Face recognition with neural network

The identification of individuals using face recognition represents a challengingtask with many applications in everyday life as well as in high security applications.

Since the human face will vary in appearance in the short time as well as in the long time
range, the inherent "slack in operation" of neural networks together with the redundancy
and possibility to generalize are suggestive for implementing such a recognition system.
The problem has important applications to automated security systems, lip
readers, indexing and retrieval of video images, videoconferencing with improved visual
sensation, and artificial intelligence. There are two basic face detection techniques:
content-based methods and color-based methods.
Content-based methods try to identify features in a human face. Most content based
methods were developed for gray scale images to avoid the complexity of
combining the features detected in the RGB color space. Unfortunately, content-based
techniques are very complex and expensive computationally. Also, if the face is rotated
or partially obscured, the technique has to incorporate other techniques to solve the image
registration and occlusion problems. In addition, it is often difficult to adapt the methods
to color images.
Face detection using artificial neural networks was done by Rowley. It is robust
but computationally expensive as the whole image has to be scanned at different scales
and orientations. Feature-based (eyes, nose, and mouth) face detection is done by Yow et
al. Statistical model of mutual distance between facial features are used to locate face in
the image. Markov Random Fields have been used to model the spatial distribution of the
grey level intensities of face images. Some of the eye location technique use infrared
lighting to detect eye pupil. Eye location using genetic algorithm has been proposed by
Wechsler . Skin color is used extensively to segment the image, and localize the search
for face. The detection of face using skin color fails when the source of lighting is not
natural. In this paper, motion information is used to reduce the search space for face
detection. It is known that eye regions are usually darker than other facial parts, therefore
probable eye pair regions are extracted by thresholding the image. The eyes pair region
gives the scale and orientation of face, and reduces the search space for face detection
across different scales and orientations. Correlation between averaged face template and
the test pattern is used to verify whether it is a face or not. Recognition of human face is
also challenging in human computer Interaction. The proposed system for face
recognition is based on eigen analysis of edginess representation of face, which is
invariant to illumination to certain extent.

Existing System.

The available or currently using software is unable to find out the bogus
applications. All the information kept in a record or other document. In a passport office,
everyday comes a number of new applications and number of renewals. The applicant has
to fill the form correctly and has to submit to the office staff. The staff will check the
form and enter the information into the system. The number of applicants will be more
and need a database to carry these information.

Proposed System

Addition of this detection facility to the current software. This can be find out the
bogus applications. This will improve the efficiency of the work in passport offices. So
the identification can be very simple and the wastage of time will be low.

System Requirements

The users of the system in one category.
· Administrator
Administrator has the complete authorization to use all the facilities of the
software .Only the administrator should have the previlige to add a new wntry
into the database.

 Operational Requirements

· Software Requirements
Java SDK1.4, Mysql DBMS.
· Hardware Requirements
AMD Athlon XP Processor, 256 MB RAM, 40 GB Hard disk space.
· Operating System
Windows XP or Windows 2000
2.5.3 Functional Requirements
The System should satisfy the following functional requirements.
1) There should be a provision for check whether an applicant has already applied
for passport or not.
2) There should be a provision to view the photo of an applicant.
3) There should be a provision to create a new user.
4) There should be a provision to change the password of a particular user



4.2 Neural Networks
A neural network is a weighted directed graph that models information processing
in the human brain. Each vertex of the network can be thought of as a neuron, while each
edge is a nerve fiber. An input vector X is fed into a set of nodes. An input at a specific
node is passed along a weighted edge, multiplied by the weight, to a neuron in the socalled
hidden layer. If the passed value exceeds a threshold function, the information is
then passed along another weighted edge to an output node, where it must also pass a
threshold function. The sum of all such information from all input nodes gives the output
vector T. Note that this network can be generalized to any number of hidden layers.

Diagram of a generalized neural network.
16
A training data set consists of a set of input vectors X with corresponding output vectors
T. The actual training consists of setting the weights so tha t, for each input X, the output
vector TNET computed by the network closely matches the desired output TACTUAL . Phrased as
optimization problem, we wish to find the collection of weights that minimizes
TNET - TACTUAL , where the norm is understood to be taken over all input-output pairs in the
training set. If the training set is chosen carefully to represent the entire space of possible
inputs, then an input similar to one in the training set should result in a similar output. Note
that training is very expensive computationally, since the determination of the weights must
be done, in some sense, simultaneously for all data in the training set. However, after
training is complete, the computation of an output T for a given input X is very efficient.
For our face detection problem, our input vector X will consist of information derived
from a passport size photo image.
If the system trains the neurons in the proper way, then also there will be chance of
error. The generated output mismatches the predicted output. So the system will fix a
tolerance level, and the error lies between this tolerance interval, then there will be a chance
of coincidence. This system will only be successful up to 70-80%. This is a disadvantage of
using this software.

want more:
Click Here

or
download the full seminar report on face recognition

download the seminar presentation olso!!