Mathematical – Facial Recognition
The result corresponds to about 1 chance in 13,000,000, or getting 24 consecutive heads for true coin flips, that Lucy I and Lucy II are the same person.
DR. BENNETT FACIAL RECOGNITION REPORT
We all have the natural God-given ability to recognize faces, a social necessity.
This recall capability is hidden internally, in our fuzzy memory of faces and not
subject to any objective external review.
But facial recognition from surveillance video, along with automated processing of
facial characteristics, allows modern software to measure, analyze and identify faces in a crowd in
real time, with high probability of success. Unlike our own visual process, automated recognition
techniques follow a known and accepted set of statistical rules. So they are objective,
reviewable and repeatable by anyone…consistent with the quantitative knowledge and sharing
of evidence required by the scientific method.
What follows is a forensic application of the same professional statistical methods
of facial recognition to two sets of photos that purport to be of the same person….the only
difference being the report computations were done manually.
From the report
Tabular computation of total probability that I = II The probability of all 6 independent facial length ratios occurring is the product of each probability, times the standard error of this estimate….sqrt(n) = sqrt(6) = 2.5 So the best estimate for the likelihood that Lucy I = Lucy II is: 2.5 * 3 x 10^-8 = 7.5 * 10^-8 = 1/(13 * 10^6) = 1/13 million Summary All ratio measurements are computed by the t test and then converted to probability estimates in the last column above. We see that the mean a/b ratio of nose length to philtrum 1 is identical for Lucy I and II; the probability is 1. Were this the only comparison, the conclusion would be that the same woman was photographed. The ratio of nose length to eye width, a/e, is the least likely match of the photo set and also the most recognizable of the face. Conclusion The result corresponds to about 1 chance in 13,000,000 …. or getting 24 consecutive heads for true coin flips…. that all the photos had the same subject. Lucy I <> Lucy II … 13 million to 1 !
