1 overview of research areas and contributions tareq y. al-naffouri director office of cooperation...

Post on 28-Mar-2015

219 Views

Category:

Documents

2 Downloads

Preview:

Click to see full reader

TRANSCRIPT

1

Overview of Research Areas and Contributions

Tareq Y. Al-Naffouri

Director

Office of Cooperation with KAUST

Assistant Professor

Electrical Engineering

2

Major Milestones

• PhD (Stanford), Jan. 2005

• Assistant Prof. (KFUPM), May 2005

• Two month summer visit (Cal Tech), Summer 2006

• Fulbright Scholar (USC), Feb-Aug. 2008

• Director of Cooperation with KAUST Office, Nov. 2008

• Applied to Assoc. Prof., Jan. 10th 2009

3

9Research Area Collaborations

Adaptive signal processing Two Brazilian Universities

Statistical signal processing USC

Wireless communications U. of Texas at Dallas

Information Theory Cal Tech

Random Matrix Theory SUPELEC, France

Seismic Signal Processing SchlumbergerGeophysics Dept. (KFUPM)

During

PhD

Research Areas and Collaborations

After

joining

KFUPM

4

An Analogy

Spanish

French

German

Training Useful Instruction

5

This happens every semester

Spanish

French

German

Portuguese

Russian

Greek

First Semester Second Semester

6

How can we reduce the training over

Spanish

French

German

Portuguese

Russian

Greek

7

Digital Communication Scenario

• Receiver needs to know the communication medium to recover the transmitted data reliably

• Some of transmission time sacrificed for the sake of training

Training Useful Transmission

0 1 0 0 1 1 …0 1 0 0 1 1 …

8

Multiple Transmit Antennas

• With multiple antennas, we can transmit more data

• … But then more training is needed

9

Problem Compounded by Mobility

• With mobility, we need to train more often as the medium keeps changing

10

How to reduce the training overhead

• Utilize correlation/similarities between different channels, e.g.

1. Spacing between antennas2. Speed of car

• The more similarities used the lower the training needed

11

Figure shows error rate comparison between our algorithm and best algorithm known in literature

� State of the art

performance

� Our algorithm

Around 5 dB gain in performance

12

Our contributions

• Identified and utilized all possible dimensions (seven) of similarity to reduce training overhead

• Training done in a transparent manner using a dynamic program

“Forward Backward Kalman Filter” • Research resulted in

1. 2 Master Theses2. 2 IEEE papers3. 2 top European Journal papers4. Book Chapter5. Patent (pending)

13

Combating Impulsive Noise in DSL

• Impulsive noise is a rare phenomenon

• But when it occurs it destroys the transmitted signal completely

• Caused by bursty disturbances

• Car ignition• Telephone Network switching• A vacuum cleaner

14

Typical DSL Signal + Impulsive Noise

• Impulse prob: 3x10^-3

• Impossible to differentiate signal from Imp. Noise

• Can not simply combat noise by clipping

• Result is a punctured signal

15

Noise Estimation and Cancellation

Time Domain

Air Domain

Immerse tire in water;

Locate & eliminate

puncture

Water Domain

Freq Domain

Transmission

Band

(signal + noise)Guard

Band

Guard

Band

16

� No impulsive noise

Estimate/Remove

Predict presence

of noise

Puncture

Impulse probability

Rate

Around 35% increase in rate

Effective Tech. for Combating Impulsive Noise

17

Our Contributions

• Work used an emerging technique for identifying sparse phenomena from few observations

• Work done jointly with a Prof. at USC• Research resulted in

1. M.S. Thesis (in progress)2. 2 IEEE papers (in progress)3. Pending patent

• Working on establishing a research group for estimation of sparse phenomena

18

Multi-user Information Theory

Math

Engl.

K subjects

N students

M Professors Phys.

Mgt.

The larger the number of students, the more difficult it is

to cover the material properly

19

Broadcasting to Groups of Users

Cartoon

20

Assume there are • K TV/Radio channels• N users• M Transmitters

What is the information rate that can received reliably?

21

• How can we maintain a constant rate with increasing no. of users?

Increase the number of Antennas

• How much should the number of antennas grow to maintain a constant rate?

and you should not grow at any lower rate

22

Our Contributions

• Study is first of its kind; researchers in past focused on independent users

• Study done jointly with researchers at Cal Tech • Research resulted in

1. M.S. Thesis (in progress)2. 2 IEEE papers3. European Journal paper

23

Research Time-Line

Year 2006 2007 2008 2009 2010

Proj.Junior Proj.

Uni. Proj 1 Uni Proj 2

SABIC ProjKACST Proj.Seismic Consrt

SABIC Proj.Uni Proj

CLBRCal TechU of TX

Schlumberger USC (Flbright)SUPELEC.Brzl. U. 1

Brzl. U. 2

Grads 2 Grads in 2 Grads out2 Grads in

1 Grad out1 Grad in

2 Grads out

JrnlsIEEE Jrnl. IEEE Jrnl.

1 Eurp JrnlBook Ch.

3 IEEE Jrnl.3 Eurp. Jrnl3 Patents

4 IEEE Jrnl1 Eurp Jrnl

24

Thank You

25

How to Reduce the Training Period?

Spanish

French

German

Portuguese

Russian

Greek

Use similarities and correlations between the various languages to

reduce the training period

top related