Recognition Using PCA Face Recognition,

¶ … Recognition using PCA

Face recognition, an idea that was once science fiction, is now science fact, as computerized facial recognition technology has impacted the world. The nascent history and reason for the facial recognition technology is varied. Recent research into the nuances and the usage of face recognition technology is explained by Bulkeley (1999).

In Texas, individuals unable to meet reserve requirements regarding bank accounts can instead cash M1 money supply such as cash at locations including kiosks "that scan their faces and dispense money." (Bulkeley, pg. 1, 2010) A community in London declares a 40% decrease in the crime rate after equipping their "surveillance-camera network with face-recognition software trained to identify neighborhood criminals." (Bulkeley, pg. 1, 2010) In areas in and around Tel Aviv, law enforcement officers are commencing with the deployment of a "system that uses a combination of computerized face recognition, hand-geometry scanning and "smart" cards in an effort to speed the passage of Palestinian day laborers from Gaza into Israel." (Bulkeley, pg. 1, 1999)

The need for facial recognition technology appears to be a function of crime prevention and surveillance to monitor transportation of goods and people. With the number of crimes ranging from bank robbery to grand larceny, the ability to identify individuals using biometric scanning techniques may enable a revolution in crime prevention and security. The implication is that the development of biometric technology will perfect the recognition matrix to include gradient identification and symmetric/asymmetric facial features that can distinguish fraternal and identical twins.

The technology to perform face recognition was brought about commencing in the 1970s and through the increasingly short half-life of the advancement in technology, supporting advancements to the architecture was brought to market, with the current ability of image processing using vector referencing and facial pattern identification and recognition developed most recently, biometrics has proliferated into systems that include applications as far reaching as cyborg communication and retina scanning technology (Liu, Chen, Lu, Chen, 2008). Up till now, many kinds of face recognition methods have been proposed by earlier researchers and among these methods, they can be mainly divided into two classes, constituent-based and face-based." (Liu, Chen, Lu, Chen, 2008)

Additionally, the need to improve on the technology for surveillance purposes will prevent the misidentification of individuals whom share similar facial features. There have been wrongful sentences handed down by circuit court rulings, which have identified innocent citizens as the defendant in a crime, committed by a family or community member with identical physical features. The need to perfect facial identification scanning techniques is to prevent the wrongful accusation of innocent citizens.

The cornucopia variety in applications for the face recognition does have its limitations. Current quandaries to the operations of face recognition however, include mathematically mapping, referencing, and cross referencing the unique features of each human face. These issues are further explained by Neerja & Walia (2008) "This is due to the diversity in the nature of face images, such as variations in head pose, illumination, size, scale, orientation, expressions, occlusions and data capture." (Neerja, Walia, 2008)

Facial identification and recognition remains elusive under the conditions of pose, illumination, database size etc., (Neerja, Walia, 2008), however, still attracts significant research efforts (Neerja, Walia, 2008). "

The main reasons of ongoing research are its many real world applications like human/computer interface, surveillance, authentication, perceptual user interfaces and lack of robust features and classification schemes for face recognition task. Principal Component Analysis (PCA) is a typical and successful face-based technique. Turk and Pentland developed a face recognition system using PCA in 1991. Belhumeur et al. proposed Fisherface technique based on Linear Discriminant Analysis (LDA) in 1997." (Neerja, Walia, 2008)

PCA is a technology that has advanced since the early advent of facial recognition technology. Described as cost efficient and garnered for its 'bang for the buck', PCA has applications where there is a need for cost efficiency and a rather heterogeneous population. Quick surveillance techniques that apply PCA can search the footage against the contours of the profile and determine whether there is a match. PCA can be useful in bank heist applications and other areas of grand larceny including museum burglaries.

Under normal conditions, algorithms used in facial recognition identification are classified inclusively rather exclusively into two areas, knowledge-based & learning based (Lekshmi, Sasikumar, 2009). "Knowledge-based, in which prior knowledge of facial patterns are depicted with some explicit rules, and learning-based, in which face patterns are modeled with distribution functions under a probabilistic network." (Lekshmi, Sasikumar, 2009)

The herculean task to translate intellectual capital into a searchable format of which to identify and match data, which is the knowledge-based method (Lekshmi, Sasikumar, 2009) offers more problems than solutions and more cost than benefit. "Support Vector Machine (SVM), Principal Component Analysis (PCA), Artificial Neural Network (ANN), Bayesian-rule method, and Fisher Linear Discriminant Analysis (LDA) are examples of the learning-based method. Here, face patterns are determined by learning models or training samples. These methods can handle more complex cases, as compared to knowledge-based methods." (Lekshmi, Sasikumar, 2009)

