Use this URL to cite or link to this record in EThOS:
Title: Human gait analysis for biometric identification and authentication
Author: Balisane, Hewa
ISNI:       0000 0004 2710 6921
Awarding Body: Manchester Metropolitan University
Current Institution: Manchester Metropolitan University
Date of Award: 2011
Availability of Full Text:
Access from EThOS:
The study of biometrics is concerned with any human physiological or behavioural characteristic which is universal, unique and measurable. Biometric systems operate by acquiring biometric data from individuals, extracting feature sets from data and comparing this feature with the enrolment set in a database. The aIm of this research is to compare the performance of gait-based user recognition of children with adults. Existing analyses techniques in gait-based recognition using wearable sensors for adults are applied to gait analyses in children. This is the first known study to be conducted on children (5-16 years old) for biometric gait recognition. Results presented here show that the performance degradation for children's walking compared to adult walking is approximately 100%. In comparable settings, a 6.21 % Equal Error Rate (EER) for adult gait recognition was reached, whilst for children's walking an EER of 12.69% was achieved. The performance of children's walking whilst carrying an object has also been studied. Results show that carrying an object actually improves the performance when walking normally, but when the children were asked to walk faster the walking becomes unstable, resulting in a higher Equal Error Rate (EER). A comparative investigation of the effects of time on gait recognition in children's walking pattern was carried out. The effects of age and gender have also been considered. In addition, children were tested six months apart; with the sensor on the hip position the performance of gait recognition shows significant variations with EER values. Abstract Finally, this thesis offers for the first time a coupled approach of statistical timedomain and frequency domain methods have been employed in order to match biometric gait signals. It has been shown that initially using root mean squared, crest-factor and kurtosis obtained similar matches in gait signals of children for the ages of 5-16 than for the traditional methods. Hence these novel methods employed can be exploited to verify these more established methods resident in gait recognition software.
Supervisor: Not available Sponsor: Not available
Qualification Name: Thesis (Ph.D.) Qualification Level: Doctoral
EThOS ID:  DOI: Not available