. Madhu, F. Hummes, R. Martin, "A robust steerable realtime multichannel noise reduction system," Joint Workshop on Hands-Free Speech Communication and Microphone Arrays (HSCMA), New Jersey, USA, 2005.
  • N. Madhu, "Autonomous mobile robots (collision avoidance and path planning)," IEEE (Bombay Chapter), Student Paper Contest, Mumbai, India, 2000.

Invited talks

  • "Target source enhancement using microphone arrays: Fmriomxtures to `clean' speech," Given at the Department of Imaging Science \& Technology, Technische Universiteit Delft, Delft, The Netherlands, 2010.
  • "Independent Component Analysis - A Tutorial Introduction," For the research group on statistical model building at the Faculty of Statistics, Technische Universität Dortmund, Dortmund, Germany, 2009.
  • "Speaker Localization - An Overview of the State of the Art," Talk at the 38th meeting of the German Information Technology Society, Ruhr-Universität Bochum, Bochum, Germany, 2004.

Back to Research Fellows

Attachments