Hayfaa Talib Hussein
Improving the accuracy of facial expression recognition systems.
Email: h.t.hussein@ncl.ac.uk
Project title
Signal processing and machine learning techniques for automatic image-based facial expression recognition
Supervisors
Project description
Our research will result in a robust image-based facial expression recognition (FER) system.
We are taking a close look at the key issues facing such a system with high performance in realistic situations. We are introducing novel techniques to surmount various challenging limitations. These techniques will provide higher levels of accuracy.
Methodology and objectives
We will develop a robust FER system to mitigate spontaneous database conditions. These are more challenging and essential for real-time. The system will use identity-independent, image-based expression recognition.
The system will reduce the effects of misalignment associated with the training images. It will extract spontaneous FE features.
We will enhance the performance of the image-based system using a novel approach. We will find an effective solution based on the fusion of three different types of features with deep networks. Deep networks are one of the most attractive methods for classification in the field of neural networks.
We will improve the accuracy of recognising facial expression. We will reduce the confusion between the expression classes using a multi-classifier system strategy. This splits the problem into two or more stages. Each stage is a simple classification task, which allows higher overall accuracy of facial expression recognition to be obtained.
Result
We will explore the issue of image-based expression recognition by identity-independent images. We will compare performance by using different types of databases. We will use a posed database and three types of spontaneous databases. Each of these has different challenges.
We will use automatic extraction of geometric features from raw image data with pre-processing steps and a deep network for the classifier. This will improve image-based facial expression recognition. We will also compare the performance of different types of features: appearance and geometric features.
We will find an effective solution for image-based spontaneous facial expression recognition based on a fusion of three types of features and multi-stage classification.
Publications
- Hussein H, Chambers J, Naqvi M. Study of sparsity-based facial expression recognition on a spontaneous database. In: 11th IMA International Conference on Mathematics in Signal Processing. 2016.
- Hussein H, Naqvi SM, Chambers J. Study of image-based expression recognition techniques on three recent spontaneous databases. In: 22nd International Conference on Digital Signal Processing (DSP). 2017, London: IEEE.
- Hussein H, Angelini F, Naqvi M, Chambers JA. Deep-Learning Based Facial Expression Recognition System Evaluated on Three Spontaneous Databases. In: 2018 9th International Symposium on Signal, Image, Video and Communications (ISIVC). 2018, Rabat, Morocco: IEEE.
- Hussein H, Naqvi SM, Chambers J. Spontaneous facial expression recognition through fusion of geometric and appearance features and deep Network based classification. In: 10th International Multi-Conference on Complexity, Informatics and Cybernetics (IMCIC 2019). 2019, Orlando, Florida, USA: International Institute of Informatics and Systemics.
- Hussein H, Angelini F, Chambers J. Multi-Stage Classification and Feature Fusion for Image-Based Spontaneous Facial Expression Recognition. Submitted to APSIPA Transactions on Signal and Information Processing. 2019
Qualifications
- MSc in Computer Science from Babylon University, Iraq