Quick Links | Home | Worldwide

Search for All Microsoft ResearchDownloadsPublicationsResearcher PagesNews & Feature Stories Microsoft Research Home About Microsoft Research Research Areas People Worldwide Labs University Relations Publications Downloads Conferences and Events Careers Visiting Microsoft Research Press Resources

Manuel J. Reyes-Gomez  Contact Information Research Projects Publications Invited Talks Education CV Machine Learning and Applied Statistics Group's home page.     I am a Post Doctoral Researcher in the Machine Learning and Applied Statistics Group. I am interested in designing models and algorithms using techniques from pattern recognition, probability theory, decision theory, machine learning and signal processing to resolve practical, challenging problems. Since I joined Microsoft Research in April 2005, I have been working in the design of machine learning models and algorithms for audio analysis and biology oriented projects. The analysis of audio signals for a wide variety of applications, such as speech recognition and content based retrieval, assumes that the signals originated from a single source. However, that is frequently not the case, I am interested in finding and separating the different sources on a single channel audio signal. On the biology part, I have been working in the design of models to characterize the multiple processes involved with the immune system. We later use these models as tools to analyze data from actual patients, finding correlations between pathogen's sequence patterns and the particular patients' immune systems. I am also interested in finding patterns of expression on microarray data for essential and non-essential genes in yeast.

• Microsoft Research
  One Microsoft Way 113/3125
  Redmond, WA 98052
  Tel. +1 (425) 707-5886
  Fax. +1 (425) 936-7329
  Email: manuelrg@microsoft.com

• Audio: Model Based Source Separation
• Computational Biology: Immunoinformatics
• Multidomain (audio/images/video) Models and Algorithms
• Audio: Content-Based Classification an Retrieval
• Audio: Speech Recognition
• Image Analysis: Content-Based Classification an Retrieval

Computational Biology: Immunoinformatics

• Manuel J. Reyes-Gomez and Nebojsa Jojic.
  "Mixed Adaptive Double Threading: Joint MHC I Clustering and Learning Peptide Binding"
  To be submitted, Bioinformatics. March 2007
• Noah Zaitlen, Manuel J. Reyes-Gomez, David Heckerman and Nebojsa Jojic.
  "Shift Invariant Adaptive Double Threading: Learning MHC II Peptide Binding"
  To be presented, International Conference on Research in Computational Molecular Biology, Recomb 07. 2007
• Nebojsa Jojic, Manuel J. Reyes-Gomez, David Heckerman, Carl Kadie and Ora Schueler-Furman.
  "Learning MHC I Peptide Binding"
  Bioinformatics. Vol. 22, Number 14, 15 July 2006.

Audio: Model Based Source Separation

• Manuel J. Reyes-Gomez and Nebojsa Jojic.
  "Signal Separation by Efficient Combinatorial Optimization"
  Advances in Models for Acoustic Processing NIPS 2006 workshop. December 2006
• Manuel J. Reyes-Gomez, Nebojsa Jojic and Dan Ellis.
  "Modelling sound dynamics using Deformable Spectrograms: segmenting the spectrogram into smooth regions"
  Invited Paper
  Asilomar Conference on Signals, Systems and Computers. October 2006

• Manuel J. Reyes-Gomez, Nebojsa Jojic and Dan Ellis.
  "The Deformable Spectrograms Model for Sound Dynamics"
  Revision in progress IEEE Transactions on Speech and Audio Processing. 2006
  Here as a MSR technical report
• Manuel J. Reyes-Gomez
  "Statistical Graphical Models for Scene Analysis, Source Separation and Other Audio Applications"
  Ph.D. Thesis, Columbia University, 2005
• Manuel J. Reyes-Gomez, Nebojsa Jojic and Dan Ellis.
  "Deformable Spectrograms"
  Artificial Intelligence and Statistics, AISTATS'05. Barbados, January 2005
• Manuel J. Reyes-Gomez, Nebojsa Jojic and Dan Ellis.
  "Towards Single-Channel Unsupervised Source Separation of Speech Mixtures: The Layered Harmonics/Formants Separation/Tracking Model".
  Research Workshop on Statistical and Perceptual Audio Processing, SAPA04, Korea October, 2004.
• Manuel J. Reyes-Gomez,Nebojsa Jojic and Dan Ellis.
  "Detailed graphical models for source separation and missing data interpolation in audio."
  Snowbird Learning Workshop Snowbird, Utah, March 2004
• Manuel J. Reyes-Gomez, Dan Ellis and Nebojsa Jojic.
  "Multiband Audio Modeling for Single Channel Acoustic Source Separation"
  ICASSP, Montreal 2004
• Bhiksha Raj, Michael L. Seltzer and Manuel J. Reyes-Gomez
  "Speech Recognizer Based Maximum Likelihood Beamforming"
  Book chapter, Speech Separation by Humans and Machines,2004 ed. P. Divenyi, Kluwer.
• Manuel J. Reyes-Gomez, Bhiksha Raj and Dan Ellis.
  "Multi-channel Source Separation by Beamforming Trained with Factorial HMMS"
  IEEE Workshop on Applications of Signal Procesing to Audio and Acoustics, New Paltz, NY 2003
• Manuel J. Reyes-Gomez, Bhiksha Raj and Dan Ellis.
  "Multi-channel Source Separation by Factorial HMMs",
  ICASSP, Hong Kong 2003

