Samuel Chindaro

Automatic classification of hand drawn geometric shapes using constructional sequence analysis

A method for automatically assessing theconstructional sequence from a neuropsychologicaldrawing task using Hidden Markov Models is presented.We also present a method of extracting and identifyingthe position of individual pen strokes relating toindividual sides of a shape within a drawing to formtraining and testing sequences. Our results from twoexperiments using data from patients with visuo-spatialneglect show the HMM classifier is able to generalise onincorrectly extracted sequences and obtain a diagnosticclassification which can be used alongside other forms ofconventional assessment.

Colour space fusion for texture recognition

In this paper we propose a novel approach to colour texture classification based on fusion of the information contained in different colour spaces. In colour texture classification the choice of the most effective colour space to use is still an open issue. However, combining the strengths of different colour spaces may offer an alternative solution to the problem of robust texture discrimination. The principal aim of the work presented here is to study the performance of such decision combination approaches using classifiers obtained through training on features extracted from a number of colour space and subspace representations of the same texture classes. To this end we performed a number of cross-validation experiments involving six different colour spaces and their chromatic subspaces. Our results strongly suggest that colour texture classification can benefit significantly from techniques based on multiple classifier combination strategies.

ASSESSING VISUO-SPATIAL NEGLECT THROUGH FEATURE SELECTION FROM SHAPE DRAWING PERFORMANCE AND SEQUENCE ANALYSIS

The reported work aims to objectively and accurately assess the post-stroke clinical condition of visuo-spatial neglect using a series of standardized geometric shape drawing tasks. We present a method implementing existing pencil-and-paper diagnostic methods and define a set of static and dynamic features that can be extracted from drawing responses captured online using a graphics tablet. We also present a method for automatically assessing the constructional sequence of the drawing using Hidden Markov Models. The method enables the automated extraction, position identification and drawing order of individual sides of a shape within a drawing. Discrimination between two populations (a neglect population and stroke subjects without neglect as determined by existing standard assessment methods) using a combination of performance features and constructional sequence is examined across three separate drawing tasks. Results from experimentation show how a combination of sequence and performance features is able to generalize across a wide variety of input samples and obtain a diagnostic classification which can be used alongside other forms of conventional assessment. Furthermore, the application of a multi-classifier combination strategy leads to a significant increase in recognition ability.

Directional Properties of Colour Co-occurrence Features for Lip Location and Segmentation

Automatic lip tracking is based on robust lip location and segmentation. Here an algorithm which can locate the position of the lips robustly without the constraints of lip highlighting or special lighting conditions is proposed. The proposed method is based on the directional properties of the features of co-occurrence matrices to distinguish between facial parts. The method essentially consists of three parts: a) a face location module b) a facial features location module c) a feature identification module which identifies the lips. The proposed algorithm uses the hue information only in the HSV colour space. The method has been tested on the XM2VTS database, with a high success rate. The use of hue and textural features to do the processing makes the algorithm robust under various lighting conditions.

A Classification Framework for Large-Scale Face Recognition Systems

This paper presents a generic classification framework for large-scale face recognition systems. Within the framework, a data sampling strategy is proposed to tackle the data imbalance when image pairs are sampled from thousands of face images for preparing a training dataset. A modified kernel Fisher discriminant classifier is proposed to make it computationally feasible to train the kernel-based classification method using tens of thousands of training samples. The framework is tested in an open-set face recognition scenario and the performance of the proposed classifier is compared with alternative techniques. The experimental results show that the classification framework can effectively manage large amounts of training data, without regard to feature types, to efficiently train classifiers with high recognition accuracy compared to alternative techniques.

Analysis and modelling of diversity contribution to ensemble-based texture recognition performance

The RGB colour space is prominent as a colour representation and display scheme, although a number of other colour spaces have been developed over the years each with its own advantages and shortcomings with regard to its usefulness for colour/texture recognition. However, the recent advent of multiple classifier systems provides the unique opportunity to exploit the diverse information encapsulated in the different colour representations in a systematic fashion. In this paper we propose the use of classifier combination schemes which utilise information from different colour domains. We subsequently use suitable measures to investigate the diversity of the information infused by the different colour spaces. Experiments with two 40-class colour/texture datasets show the benefit of our multiple classifier approach, and reveal the existence of strong correlations between the accuracy achieved and the diversity measures. Finally, we illustrate, using quadratic regression, that there is significant scope to build and explore further (potentially causal) models of the observed relations between ensemble performance and diversity metrics. Our results point towards the use of diversity along with other statistical measures as possible predictors of the ensemble behaviour.

Modelling multiple-classifier relationships using Bayesian belief networks

Because of the lack of a clear guideline or technique for selecting classifiers which maximise diversity and accuracy, the development of techniques for analysing classifier relationships and methods for generating good constituent classifiers remains an important research direction. In this paper we propose a framework based on the Bayesian Belief Networks (BBN) approach to classification. In the proposed approach the multiple-classifier system is conceived at a meta-level and the relationships between individual classifiers are abstracted using Bayesian structural learning methods. We show that relationships revealed by the BBN structures are supported by standard correlation and diversity measures. We use the dependency properties obtained by the learned Bayesian structure to illustrate that BBNs can be used to explore classifier relationships, and for classifier selection.

A multibiometric face recognition fusion framework with template protection

In this work we present a multibiometric face recognition framework based on combining information from 2D with 3D facial features. The 3D biometrics channel is protected by a privacy enhancing technology, which uses error correcting codes and cryptographic primitives to safeguard the privacy of the users of the biometric system at the same time enabling accurate matching through fusion with 2D. Experiments are conducted to compare the matching performance of such multibiometric systems with the individual biometric channels working alone and with unprotected multibiometric systems. The results show that the proposed hybrid system incorporating template protection, match and in some cases exceed the performance of corresponding unprotected equivalents, in addition to offering the additional privacy protection.

Face Recognition Using Balanced Pairwise Classifier Training

This paper presents a novel pairwise classification framework for face recognition (FR). In the framework, a two-class (intra- and inter-personal) classification problem is considered and features are extracted using pairs of images. This approach makes it possible to incorporate prior knowledge through the selection of training image pairs and facilitates the application of the framework to tackle application areas such as facial aging. The non-linear empirical kernel map is used to reduce the dimensionality and the imbalance in the training sample set tackled by a novel training strategy. Experiments have been conducted using the FERET face database.format.

Non-linear fusion of local matching scores for face verification

This paper presents a face verification framework for fusing matching scores that measure similarities of local facial features. The framework is aimed to handle an open set verification scenario when users who try to enroll can be unknown to the system at the training phase. The kernel discriminant analysis is adopted within the framework to explore the discriminatory information of local matching scores in a high-dimensional non-linear space. A large sample size problem is raised for system training and an effective strategy is provided for tackling this problem. We demonstrate the framework by fusing the scores calculated using local binary pattern features. The experimental results show that our method improves the verification performance significantly when compared to a number of competitive techniques.