DSP PROJECTSThe following projects, could be implemented in a group of 2 people (for a +0.5 bonus extra points with

respect to the exam), or by every single student (to get a +0.75 extra points). Some of the projects may

be gifted more, as they are indicated in the following paragraphs related to the projects’ list.

What to be delivered:

1. A report containing the theory, explanation of the problem (understanding), details of

implementation along with its associated MATLAB (or python, or C++) code related to it, plus

the results (curves, tables, simulation results, etc.), not more than 20 pages, in Latex(with *.tex

file included) or Microsoft office format (*.doc/*.docx).

2. A complete set of codes in a zipped/rarred format, with necessary comments about the

implementation issues related to the general and local structures of the code.

A proposed list of the topics for the projects:

1. Nonlinear Beamformer using Neural Networks. (+0.75 for a group of 2, or +1 for a single person

group)

2. Dereverberation of the speech signals using wavelets/filterbanks.

3. Speaker identification using LPC features.

4. Speech diarization (+0.75 for a group of 2, or +1 for a single person group)

5. Adaptive kernel Beamforming of multiple sources in an echoic (reverberant) environment.

6. Online tracking of a moving object (or moving speaker) using adaptive /online methods.(RLS,

Kalman, etc.) (+0.75 for a group of 2, or +1 for a single person group)

7. Implementation of a General Sidelobe Canceller (GSC) Beamformer, using LMS adaptive filter.

8. Online voiced/unvoiced/silent detection of the speech frames. (e.g., using wavelets, speech

features, etc.) (+0.75 for a group of 2, or +1 for a single person group)

9. Online detection/tracking of pitch periods for single (or multiple combined) source(s).

10. Online localization of multiple speech sources in a reverberant/noisy environment. (+0.75 for a

group of 2, or +1 for a single person group)

11. Optimum Bayesian Beamforming processing.

12. Localization and separation (including the beamforming processing) of multiple speech sources

in a reverberant environment using ICA, SCA, or NMF based methods. (+1 for a group of 2, or

+1.25 for a single person group)

When, and how to deliver the project related files?

The necessary files associated to the projects, should be delivered electronically (by email) any time

before the exam date to the following email directly: mahdian.t.r@gmail.com or a download link to your

personal dropbox could be emailed to this email address.