Audio: Generic Acoustic Modelling

• Manuel J. Reyes-Gomez and Dan Ellis(2003).
  "Selection, Parameter Estimation and Discriminative Training of Hidden Markov Models for Generic Acoustic Modeling",
  International Conference on Multimedia and Expo, ICME 03, Baltimore 2003

Audio: Acoustic Modelling for Speech Recognition

• Manuel J. Reyes-Gomez and D.P.W. Ellis.
  "Error visualization for tandem acoustic modeling on the Aurora task",
  ICASSP, Orlando, May 2002.
• Dan Ellis and Manuel J. Reyes-Gomez (2001).
  "Investigations into Tandem Acoustic Modeling for the Aurora Task",
  Eurospeech-01, Denmark, September 2001.

Multidomain Models/Algorithms

• Manuel J. Reyes-Gomez and Nebojsa Jojic.
  "Signal Separation by Efficient Combinatorial Optimization"
  Advances in Models for Acoustic Processing NIPS 2006 workshop. December 2006
• Nebojsa Jojic, Yaron Caspi and Manuel J. Reyes-Gomez .
  "Probabilistic Index Maps for Modeling Natural Signals"
  Conference on Uncertainty in Artificial Intelligence. Banff, Canada, July 2004

Image Analysis

• Manuel J. Reyes-Gomez.
  "User-Defined Object Classifier based in Neural Network with Optimal Selection of Visual Features using a Genetic Algorithm"
  Technical Report. Columbia University 1999

Power Electronics

• Manuel J. Reyes-Gomez.
  "Efficient Energy Converters for Energy Savings in Air Conditioners"
  Undergraduate Thesis.Universidad de las Americas, 1995

• Manuel J. Reyes-Gomez
  "Probabilistic Graphical Models for Audio and Biology"
  15th International Conference on Computing, Mexico City, Mexico, November 2006
• Manuel J. Reyes-Gomez and Nebojsa Jojic
  "Learning MHC I Peptide Binding"
  Microsoft Research: Latinamerican Academic Summit, Guadalajara, Mexico, May 2006
• Manuel J. Reyes-Gomez, Nebojsa Jojic and Daniel P.W. Ellis.
  "Deformable Spectrograms"
  IBM Research - Multimedia - Seminars, Hawthorne NY, March 2005
• Manuel J. Reyes-Gomez, Nebojsa Jojic and Daniel P.W. Ellis.
  "Deformable Spectrograms"
  National Science Foundation Workshop on Speech Separation and Comprehension in Complex Acoustic Environments. Montreal Quebec November 2004
• Manuel J. Reyes-Gomez, Nebojsa Jojic and Daniel P.W. Ellis.
  "Harmonics/Formants Separation and Tracking Model''.
  International Conference "Auditory Scene Analysis and Speech Perception by Human and Machine''. Hanse Institute for Advanced Study, Delmenhorst, Germany August 2004.
• Manuel J. Reyes-Gomez.
  "Statistical Models for Source Separation, Scene Analysis and Other Audio Applications''
  Victoria University, Victoria Canada, July 2004.
• Manuel J. Reyes-Gomez.
  "Graphical Models and Variational Methods for Source Separation and other Audio Applications.''
  2003 IBM User Interface Technology (UIT) Student Symposium. IBM T.J. Watson Research Center, NY. September 2003

• Ph.D. in Electrical Engineering, May 2005
  Laboratory for the Recognition and Organization of Speech and Audio, LabROSA,
  Columbia University, New York, USA
• M.S. in Electrical Engineering, 2000
  Specialization on Multimedia Technologies,
  Columbia University, New York, USA
• B.S. in Electronics and Communications Engineering, 1995
  Power Electronics,
  Universidad de las Americas, Puebla, Mexico