Biometrics for smartphones: Age recognition, gender recognition and identification

Abstract

Identication, gender and age recognition are directly used especially in the security eldand there are many researches that are actively conducted nowadays about these topics.Sensors are required for this studies. But the large and constrained sensors make itdicult for the experiments to be practicable. On the other hand, the use of smartphonesare becoming increasingly popular and becoming an integral part of our lives. Therefore,the method of providing the necessary sensors via smartphones is increasing. For thisstudy, accelerometers were used which are built-in smartphones and preferred especiallyto recognize motions. In a nutshell, the purpose of this thesis is to present, the studywas done to recognize identication, age, and gender using the accelerometer sensor ina smartphone and the results of these studies. To accurately classify the identication,age groups, and genders that make the activities of walking and running. We also aimedto determine that which motions gives us better success rates in this classication. Inaddition, we aim to compare between the ve dierent classication algorithms. Forthis purpose, to collect data, and to be able to use built-in sensors, compatible Androidand IOS applications have been developed. Also, feature vectors were derived from theMatlab tool and classied by using the Weka Data Mining tool. Five dierent classiershave been studied and the performance evaluation and comparisons were made.Keywords: Biometric, Identication, Classication,Feature, SmartPhone, Sensor, Accelerometer,Application

Kimlik tanma ba³ta olmak üzere, cinsiyet tanma ve ya³ tanma konular özellikle güvenlikalannda do§rudan uygulamalar olan ve günümüzde aktif olarak çal³lan ara³trmakonulardr. Tanmlama yaplrken gerekli olan sensor için ise hayatmzn ayrlmaz birparças haline gelen akll telefonlarn kullanm da artmaktadr. Bu çal³mamzda özellikleeylem tanma için kullanlan ve akll telefonlarda bulunan ivmeölçer sensörü kullanlm³tr.Özetle bu tezin amac telefonlar üzerinde bulunan ivmeölçer sensörünü kullanarak,kimlikleri, ya³ gruplarn ve cinsiyetleri tanmlamaya yonelik yaplan çal³malarve bu çal³malarn sonuçlarn de§erlendirmektir. Yürüme ve ko³ma hareketlerini yapankimlikleri, cinsiyetleri ve ya³ gruplarn do§ru bir ³ekilde snandrabilmek ve yüksekba³ar orannn hangi hareket ile sa§land§nn tespitini yapmaktr. Bununla birliktebe³ farkl snandrma algoritmas arasnda ba³arm kyaslamas yapmaktr. Bu amaçlaveri toplayabilmek için, sensor olarak kullanabilecegmiz, hedeedi§imiz telefon modelleriile uyumlu Android ve IOS uygulamalar geli³tirilmi³tir. Ayrca toplanlan verilerdenöznitelik vektörleri Matlab aracl§yla elde edilmi³ ve snandrma Weka Data MiningTool aracl§yla yaplm³tr. Be³ farkl snf üzerinde çal³ma yaplm³ olup, performansde§erlendirmesi ve kar³la³trmas hedeenmektedir.Anahtar Sözcükler: Biyometri, Kimlik Tanmlama, Snandrma, Öznitelik, AkllTelefon, Sensör, vmeölçer, Uygulama