Knowledge based applications are known to be less costly and work well with the aforementioned approach of bank heist profiling using the PCA. The learning-based approach involves artificial intelligence, and neural networks that function on advanced algorithms that determine precise changes in depth and concavity with regard to how a face should move and respond given a specific facial movement and then monitor the variance to determine reliability of data according to whether the face matches what is predicted by the intelligence. Examples that include knowledge and learning-based approaches are mentioned below.

"Advances in face recognition come amid heightened interest in using biometrics such as iris scanners, hand-geometry readers, fingerprint sensors and voice-recognition systems for everyday identification. This is mainly because of a widespread belief that passwords and PINs are easy to compromise. And face recognition is relatively palatable to the general public: It doesn't have the stigma of fingerprinting, the strangeness of sticking your hand into a hand-geometry reader or the spookiness of eyeball reading." (Bulkeley, pg. 1, 1999)

A facial recognition company based in N.J. is operated by a scientist whom has many years of experience with face recognition systems development. He investigated facial recognition algorithms by supposing how the human brain via neurotransmitters and neurochemicals can process what he labels as roughly 500 books in a matter of seconds. Duplicity of information he blames as the cause for errors in judgment. Should one discard the erroneous and duplicative information then the notion of less errors becomes rather apparent.

"Visions Inc., a small Jersey City, N.J., outfit, says its face-recognition system on a powerful PC can compare one photo with 60 million pictures in a database and select a match. "We've gone a lot further than the human brain, which only needs to recognize a few members of society," says Visionics' chief executive officer, Joseph Atick, who is a former mathematics professor at Rockefeller University." (Bulkeley, pg. 2, 1999)

Dr. Atick and associates could not comprehend the manner of how individuals were able to processes the incredible quantity of data information as visual inputs (Bulkeley, pg. 2, 1999) "A human is bombarded by data equivalent to three to four books a second," he says. The scientists concluded we must automatically eliminate redundant information and remember only unique features that are different. That led him to develop computer models of facial characteristics, or landmarks, that would differentiate them from all other faces." (Bulkeley, pg. 2, 1999)

The research supposition a high limit range of "80 landmarks" (Buckeley, pg. 2, 2010) covering the entirety of the face. Additionally the requirements to conclude a positive match, the algorithm need only, according to Buckeley (2010), "find 14 points that are alike, usually located where the curvature of the face changes. Once the computer creates a template of the face, it can search a photo database for a match. The computer ignores changeable characteristics like hair color and style or facial expressions, but its focus on the immutable has a major flaw: It can't differentiate between identical twins." (Buckeley, pg. 2, 2010)

The duplicity was eliminated by Visionics and the resulting 80 landmarks (Buckeley, pg. 2, 2010) reduced the overall variability tremendously by identifying the most critical areas to enable facial identification on the most consistent basis. However, the one flaw in Visionics ability to identify individuals using facial recognition is its inability to use the knowledge-based approach to discern between identical twins.

The 14 point algorithm however, provides the basis to the critical link of the 80 landmarks regarding the facial expressions and contact points regarding movement. These algorithms are enabled by large databases that contain copious information regarding facial profiles. These algorithms are mentioned by Carts-Power, below.

First-generation algorithms (Carts-Power, pg. 127, 2005) derived for use within large databases (Carts-Power, pg. 127, 2005) include Eigenfaces, Fisherfaces, Bayesian intrapersonal/extrapersonal classifiers, and elastic bunch graphing (Carts-Power, pg. 127, 2005). Eigenfaces, or principal component analysis (PCA) (Carts-Power, pg. 127, 2005), built and refined at the Massachusetts Institute of Technology (MIT; Cambridge, MA) in the mid-1990s (Carts-Power, pg. 127, 2005).

