Evaluafion and Development of Strategies for Facial Features

Comments

Transcription

Evaluafion and Development of Strategies for Facial Features
 Evalua'on and Development of Strategies for Facial Features Extrac'on for Emo'on Detec'on by So;ware Proposal presenta4on. February 27th, 2015 Carolina González-­‐Restrepo Sebas4án Rincón-­‐Montoya Advisors: Olga Lucia Quintero-­‐Montoya -­‐ GRIMMAT René Restrepo-­‐Gómez – Applied Op4cs Daniel Sierra-­‐Sosa – Applied Op4cs Outline 1.Concepts 2.Problem Descrip4on 3. Approach 4.Timeline 5.Results 6.Objec4ves 7.Different Approach 8.Applicability 9.References Useful Concepts Set of propor4ons most people follow. (Ricke?s, 2002) Beard Bang Glasses Skin tone Facial Canon “Noisy Images” x
x
x
x
x
x x x
Par4al feature detec4on Establishment of feature marks Global Vision of the Problem Pre processing Ideas Picture I
Anfis A Anfis Results Emo'on Map Processing Voice Audio Signal Russell’s taken from hfp://blogatufotondo.altervista.org/category/adolescen4/ Progress Carolina Gonzalez R Sebas'an Rincon M Improvement for Feature Detector Susana Mejía A Emo4on Detector for Children (Mejía & Quintero & Castro, 2011) 2013 2001 2011-­‐2012 Viola Jones Face recogni4on (Viola & Jones, 2004) Sebas'an Isaza M Feature Detector 2014 GRIMMAT Emo4on Detector based on Audio Signals Carolina Gonzalez R Sebas'an Rincon M Improvement for Feature Detector in Noisy Images NOW Jose Daniel Mejia Z Feature Detector based on a physical approach The current research will start with a compila4on of informa4on, collected from previous work of the people men4oned above. Focusing on the most promising results and untested ideas. Why us? Ability to master certain areas Mathema4cal Logic How? Valida4on and op4miza4on skills Inves4ga4ng the most recent advances Understanding the problem with a mathema4cal
approach Studying diverse topics. ex: Fourier spectrum analysis Filters Image Processing Valida4ng the results Finding errors to probe the precision of the algorithm Re-­‐training other’s algorithms Past Results x x
x
x
x
x
x
x
x
x
x
x
x x x
x
x
x
Pre-­‐processing x
x
x
x
x
x
xx
x x
x
x
x
x
Canon’s propor4ons Lack of accuracy while stablishing marks Objec4ves To improve the methodology used in the previous research prac4ce, seeking for a more robust algorithm to extract features from noisy images. • 
To evaluate different facial feature extrac4on techniques, previously used, in a larger data base. • 
To search the state of art, looking for new techniques in feature extrac4on; and evaluate them. • 
To study different pre-­‐processing techniques for images, in order to strengthen the algorithm in feature extrac4on. • 
To design and use filters in images, to minimize the noises that difficult the feature extrac4on and therefore the emo4on detec4on. • 
To compare the results obtained in the current research and the ones obtained with the physics engineering approach. • 
To verify whether different features can be treated as noise. Other’s approach •  Op'cal Image processing Signal processing for which the input is an image, and the output may be a modified image or informa4on related to it. •  Biometrics Refers to technologies for measuring and analyzing human body characteris4cs. •  FACS (Facial Ac'on Coding System) (Ekman & Friesen, 1977) Tool used to measure facial expressions based on compound facial movements. Applica4ons There is psychological evidence that proves that changes in emo'ons can lead to a possible lie. Academic Purposes Governmental (Kroll, 2013) Social Rela4onships Costumer Service Business Area (T. Interna'onal, 2013) References Ricke?s, R. (2002). La divina proporción. Goldstein R. Odontología esté'ca, principios, comunicación, métodos terapéu'cos. 193-­‐21. Viola, P., & Jones, M. J. (2004). Robust real-­‐'me face detec'on. Interna'onal journal of computer vision, 57(2), 137-­‐154. S. Mejía, O. Quintero, and J. Castro. (2011). “Analysis of emo'on: An approach from ar'ficial intelligence perspec've”. Ekman, P., & Friesen, W. V. (1977). Facial ac4on coding system. Kroll. (2013). “2013/2014 Informe Global sobre Fraude”. T. Interna'onal. (2013). “Global corrup'on barometer 2013: Report”.