Continue to Site

Welcome to EDAboard.com

Welcome to our site! EDAboard.com is an international Electronics Discussion Forum focused on EDA software, circuits, schematics, books, theory, papers, asic, pld, 8051, DSP, Network, RF, Analog Design, PCB, Service Manuals... and a whole lot more! To participate you need to register. Registration is free. Click here to register now.

Register Log in

Face Recognition in neural networks

Status
Not open for further replies.
A

ali

Member level 2
Joined
Feb 16, 2003
Messages
43
Helped
1
Reputation
2
Reaction score
1
Trophy points
1,288
Activity points
454
face recognition matlab

Hi everybody,
I recently start to work on face recognition ,and be very happy if
of you share its knowleg with me and let me to know about its
work .

Best wishes.
Ali
 

what's your work about?
specifically,

good viewpoint flatulent jeje

I suspect these steps:
digital image processing (filtering),
image segmentation
some image transforms to enhance aspects
finally a statitical or neural -fuzzy pattern recognition scheme
 

suspect secrecy

I suspect that this work is secret because knowing how it is done will allow better disguise to be made.

Don't forget to add it to he/she to make he/she/it. You never know if someone is a space alien disguised as an earthling. I have been told that you can recognize them by their talking to themselves to keep in practice.
 

I have seen some aliens with two heads. So I suggest an upgrade to he/she/it to he/she/it/they. This also completes the English language requirements.
 

I think you should begin to study anatomic structure of the cranium, musculature of the face.
What is recognition. This is comparing of the image with huge number of storied images plus details of the anatomic structure. All these should be multiplied on the image(contour lines) comparing with different view points.
I think that your face images will have different conditions of light. May be you will work with motion object.
Also your camera resolution and S/N ratio should be according to your task.
And what is with requirement of real time processing?
Very difficult task.
 
crosscorrelation

I suspect that it is done by crosscorrelation with stored images which can be rotated and morphed. My further guess is that parts of the face that are hard to change, like the spacing between the eyes have heavier weighting than easily changed things such as hair style.
 

I saw a documentary about a security system that used a simple webcam to take the picture of te user and then found some 10 - 15 points on the face like nose tip, eyes corners, etc. It then compared teh positions of these with teh ones stored in the database to recognise the face.

Ofcourse, the error due to 3D effects have to be corrected before doing that
 

I think Romanp is right. Some little hurts on face may cause adverse affects on recognition. After analyzing human anathomy, fuzzy logic might help you.
 

Hi

Hi friend..
i think u must use backprooagation algorithm for this.

first u must standart photo example 150 x 150 photo ..
u must use one pixel one neuron input layer.
u may select your hidden layer .. i think 2 layer enought .
and out layer u must add out neuron for every person. how many person must be recoqnize ex .. 5 your outlayer must have 5 ...

u must trained ..
and u may use..

this very easy with matlab + nn toolbox.

best regard ..

if u have prb call me..

murat
 

Hi Ali,
I think that right face recognition should be focused on the comparing of well captured face contours. Information about low frequancy face paterns is important for a next algorithm layers.
By the way, so works the human vision.
Regards
 

**broken link removed**
 

Re: Face Recognition

hi Ali

I try to develop face recognition system, too. I work on DSP from TI. first of all you learned image enhencement techniques then you should start to work on face recognition. face recognition can be implemented by support vektor machines and especially fuzzy-neural approximation. if you would like to wonder image enhencement , you can download "DIPToolBox" from e-book section.

tmk
 

Face Recognition

check the following site
**broken link removed**
 

Re: Face Recognition

Visit citeseer and download papers regarding image processing. Many good papers there that should help you out.
 

Face Recognition

Face recogntition is well done by autocorrelation of the image itself with images stored in a database. The one that match or closesly match is the face you are looking for.

Must have a database full of pictures, then you need also aging of people, definitely some neural network mapping.
 

Re: Face Recognition

The best nethods for doing face matching are not by correlations of pictures ..Otherwise BINLADEN or others might just shave put some sun glasses and they will go undetected ..It is now done with SKULL METRICS
(TRIANGLES and the defining angles of those !!)
those propotions never change and remain constant in human adults
on top the matchings can be done very quickly
 

Face Recognition

I think they go on noise,eyes, upper face area. There are course can be bew methods which world does not yet know in publicity

I think face recognition must be really hard.
 

Re: Face Recognition

VeriLook
 

Face Recognition

this is very good tools
 

Status
Not open for further replies.

Similar threads

Part and Inventory Search

Welcome to EDABoard.com

Sponsor

Back
Top