Two Ph.D. fellowships available at  

Department of Electronics and Telecommunications

The Norwegian University of Science and Technology ( NTNU)

Trondheim, Norway

The Ph.D. scholarships are part of the projects SIRKUS (Spoken Information Retrieval by Knowledge Utilization in Statistical Speech Processing) and SMUDI (Voice Control in Multi-Modal Dialogue).

PhD Fellowship in the SIRKUS project:
The vision of SIRKUS is to develop new architectures for speech recognition that can achieve human-like performance. There is evidence indicating that system performace based on the current paradigm is reaching an upper bound. We believe that in order to achieve our vision, knowledge about speech production and perception as well as a better understanding of speech per se needs to be incorporated into a statistical framework. This implies that new approaches to speech analysis need to be investigated and developed, and that a new statistical framework for modelling speech based on a set of analysis results needs to be defined.

The PhD project
An on-going PhD project will investigate and develop speech analysis methods that will supplement and augment the current methods that are based on spectral analysis. In particular, development of detectors of phonologically significant events, i.e. speech attributes, will be central. A set of speech analyses, including speech attribute detectors, will produce speech event sequences, which can constitute temporally asynchronous observation streams and may be correlated both in time and across observation streams.
This  PhD project will in cooperation with the on-going project develop novel statistical modelling approaches suitable for modelling speech based on event sequences. Integration of the observation streams from the speech analyses in a statistical description of various linguistic units such as phonemes, syllables, words and sentences will be central.
The project will incorporate collaboration with several foreign partners. The scholarships will also include an international visiting research scholarship.

Qualifications:
We seek highly motivated individuals holding a masters degree in electronics engineering, signal processing, statistics, or other relevant disciplines. Experience in speech technology is desirable, but not an absolute requirement.

PhD Fellowship in the SMUDI project:
Large groups of disabled persons have great difficulties accessing information that is available on the internet. Many government and municipal agencies are in the process of changing their preferred interaction with the public, moving to internet based systems for submission of applications and information requests. In many cases it will be necessary for the user to provide information by filling in forms. Some examples are online shopping for goods and services, e.g. air travel and use of public services such as filling in tax return forms.

The PhD project
The goal of the PhD project is to develop a speech based system for filling in internet forms. The work will include methods for interpretation of the internet forms, speech recognition to transform the user speech to text, and integration of speech tehnology with other modalities for information presentation and user input.

In order to achieve a best possible performance for this task, the speech recognition ought to be at the user’s computer. This will have the advantage that the speech recognizer can be adapted to the user’s voice and pronunciation, and in addition provide a situation where the speech signal is not band limited or noise corrupted by the transmission channel. The challenge is that the recognizer cannot be tailored for filling in one particular form, and thus it will be necessary to develop a general large vocabulary speech recognizer (a “dictation engine”) for Norwegian. The system should be able to define vocabulary and syntax dynamically by interpreting the content of the internet form.

Qualifications:
We seek highly motivated individuals holding a masters degree in electronics engineering, signal processing, statistics, natural language processing or other relevant disciplines. Experience in speech technology is desirable, but not an absolute requirement.

Information for both fellowships:
The PhD fellows will be associated with the Signal Processing Group at  NTNU and will work in a strong and active scientific environment.

Award holders at NTNU are normally appointed for up to 4 years with 25% of the time spent on specified work. This work is primarily linked to teaching and is usually divided so that a relatively large part is done in the first half of the period of the appointment.
The appointments is at code 1017, salary level 43-58 in the national salary scheme, gross NOK. 325.600 – 423.800 per annum (1 NOK ~ 0.125 EUR), and are normally remunerated at wage level 43 of which 2% is deducted for the State Pension scheme. The salary might be adjusted after negotiation with the employer to reflect the applicant’s experience.

Further information:
For more information on the SIRKUS position and the application requirements, please see http://www.iet.ntnu.no/projects/SIRKUS/Positions.html
For more information on the SMUDI position and the application requirements, please see http://www.iet.ntnu.no/projects/SMUDI/Positions.html

Interested candidates are encouraged to contact Professor Torbjørn Svendsen (phone: +47-735-92674, email: torbjorn@iet.ntnu.no) for further information.