Laboratory Course  

Content

During the winter term 2008/2009 the Lehrstuhl für Informatik 6 offers the laboratory course Speech- and Image Processing for the main studies. The course covers the implementation of basic methods of speech and image recognition using the programming language C/C++. Currently the following topics are planned:
  • image object recognition:
    • Gaussian single and mixture distributions
    • Simple image classification
    • Application of the image object recognition software of the chair

  • automatic speech recognition
    • Introduction to the speech recognition software of the chair
    • Signal processing: Cepstrum-analysis
    • Acoustic modeling: Hidden Markov Models, Expectation Maximization training
    • Search: Linear search, analysis of the search space
    • Language modelling: Extraction of vocabulary, construction of a language model

Schedule

The first meeting takes place on Thursday, November 6, 2008 at 10:00h in the seminar room 6124 of Lehrstuhl für Informatik 6. The course itself will take place as a block at the end of the lecture period.

The lab course begins on Tuesday, February 10, and ends on Friday, February 27, 2009. Each day will start with a meeting at 10:00 in the seminar room of Lehrstuhl für Informatik 6 (room 6124), where the solution to the last exercise will be discussed and the exercise of the day will be handed out. Presence on the introductory meetings is mandatory. The exercises have to be solved before the next morning's meeting.

Calendar Week (KW) 7
  • Tues-Wed: Image processing (Philippe Dreuw "dreuw@i6...")
  • Thur-Fri: Signal analysis (Jonas Lööf "loof@i6...")
    Calendar Week (KW) 8
  • Mon-Wed: Acoustic model training (Christian Plahl "plahl@i6...")
    Calendar Week (KW) 9
  • Tues-Wed: Connected digit recognition (Christian Plahl "plahl@i6...")
  • Thur-Fri: Language modelling (Sasa Hasan "hasan@i6...")

    Among others the goal of this lab course is to introduce candidates for research supplementals (a.k.a. "hiwi-jobs") and diploma works to the methods applied at Lehrstuhl für Informatik 6.

    Documents (Access only permitted within the RWTH domain)

    It is recommended to review the appropriate lecture notes and slides from corresponding courses:
    • Handout from introductory session  
    • Speech Recognition, lecture slides    
    • Pattern Recognition and Neural Networks, lecture notes    
    • Statistical Methods in Natural Language Processing, lecture slides    

    Requirements

    Participants should have passed their Vordiplom/Bachelor and must have attended at least one of the lectures "Speech Recognition", "Digital Processing of Speech and Image Signals", or "Pattern Recognition and Neural Networks".
    Practical experience with the programming language C/C++ is helpful.

    Registration

    Candidates must register using the online registration form. The registration will be open from Monday, June 16, 2008 until Sunday, June 29, 2008.

    Assignment

    The course is assigned to applied computer science or as specialization.

    Recurrence

    This course will be repeated each winter term.

    Enquiries

    Enquiries concerning the organization of the lab course may be sent to Jonas Lööf, Lehrstuhl für Informatik 6, Phone: 80-21621.

    •