Independent thesis Advanced level (degree of Master (Two Years)), 20 credits / 30 HE credits
Identification and verification first line of defence of every secure system. Due to the advancement of technology and miniaturization, the use of mobile devices is highly increased since last decade. Nowadays, mobile phones are more powerful than the computers of early days of last decade. Today, mobile phones hold a lot of personal and sensitive data, which must be protected using most reliable, robust, convenient and cost effective authentication mechanisms.
Biometric authentication is one of the three methods of identi cation. Gait authentication is one type of biometric authentication that operates on behavioral characteristics of human beings. There are dierent approaches in gait authentication. In this thesis, we have used wearable sensor based approach and for this purpose Google G1 smart phone is used to collect gait data. This thesis is an attempt tond out the inuence of dierent walking speeds (slow, normal and fast) and surfaces (at carpeted, grass, gravel and inclined) on gait recognition. This gait data is collected from 48 subjects for these six dierent walk settings in two sessions on dierent days to measure same day and cross day performance. Later, dierent mathematical and machine learning concepts are used to analyze recorded data and extract typical gait cycle for each subject. Dierent evaluations are conducted to nd out best parameter settings of these methods.
For different walking speeds, same-day results vary between 14% to 29%, and cross day results vary between 29% and 35%. Similarly, for different walking surfaces, same-day results vary between 9.78% to 39%, and cross day results vary between 28% to 42%. Results of tests conducted for one walk setting clearly reects that slight change in parameters also inuences the results.
2012. , 104 p.
Master of Science - School of Electrical Engineering (EES) - Master of Science - Research on Information and Communication Technologies