Informatik 6 - Automatic Speech Recognition, WS 2009/2010

Automatic Speech Recognition, WS 2010/2011

News - Download - Schedule - Exercises - Info - Literature - Contact

News

Nov. 4, 2010

Please note: As pointed out on the first exercise sheet, the first exercise is to be submitted by next Monday, Nov. 8, 2010, in or before the exercise lesson at 8:15h.

Download

Here you can find the download material for the lecture. The exercise sheets as well as other documents needed for the solution of the exercises will be made available on this page.

Lecture Notes:

Download lecture notes

Exercises:

Type	Date	Download	others

The exercises should be treated in groups of 3 students.
Requirement for the issuing of certificates (Schein/Leistungsnachweis) will be that at least 50% of the exercise points are obtained.
Solutions to problems without points will be presented in the exercise lesson.

For the later exercises you will need a mathematical toolkit, for which we strongly suggest the usage of the public domain toolkit Octave (similar to MatLab). Solutions to the exercises will be given only in Octave code.

You can download Octave and its documentation here.

The RWTH Aachen University Speech Recognition System can be downloaded here.

Exercises

1. Exercise Sheet (Submission: November, 15^th 2010)

2. Exercise Sheet (Submission: November, 22^nd 2010)

3. Exercise Sheet (Submission: November, 29^nd 2010)

4. Exercise Sheet (Submission: December, 6^th 2010)

5. Exercise Sheet (Submission: December, 13^th 2010)

6. Exercise Sheet (Submission: December, 20^th 2010)

7. Exercise Sheet (Submission: January, 19^th 2011)

spoken.txt recognized.txt ref.dat test.dat lev_skeleton.cc

8. Exercise Sheet (Submission: January, 24^th 2011)

probe1.ascii obs.data segmentation.pdf

Schedule

Type	Date	Room	Start	Instructor
V3/4	Mo, 10:00 - 11:30	AH 6		Dr.rer.nat. R. Schlüter, Prof. Dr.-Ing. H. Ney
	Do, 11:45 - 13:15	AH 6	26.10.2010
Ü2	Mo, 08:15 - 09:45	seminar room i6	08.11.2010	Nussbaum-Thom, Schlüter, Ney

info

Content:
The aim of automatic speech recognition is to extract the spoken word sequence from the acoustic signal. An automatic speech recognition system consists typically of four parts:

the signal analysis extracts time-dependent features from the acoustic signal
the acoustic model establishes a link between features and phonemes
the language model covers syntactic and semantic constraints of the language
the search process determines the word sequence that fits best to the features taking the acoustic and language model into account.

Assignment:
Applied Computer Science or Field of Specialization.

Requirements:
4 terms of university (Fachsemester), knowledge of probability theory / statistics

Language of instruction:
English

Literature

L. Rabiner, R. Schafer: "Fundamentals of Speech Recognition", Foundations and Trends® in Signal Processing: Vol. 1: No 1-2, pp 1-194, 2007. (free download)
T. Lehmann, W. Oberschelp, E. Pelikan, R. Repges: "Bildverarbeitung für die Medizin. Grundlagen, Modelle, Methoden, Anwendungen.", Springer-Verlag, Berlin, 1997.
A. V. Oppenheim, R. W. Schafer: "Discrete Time Signal Processing",Prentice Hall, Englewood Cliffs, New Jersey, 1989.
L. Rabiner, B. H. Juang: "Fundamentals of Speech Recognition", Prentice Hall, Englewood Cliffs, New Jersey, 1993.
M.Gales, S. Young: "The Application of Hidden Markov Models in Speech Recognition", Foundations and Trends® in Signal Processing: Vol. 1: No 3, pp 195-304, 2008. (free download)
W. K. Pratt: "Digital Image Processing",John Wiley and Sons, New York, NY, 1991.

Contact

Dr. Ralf Schlüter	schlueter@informatik.rwth-aachen.de	Tel. 80-21612
Markus Nussbaum-Thom	nussbaum@informatik.rwth-aachen.de	Tel. 80-21602

Office hours: by arrangement