course: Fundamentals of Automatic Speech Recognition

number:
141044
teaching methods:
lecture with integrated lab excercises
media:
Videoübertragung, overhead transparencies, Moodle
responsible person:
Prof. Dr.-Ing. Do­ro­thea Kolossa
lecturer:
Prof. Dr.-Ing. Do­ro­thea Kolossa (ETIT)
language:
german
HWS:
4
CP:
6
offered in:
summer term

dates in summer term

  • start: Tuesday the 13.04.2021
  • lecture Tuesdays: from 14:15 to 15.45 o'clock
  • tutorial Wednesdays: from 12:15 to 13.45 o'clock

Exams

Die Angaben zu den Prüfungsmodalitäten (im WiSe 2020/2021 | SoSe 2021) erfolgen vorbehaltlich der aktuellen Situation. Notwendige Änderungen aufgrund universitärer Vorgaben werden zeitnah bekanntgegeben.

Date according to prior agreement with lecturer.

Form of exam:oral
Registration for exam:FlexNow
Duration:30min
description of exam:

Änderung der Prü­fungs­form im WiSe 20/21

Die Angaben zu den Prüfungsmodalitäten (im WiSe 2020/2021 | SoSe 2021) erfolgen vorbehaltlich der aktuellen Situation. Notwendige Änderungen aufgrund universitärer Vorgaben werden zeitnah bekanntgegeben.
Form of exam:written
Registration for exam:FlexNow
Date:10.09.2021
Begin:08:30
Duration:120min
Rooms : HMA 10,  HMA 20
Individual appointments of students to each exam location will be issued by the responsible chair.

goals

The participants understand the theoretical foundations and practical realization concerns of automatic speech recognition systems. They can implement the core algorithms of automatic speech recognition and they understand the principle of operation of state-of-the art small- and large-vocabulary recognizers.

content

The lecture teaches foundations and application of machine speech recognition in that form, in which they are being used in current systems for continuous speech recognition. The following topics are addressed

  • Foundations: Phonetics, Speech Perception
  • Deep neural networks and statistical methods for classification and posterior probability estimation
  • Feature extraction in the time, frequency and cepstrum domain
  • Speech Recognition using Hidden Markov Models

Simultaneously, in practical excercises, a connected-word recognizer is implemented in Python in small teams of 2-3 students.

requirements

keine

recommended knowledge

  • Basic knowledge of digital signal processing
  • Basic programming knowledge

materials

script:

miscellaneous

Due to the currently implemented emergency regulations at the RUB for the summer term 2021, this course will be offered as an online class. Therefore, all lectures and exercises are carried out with the help of video conferences or self-learning tutorials. The details will be presented in the first video lecture, which will be available in Moodle from 2 p.m. Tuesday, April 13, 2021. Any questions in this respect will be answered in the discussion forum of the course in Moodle and/or in the digital office hours via zoom.

Registration for the course in advance is mandatory!

Please send an email until April 11, 2021, 11:59 p.m. using your RUB email address with the subject "An­mel­dung Kurs 141044 SoSe2021" to benedikt.boenninghoff[at]rub.de and steffen.zeiler[at]rub.​de. All further information, in particular the password to the Moodle course, the use of the video conference system and details to the exercises, will be sent to the participants by email on April 12, 2021.

(Digital) office hours in the semester:

  • Prof. Dr.-Ing. Dorothea Kolossa Wednesdays 9:30 a.m.- 11:00 a.m.
  • Exercise Q&A 1: Thursdays, 10:00-11:00 a.m.
  • Exercise Q&A 2: Fridays, 11:00 a.m. - 12:00 a.m.