Dansk - English
Short version - Full version
Digital signal processing and mathematics 4 (Spring 2013) |
|||
| Course code : | EDSM4-U01 | ||
| ECTS Credits : | 10 | Status : | Compulsory |
| Revised : | 06/02 2010 | Written : | 06/12 2001 |
| Placement : | 4. semester | Hours per week : | 8 |
| Length : | 1 semester | Teaching Language : | English if English students are present |
| Objective : | Purpose: Development of mathematical key competences for electronic and IT engineers. The course mixes a theoretical and applicational aim with emphasis on problem-orientation and the use of Matlab to support the learning. The study module should make the student able to: Analyse discrete-time signals and systems in the frequency domain. This includes the understanding and evaluation of the problems and limitations of computer-based frequency analysis of analogue signals. Understand elementary definitions, concepts and calculation methods concerning probability and statistics and to apply these in the description and analysis of stochastic signals. Evaluate the possibility for data compression of a signal source on basis of the source’s statistical properties. Explain the fundamental principles for speech coding and evaluate different standards. Explain in theory the principles of automatic speech recognition and utilize these to produce a single word speech recognizer. Work out, elaborate and defend the documentation of the compulsory group assignments of the course. |
||
| Principal Content : | The main content is divided into four parts: A, B, C and D: A. Discrete time Fourier analysis: Discrete time Fourier transform (DTFT) and inverse (IDTFT). Discrete Fourier transform (DFT/FFT) and inverse (IDFT/IFFT). Practical spectral analysis: Spectrum analysers, frequency resolution, amplitude precision, window functions, leakage, zero padding. B. Probability and stochastic signals: Probability: definitions, calculus and Probability Functions. Stochastic Signals. Elementary Information Theory and Source Coding. C. Speech Coding in telecommunication: Quantization noise, linear Pulse Code Modulation (PCM) and companding. Differential PCM (DPCM) and linear prediction, adaptive DPCM (ADPCM), Pitch-excited linear predictictive coding (LPC). Possibly modern speech coding principles and standards (GSM , CELP etc). D. Automatic Speech Recognition: Cepstrum Analysis, Feature Extraction, Time Warping, Model Word production, Distance computation and Word Recognition. |
||
| Teaching method : | In the beginning of the semester the teaching will be a mixture of classroom lecturing and problem work solving, partly based on MATLAB as the simulation tool. This form makes up 50% to 75% of the scheduled time in the semester, distributed as 100% in the beginning of the semester and 0% towards the end of the semester. In relation to the contents items A and B two large course assignments have to be carried out and documented in writing. In relation to the contents item C and D a large course assignment involving speech coding practise and design of a speech recognizer has to be carried out and documented in a report. | ||
| Required prequisites : | Documented knowledge similar to DSM2Documented knowledge similar to DSM2 | ||
| Recommended prerequisites : | - | ||
| Relations : | Prerequisite for the elective study modules: DCEC, DCM, WWL, DIP, LAAF, ANN. | ||
| Type of examination : | Look under remarks | ||
| External examiner : | External | ||
| Marking : | 7 step scale | ||
| Remarks : | - Scale of 7 stages, three of which are described in the following: 12 is rewarded for a brilliant performance demonstrating knowledge, routine, professional depth and reflective skills with respect to the learning outcomes - with no or only a few minor flaws. 7 is rewarded for a more average performance demonstrating knowledge, routine and depth with respect to almost all of the learning outcomes- but with some more essential flaws. 2 is rewarded for a performance just acceptable in terms of basic knowledge and routine. - Oral examination mainly based on the compulsory assignments and the teaching as a whole with individual grading. In the beginning of the semester, the students form project groups of 3-4 students. Other group formations must be accepted by the teacher. If a group member repeatedly violates the group agreements, the group may recommend him/her for exclusion. If a student does not contribute to the project work or the compulsory assignments are not accepted, he/she will be refused admission to the exam. Before the examination: A presentation by each group member of a self-chosen topic from the items C/D of the final project in presence of the entire group. It is required that the topic selection is coordinated between the group members to be non-overlapping and cover major points of the item C and D assignment with professional width and depth. Up to 10 minutes per student is available for this presentation. Examination: Internal and external examiner ask supplementary questions to the presentation. After this the students are examined in a topic, selected at random from a pool of topics originating from both the first two course assignments and the teaching as a whole. Approximately 10 minutes per student is available for this examination. The grade is rewarded on 1) the basis of the quality of the report of the item C/D assignment and the presentation/defence of the self-chosen topic (50%) and 2) the individual examination in the randomly selected topic (50%). The grade is awarded as an average of the assessment of part 1) and 2) with the exception that neither of the parts 1) and 2) must be assessed as completely unacceptable ( -3 in the scale of 7). In that case the grade will be -3 or 0. |
||
| Teaching material : | James McClellan, Ronald Schafer, Mark Yoder: Signal Processing First. (Prentice Hall 2003). ISBN 0-13-120265-0. Croft, Davison, Hargreaves: Engineering Mathematics (Addisson Wesley) Electronic notes available on the Campusnet fileshare and printed handouts. |
||
| Responsible teacher : | John Kryger Sørensen
, jksor@dtu.dk |
||