*…show more content…*

Facial recognition and detection algorithms work together to improve the accuracy of today’s facial recognition software. The job of today’s computer scientists and mathematicians is to mimic the human eyes and brain’s ability to detect and recognize human faces through an attempt to replicate this complicated process using a series of highly sophisticated algorithms. Infants learn these techniques shortly after birth, and today’s programmers are only beginning to scratch the surface of the possibilities created with this

*…show more content…*

P_I= I ̂/I^* where I ̂ equals correctly identified and I* represents the size of the probe set.

The final comparisons are made by using all pairs for (PI,PJ) and for all of the ROC values which were measured by letting the ROC = R_k/P. As with the algorithms already discussed, the Linear Discriminant Analysis (LDA) model begins with Gaussian data. LDA makes use of two values, the mean and the variance, which are produced for each class. Classes can be thought of as facial features such as chins, noses, eyes, ears, hair lines, hair styles, etc. Four actual pieces of data are calculated during before the final algorithm can be used. The mean value is used to find the muk value.

Muk = 1/nk * sum(x). After this, the variance is found for all classes by using this mean value.

Sigma2 = 1 / (n-k) * sum((x-mu)2). There are two additional steps required before making a final prediction using the LDA method, this paper will only look at the final function.

Dk(x) = x * muk/〖sigma〗^2 -(〖muk〗^2/〖2sigma〗^2 +ln(Plk)) where “Dk(x) is the discriminate function for class k given input x, the muk, sigma2 and Plk are all estimated from your data. (Browniee). Once this discriminant is calculated, this data then can be further manipulated to assist in the facial recognition