Cardiff University | Prifysgol Caerdydd ORCA
Online Research @ Cardiff 
WelshClear Cookie - decide language by browser settings

Modelling facial dynamics change as people age

Al-Meyah, Khtam 2017. Modelling facial dynamics change as people age. PhD Thesis, Cardiff University.
Item availability restricted.

[thumbnail of 2018al-meyahkphd.pdf]
PDF - Accepted Post-Print Version
Download (49MB) | Preview
[thumbnail of al-meyahk.pdf] PDF - Supplemental Material
Restricted to Repository staff only

Download (2MB)


In the recent years, increased research activity in the area of facial ageing modelling has been recorded. This interest is attributed to the potential of using facial ageing modelling techniques for a number of different applications, including age estimation, prediction of the current appearance of missing persons, age-specific human-computer interaction, computer graphics, forensic applications, and medical applications. This thesis describes a general AAM model for modelling 4D (3D dynamic) ageing and specific models to map facial dynamics as people age. A fully automatic and robust pre-processing pipeline is used, along with an approach for tracking and inter-subject registering of 3D sequences (4D data). A 4D database of 3D videos of individuals has been assembled to achieve this goal. The database is the first of its kind in the world. Various techniques were deployed to build this database to overcome problems due to noise and missing data. A two-factor (age groups and gender) analysis of variance (MANOVA) was performed on the dataset. The groups were then compared to assess the separate effects of age on gender through variance analysis. The results show that smiles alter with age and have different characteristics between males and females. We analysed the rich sources of information present in the 3D dynamic features of smiles to provide more insight into the patterns of smile dynamics. The sources of temporal information that have been investigated include the varying dynamics of lip movements, which are analysed to extract the descriptive features. We evaluated the dynamic features of closed-mouth smiles among 80 subjects of both genders. Multilevel Principal Components Analysis (mPCA) is used to analyse the effect of naturally occurring groups in a population of individuals for smile dynamics data. A concise overview of the formal aspects of mPCA has been outlined, and we have demonstrated that mPCA offers a way to model the variations at different levels of structure in the data (between and within group levels).

Item Type: Thesis (PhD)
Date Type: Completion
Status: Unpublished
Schools: Computer Science & Informatics
Date of First Compliant Deposit: 18 February 2019
Last Modified: 26 Oct 2021 01:26

Actions (repository staff only)

Edit Item Edit Item


Downloads per month over past year

View more statistics