Abstract –Face detection is an important part of face recognition system used in computer vision. There are many challenges for effective and efficient face detection. In this paper, the framework for efficient face detection using fusion of PCA and Artificial neural network is presented. The image features are represented as reduced features space by using PCA which is a dimensionality reduction technique. Further these features are given as input to the ANN for training. We have used multilayer perceptron network for accomplishing this task. Keywords:- Face detection, LDA, PCA, ANN, face recognition , SVM, Adaboost. I. INTRODUCTION
Human Face detection is the process of identifying the features of faces to detect the faces on the basis of the discriminant features. Features of faces are eyes, ears, eyebrows, nose, lips, hairs, chicks, forehead etc. Face detection can be carried out using these features of faces. Face is important part to identify the person. It can be used as the computer visual application. Face is the important part of our body by which it is easy to identify and recognize the person. Face detection is one of the challenging tasks as there are many issues such as changes in the appearances of faces, variations in poses, noise, distortion and illumination condition. Complications occur in discriminating the two identical faces for example in case of twins. There are several techniques for face detection that exist in the literature. Principal discriminant analysis (PCA)  and Linear discriminant analysis (LDA)  are most commonly used techniques for face detection. Handsdorff distance measure for face recognition , Elastic Graph Matching (EGM) , eigenspace-based face recognition [ 4], a novel hybrid neural and dual eigen spaces methods for face recognition , eigenfaces and Fisherfaces methods . In order to capture the frontal face image accurately and timely, many face detection methods have been proposed, such as face detection in color images based on the fuzzy theory
, the discriminating feature analysis and Support Vector Machine (SVM) classifier for face detection , neural network-based face detection . Face color information is an important feature in the face detection. In reference , a latest survey of skin-color modeling and detection methods was presented. Statistical color modules with application to skin detection was reported in reference . The quantized skin color regions for face detection were given in reference . Eye is another important feature for face detection and recognition. For example, a robust method for eye feature extraction on the color image was reported in reference . Using optimal Wavelet packets and radial basis functions for eye detection was introduced in reference. Face detection has the application in variety of fields used in the teleconference system, medical imaging intelligent video surveillance, smart cards and security based public places such as airport and railway station. The other face detection techniques are Adaboost, fisher technique, float boost etc. Over the past decade, many approaches to improve the performance of face detection were proposed. These approaches are categorized into two types : 1) Knowledge-based approach, and 2) Feature invariant approach. Knowledge-based methods use a priori rules to carry out face detection, such as the face is usually symmetrical with each other eyes. Feature invariant approach include extraction of features, filtering of images with respect to size , noise, distortion illumination etc. Face detection is important considering the new medical science research such as plastic surgery. It has become challenging to detect and recognize face many due to the changes and variation in the face that occurs due to aging, hormonal changes, emotion changes ,skin color changes etc. In this paper the different techniques used in the face detection are explained. Review of recent face detection method is studied in this paper.
Adaboost technique for face detection was introduced by viola and jones. They proposed a training classifiers to detect the discriminating features of the faces. A framework was introduced such that the training classifiers checked the faces in the cascading style. They used the boosting method in their work by which they tried to detect the faces using the discriminating features. As shown in the fig.3 c0,c1,c2…cn are the classifiers in the cascading manner. The input is provided to the first classifier. The first classifier checks whether the input is face or non-face object .If it is face object then it is passed to the next classifier otherwise it is discarded as non-face object. The process of testing faces by several classifiers continues till it passes to the final node. The output is the face object only if it passes through every node. In this method much of the effort has been spent on improving the boosting method. But the major drawback of Adaboost method is that training classifiers reject the most of the non-face objects before reaching the final node. Secondly the accuracy in the detecting the faces was not much as expected. Adaboost method takes much time to check the face and non-face object Thus time complexity is also a major issue in the Adaboost technique as more number of classifiers are used. Still many analyst used this method in their work.
Fig.3. Cascading of boosted classifier
The existing algorithm based on Adaboost algorithm suffers the problem of computation complexity for training of weak classifier. However some researchers have proposed much simpler approach for face detection without using Adaboost algorithm. Padma Polash Paul and Marina Gavrilova presented PCA based geometric modeling for automatic face detection.They have presented PCA base modeling of geometric structure of the face foe automatic face detection. They have used skin color modeling for efficient feature extraction.
Alireza Tofighi and S.Amirhassan Monadjemi proposed face detection recognition using skin color and Adaboost algorithm  combined with Gabor features and SVM Classifier. They have proposed this technique to enhance the performance of face detection and recognition system. This method consist of detection of faces and recognition of detected faces. In detection step they have used skin color segmentation with gosian skin color model combined with Adaboost Algorithm. Further the series of morphological operators are used to improve the face detection performance.
P.E Robinson introduced the comparison of PCA AND LDA techniques of face detection  by implementing the nearest neighbor classifier. He used the set of database containing the face images in different rang of poses, lighting condition and occlusion. Then he compared performance of face detection using PCA and LDA. In PCA, a training set is used to create a covariance matrix containing the strongest eigenvalues and thus form a vector face also called as Face Space. Secondly the face to be recognized will be is converted into vector and is compared with the nearest closest face class present in the dataset .The result obtained will be the face that is to be detected. It was noted that PCA has better compact detection rates and less processing time as compared to PCA AND LDA combined work.
Sanjay A. Pardeshi and Dr. S.N. Talbar  introduced a automatic face detection technique. In this method first they extracted the fiducial face point of the color images. Then using Principal component Analysis (PCA) technique tried to reduce the dimensionality of Gabor filter .The face is recognized by measuring the similarity between the Gabor filter (jets) in the Eigen faces. The proposed work showed that the automatic face detection recognition system is better for machine face recognition. The result showed 72% face detection rate is acquired using the 30 facial feature point with L-1 norm and COS distance metrics but all distance metrics are necessary to be equally good.
According to Khalid Mohamed Alajel, Wei Xiang, and John Leis face can be detected using the skin color. They divided the process in three stages .First image acquisition is done in which the set of color images containing the face and non-face images .Secondly they using skin color segmentation non-face region in the color images are detected using the low probability region. Finally the modified Hausdorff distance is used to detect the faces using output of skin color segmentation. The performance of this algorithm achieved 87.5 % accuracy in detection faces.
Kanhaiya kumar presented artificial neural based face detection using Gabor feature Extraction .They have discussed the desirable characteristic of spatial locality and orientation selectivity of the Gabor filter The feature vector based on Gabor filter is used as the input of the classifier which is the feed forward neural network on reduced feature of space.
Karhunen introduced the method of Principal Component Analysis(PCA) technique of detecting the faces. Principal Component Analysis(PCA) is used t reduce the redundancy and used to reduce minimize the distance between the data space and projection space. The merit of Principal Component Analysis(PCA) is that it can detect the faces even in presence of random noise and rotation scaling translation distortion.
Figure1. Images of distortion due to pose, expression, noise , illumination.