|
|
|
|
|
|
Hybrid Algorithm with Variants for Feedforward
Neural Network
Thinakaran Kandasamy1 and Rajasekar Rajendran2
1Sri Venkateswara College of Engineering and Technology, India
2Excel Engineering College, India
Abstract: Levenberg-Marquardt back-propagation algorithm, as a Feed forward Neural Network (FNN) training method, has some limitations associated with over fitting and local optimum problems. Also Levenberg-Marquardt back-propagation algorithm is opted only for small network. This research uses hybrid evolutionary algorithm based on PSO in FNN training. This algorithm includes a number of components that gives advantage in the experimental study. Variants such as size of the swarm, acceleration coefficients, coefficient constriction factor and velocity of the swarm are proposed to improve convergence speed as well as to improve accuracy. The integration of components in different ways in hybrid algorithm produces effective optimization of back propagation algorithm. Also, this hybrid evolutionary algorithm based on PSO can be used for complex neural network structure.
Keywords: Back propagation, hybrid algorithm, Levenberg-Marquardt, Particle swarm optimization, variants of PSO algorithm
Received August 31, 2014; accepted April 12, 2015
|
|
|
|
|
|
A Fuzzy Based Matrix Methodology for Evaluation
and Ranking of Data Warehouse Conceptual Models Metrics
1Naveen Dahiya,
2Vishal Bhatnagar, and 3Manjeet Singh
1Maharaja
Surajmal Institute of Technology, C-4, Janakpuri, India
2Ambedkar
Institute of Advanced Communication Technology and Research, India
3YMCA
University of Science and Technology, Sector-6, India
Abstract: The authors present a
methodology for ranking data warehouse conceptual models metrics based on
opinion of experts using fuzzy inference technique. The fuzzy based approach
gives a precise ranking methodology due to its ability to handle imprecise data
involved ranking of metrics and ambiguity involved in expert decision making
process. The proposed work aims towards ranking of quality metrics already
proposed and validated by Manuel Serrano along certain identified parameters
based on expert opinion and evaluation of criteria matrix using permanent
function. The results obtained are also compared with the actual experts
ranking. The achieved results are better as the imprecise human thinking is
taken into consideration during calculation of results to give realistic
results.
Keywords: Fuzzy, Data warehouse, Conceptual models, Quality
metrics, Criteria matrix.
Received October
23, 2014; accepted July 7, 2015
|
|
|
|
|
|
|
An Optimized Model For Visual Speech Recognition Using
HMM
Sujatha Paramasivam1
and Radhakrishnan Murugesanadar2
1Department of
Computer Science and Engineering, Sudharsan Engineering College, India
2Department of
Civil Engineering, Sudharsan Engineering College, India
Abstract: Visual Speech Recognition (VSR) is to identify spoken
words from visual data only without the corresponding acoustic signals. It is
useful in situations in which conventional audio processing is ineffective like
very noisy environments or impossible like unavailability of audio signals. In
this paper, an optimized model for VSR is introduced which proposes simple
geometric projection method for mouth localization that reduces the computation
time. 16-point distance method and chain
code method are used to extract the visual features and its recognition
performance is compared using the classifier Hidden Markov Model (HMM). To
optimize the model, more prominent features are selected from a large set of
extracted visual attributes using Discrete Cosine Transform (DCT). The
experiments were conducted on an in-house database of 10 digits [1 to 10] taken
from 10 subjects and tested with 10-fold cross validation technique. Also, the
model is evaluated based on the metrics specificity, sensitivity and accuracy.
Unlike other models in the literature, the proposed method is more robust to
subject variations with high sensitivity and specificity for the digits 1 to
10. The result shows that the combination of 16-point distance method and DCT
gives better results than only 16-point distance method and chain code method.
Keywords: Visual speech recognition, feature extraction,
discrete cosine transform, chain code, hidden markov model.
Received March
20, 2015; accepted August 31, 2015
|
|
|
|
|
|
|
Progressive Visual Cryptography with Friendly and Size
Invariant Shares
Young-Chang Hou1,
Zen-Yu Quan2, Chih-Fong Tsai2
1Department of Information
Management, Tamkang University, Taiwan
2Department of Information Management,
National Central University, Taiwan
Abstract: Visual
cryptography is an important data encoding method, where a secret image is
encoded into n pieces of noise-like shares. As long as there are over k shares
stacked out of n shares, the secret image can be directly decoded by the human
naked eye; this cannot be done if less than k shares are available. This is
called the (k, n)-threshold visual secret sharing scheme (VSS). Progressive visual
Cryptography (PVC) differs from the traditional VSS, in that
the hidden image is gradually decoded by superimposing two or more shares. As
more and more shares are stacked, the outline of the hidden image becomes clearer.
In this study, we develop an image sharing method based on the theory of PVC,
which utilizes meaningful non-expanded shares. Using four elementary matrices (C0
- C3) as the
building blocks, our dispatching matrices (M0 - M3) are designed to be expandable so that the
contrast in both the shares and the restored image can be adjusted based on
user needs. In addition, the recovered pixels in the black region of the secret
image are guaranteed to be black, which improves the display quality of the
restored image. The image content can thus be displayed more clearly than that
by previous methods.
Keywords: Visual Cryptography, Progressive Visual
Cryptography, Secret Sharing, Unexpanded Share, Meaningful (Friendly) Share.
Received April 8,
2015; accepted October 7, 2015
|
|
|
|
|
|
|
Real Time Facial Expression Recognition for Nonverbal
Communication
Md. Sazzad Hossain1
and Mohammad Abu Yousuf2
1Department of Computer Science and
Engineering, Mawlana Bhashani Science and Technology University, Bangladesh
2Institute of Information Technology, Jahangirnagar
University, Bangladesh
Abstract: This paper
represents a system which can understand and react appropriately
to human facial expression for nonverbal communications. The considerable
events of this system are detection of human emotions, eye blinking, head
nodding and shaking. The key step in the
system is
to appropriately recognize a human face with acceptable labels. This system
uses currently developed OpenCV Haar Feature-based Cascade Classifier for face
detection because it can detect faces to any angle. Our system can recognize
emotion which is divided into several phases: segmentation of facial regions,
extraction of facial features and classification of features into emotions. The
first phase of processing is to identify facial regions from real time video. The second phase of
processing identifies features which can be used as classifiers to recognize
facial expressions. Finally, an artificial neural network is used in order to
classify the identified features into five basic emotions. It can also detect
eye blinking accurately. It works for the active scene where the eye moves
freely and the head and the camera moves independently in all directions of the
face. Finally, this system can identify the natural head nodding and shaking that can be recognized in real-time using
optical flow motion tracking and find the direction of head during the head
movement for nonverbal communication.
Keywords: Haar-Cascade Classifier, Facial Expression, Artificial Neural Network, Template matching,
Lucas-Kanade Optical flow.
Received April 4, 2015; accepted November
29, 2015
Full text
|
|
| |
|
