Performance Analysis of Two Methods for Dimension Reduction in Face Recognition
Journal Title: International Journal of Research in Computer and Communication Technology - Year 2014, Vol 3, Issue 6
Abstract
Face recognition has been a fast growing, challenging and interesting area in real time applications. This work aims to compare the two renowned techniques of feature dimension reduction on the basis of the classification results of three classifiers used to fulfill the task of face recognition. With the fast increasing quantity and complexity of data in an information-rich age, it becomes difficult, challenging or even impossible for engineers or analysts to deal with raw data directly. Dimensionality reduction provides an efficient way for data abstraction and representation as well as feature extraction. It aims to detect intrinsic structures of data and to extract a reduced number of variables that capture and retain the main features of the high-dimensional data. Principal component analysis (PCA) has long been a simple, efficient technique for dimensionality reduction. A newer nonlinear method local linear embedding (LLE) has been proposed for increasingly complex nonlinear data. In this research work, we investigate and compare linear PCA and nonlinear LLE for face recognition. Experimental results on real-world face database show that these linear and nonlinear methods when compared over performance, the non linear method LLE yield better performance. The classification is carried out using knearest neighbor, probabilistic neural network and support vector machine to identify the face.
Authors and Affiliations
Anika Gugnani, R. B Dubey
Keywords
Related Articles
- EP ID EP27944
- DOI -
- Views 215
- Downloads 0
Butterfly Design for RADIX-4K DIF FFT
VLSI and Digital Signal Processing are the two emerging technologies in the present world. Design of high performance, low area and low power VLSI circuits are needed for the DSP applications. This paper is dealing w...
Accurate Alignment of Search Result Records from Web Data Base
Most search engines are returning results page that are come from structured databases. This type of search engines also referred as web databases. Each results page contains search results records and each record con...
Performance Issues of WMNs by implementing Coordinator in IEEE 802.11 Networks
The wireless networking is the current trend that most of the peoples that used and fastly grow in the world. The number of topologies has been used in this network and few are Star, Bus , Mesh etc. but the Mesh Topol...
Echo Noise Reduction Using Dynamic Selection Of Sub Band Adaptive Filters
Versatile filters assume a paramount part clinched alongside advanced day sign transforming with provisions for example, such that commotion cancellation, sign prediction, versatile reaction cancellated What's more r...
A Novel Technique for Creative Problem-Solving by using Q-learning and Association algorithm
for a single problem there are multiple solutions or multiple ideas from different users. Means every one gave an idea (solution) for solving the given problem if we distribute that particular problem to the different...