An Eigenface representation (Carts-Power, pg. 127, 2005) created using primary "components" (Carts-Power, pg. 127, 2005) of the covariance matrix of a training set of facial images (Carts-Power, pg. 127, 2005). This method converts the facial data into eigenvectors projected into Eigenspace (a subspace), (Carts-Power, pg. 127, 2005) allowing copious "data compression because surprisingly few Eigenvector terms are needed to give a fair likeness of most faces. The method of catches the imagination because the vectors form images that look like strange, bland human faces. The projections into Eigenspace are compared and the nearest neighbors are assumed to be matches." (Carts-Power, pg. 127, 2005)

The differences in the algorithms are reflective in the output of the resulting match or non-match of real facial features against the biometric database or artificial intelligence generated via algorithm. The variances generated by either the Eigenspace or the PCA will vary according to the use of the approach. Eigenspace work on the premise of vectors, contours, and gradients, which are all essentially geophysical descriptors used in earth science technology. However, the human face is very similar to a geophysical landscape, similar to an arid desert with hills, valleys, and peaks.

Many regard the principle component analysis (PCA) or eigenface approach (Liu, Chen, Lu, Chen, 2006) as highly beneficial. As such, the industry early on has relied on "PCA-based face recognition systems" (Liu, Chen, Lu, Chen, 2006). The PCA approach is able to locate variances in the details and intricacies when reviewing the "scaled and aligned human face, but it will degrade dramatically for not-aligned faces." (Liu, Chen, Lu, Chen, 2006) The prevailing over the limit of this approach (Liu, Chen, Lu, Chen, 2006), is what Liu et al. regard as "a better method named independent component analysis (ICA) is presented" (Liu, Chen, Lu, Chen, 2006), developed to find "basis functions which are local and give good representation of face images." (Liu, Chen, Lu, Chen, 2006)

Issues with parametric modeling of the facial sub-features hidden from shading issues are posed for solution. The use of PCA to solve this issue (Zhao, Chellappa & Rosenfeld, Phillips) provides a means to create a mathematical framework to identify the hidden parameters where the shadow subspace is shading.

Principal Component Analysis (PCA) (Zhao, Chellappa & Rosenfeld, Phillips), recommended as an enabler for to render a solution to the "parametric shape-from shading (SFS) problem." (Zhao, Chellappa & Rosenfeld, Phillips) "An eigen-head approximation of a 3D head" (Zhao, Chellappa & Rosenfeld, Phillips) "was received after training on about 300 laser-scanned range images of real human heads." (Zhao, Chellappa & Rosenfeld, Phillips) The SFS quandary described by Zhao et al. morphs to a "parametric problem" (Zhao et al.) however, "a constant albedo is still assumed." (Zhao et al.) "This assumption does not hold for most real face images and it is one of the reasons why most SFS algorithms fail on real face images. To overcome the constant albedo issue, suggests including the use of a varying albedo reflectance model." (Zhao, Chellappa & Rosenfeld, Phillips)

In the face of stellar results performed by the PCA, this approach has now been understood to possess the "disadvantage of being computationally expensive and complex with the increase in database size" (Neerja, Walia, 2008), as each pixel in the entire image in aggregate, is required to generate representation needed "to match the input image with all others in the database." (Neerja, Walia, 2008)

Neerja & Walia put forth a "new PCA-based face recognition approach" (Neerja, Walia, 2008), "using the geometry and symmetry of faces, which extract the features using fast Fuzzy edge Detection to locate the vital feature points on eyes, nose and mouth exactly and quickly." (Neerja, Walia, 2008) With regard to each feature, each subgroup repository for database images are created. "During recognition only the images falling in same group as test image, will be loaded as image vectors in covariance matrix of PCA for comparison." (Neerja, Walia, 2008)

The aforementioned approach is expensive, however such governmental agencies including the FBI, CIA, and departments such as the DoE, DOD, and the Secret Service will use these approaches to ensure that the SFS problem is eliminated. Additional algorithms are described below.

"The Fisherfaces algorithm, also known as linear discriminant analysis (LDA), was developed at the University of Maryland (College Park, MD)." (Carts-Power, pg. 127, 2005) This method is akin to the PCA application, however incorporates addendums that accentuate the differences between faces as more evident. (Carts-Power, pg. 127, 2005) "Instead of looking for the nearest neighbor in a subspace (like PCA and LDA), the Bayesian intrapersonal/extrapersonal classifier looks at the distance between two face images." (Carts-Power, pg. 127, 2005) Each differing image may undergo reclassification into two classes as they either are a function of "two images of the same subject or derived from images of different subjects." (Carts-Power, pg. 127, 2005) Each of the aforementioned classes will unfold as a distribution that is Gaussian (Carts-Power, pg. 127, 2005) in appearance. The Gaussian distribution can have layered results without obfuscation. (Carts-Power, pg. 127, 2005)