System And Method For Creating Individualized Mobile and Visual Advertisments Using Facial Recognition

摘要

A computer implemented advertisement method for matching a user with a computer generated advertisement that is compatible to a users of the methods preferred aesthetic. In one example the method comprises: receiving a profile generated by recording user choices from a field of images displayed to the user; applying a logistics regression model to the image choices to create a baseline aesthetic template used to determine the most appropriate model or actor to be inserted into the background of a still or video advertisement. Any results within a desired confidence level are returned to the user as and advertisement with the appropriate model or actor inserted by means of current and available technology into the ad to be displayed on the user's mobile or video device

主权项

1- A computer implemented predictive advertising method, for matching a user with one or more aesthetically pleasing models or actors to be inserted directly into ads directed to the individual user on mass marketing level, operable on a server comprising a processor and memory and connected to a mobile network and device, comprising:
Recording a user's inputs from a field or page of multiple thumbnail or other visual images performing a facial recognition process on the image and storing the result of the process on the server as a template on a unique server with the intention of using the stored template as a reference in generating and optimizing a baseline template or model of the aesthetic preferences of the user; receiving a template image comprising an image of a human face, generated from a statistical valid set of models and actors, and storing a result of the process on the server for matching and creating advertisements with users of the method; user is expected to choose photographs of other users of the method out of field, most likely thumbnails on a mobile device for purposes of creating and further enhancing a baseline aesthetic model; performing facial recognition processes amongst the chosen facial images and creating a template for each including additional data such as ordinal choice in field, time between screen populated and choice, time between second choice, third, number of choices per screen and number of pictures actually chosen and storing the process on the create baseline template for a user's aesthetic ideal using a facial recognition morphing function using all chosen pictures; creating a predictor model using binomial logistic regression weighting each image according to above measures, ordinal, time, range, so that morphing function creates a more accurate baseline facial template; determining goodness of fit using R2 processing each co-efficient through a statistical package to determine the Wald Statistic:Wj=Bj2SEBj2or Likelihood Ratio Test
applying the following binomial regression equation to each subsequent template to determine predictability of match:L(Yμ)=∏i=1n(1yi=1(μi)+1yi=0(1-μi)),;where L=likelihood of an event occurring, Y=attraction given M or the mean of the users choices. The second half of the equation describes in more detail how this is determined through the use of multivariate regression, 1yl representing a yes choice in a binomial system and 1yl=0 being the not set or false.
user then has one template of aggregate, preferred facial characteristics that will be used to compare to the set or sets of human models and/or actor's templates previously created for use in determining an appropriate type or image to be used in an individualized mass advertisement campaign; alternatively a template is created by determining mode of choices and morphing baseline template from all applicable images; determining mode and mean of choices and running standard deviation of each with which to create sample population for user which is then applied and updated to subsequent image fields; recording future choices and repeating the process to fine tune baseline facial template; processing database of users and returning field of images to querant within one degree of standard deviation from mean or mode;processing each image using threshold and predictability of matching variables at certain confidence intervals utilizing; Y*=Xβ+ε, Where Y=the response (confidence of attraction), X=predictive variable (time,face match,location), e=error rate and the B is the function feature general set to 1 if unknown.
recording choices of user and repopulating field of images with tightening degree of accuracy based on process in claim one; assigning chosen each image a z score; determining most accurate method of matching by comparing subsequent choices using either binomial or latent regression models template match's mean, mode or other factors and assessing method best suited to predict user compatibility by then creating binomial distribution table to determine potential for mutual compatibility The method of claim 1, the method further comprising: aggregate outlying data to determine if there is any secondary or tertiary facial type's user would be attracted to and applying claim one to these to determine if there is any alternate template possible to that of the baseline;