Performance evaluation of multimodal biometric systems using fusion techniques


Ho, Chiung Ching (2013) Performance evaluation of multimodal biometric systems using fusion techniques. PhD thesis, Multimedia University.

Full text not available from this repository.


This thesis deals with the performance evaluation of fused multimodal biometric systems comprising accelerometer-based gait, speaker and face modalities. Performance measures such as accuracy rate, true positive rate (TPR), false positive rate (FPR) and area under ROC curve (AUC) are employed for both individual biometric modalities as well as for multimodal biometric combinations. The same performance measures are used to evaluate accelerometer-based gait verification performed under the effects of different factors such as footwear, walking speed, encumbrance, and traditional articles of clothing. Multimodal biometric systems are being increasingly constructed using commercial off the shelf (COTS) equipment and sensors. In this thesis, COTS equipments are used due to reasons such as ease of acquisition, low cost and ubiquitous usage. Though combining multiple biometric modalities is not a new idea, the combination of accelerometer based gait, speech and face biometric modalities is a novel combination which has not been investigated by any researchers so far. Currently there is no benchmarked multimodal database available for the combination of gait, speech and face biometric modalities. In this work, a new multimodal biometric database named MMU GASPFA (Gait- Speaker- Face) comprising accelerometer based gait, speech and face biometric data is created. Collection of the accelerometer based gait is done with subjects using different encumbrance, footwear and clothing in order to see the effects of such factors on the verification process. Evaluation and comparison of the performances of multimodal biometric fusion strategies on different combinations of chimeric accelerometer gait, speech and face modalities using the BioSecure feature database, VidTIMIT face and speech database as well as the HumanSense accelerometer gait database are carried out. In addition, the performance comparison is done using the MMU GASPFA also. An assessment of the effects of different encumbrance, articles of clothing and footwear on accelerometer based gait authentication is also performed. Further, this thesis presents a conceptual framework for consolidating Malaysian fingerprint databases using a verification approach.

Item Type: Thesis (PhD)
Additional Information: Call No.: TK7882.B56 H63 2013
Subjects: T Technology > TK Electrical engineering. Electronics Nuclear engineering
Divisions: Faculty of Computing and Informatics (FCI)
Depositing User: Ms Nurul Iqtiani Ahmad
Date Deposited: 21 Apr 2014 03:03
Last Modified: 21 Apr 2014 03:03


Downloads per month over past year

View ItemEdit (login required)