Alex I. Bazin

A floor sensor system for gait recognition

This paper describes the development of a prototype floor sensor as a gait recognition system. This could eventually find deployment as a standalone system (e.g. a burglar alarm system) or as part of a multimodal biometric system. The new sensor consists of 1536 individual sensors arranged in a 3 m by 0.5 m rectangular strip with an individual sensor area of 3 cm/sup 2/. The sensor floor operates at a sample rate of 22 Hz. The sensor itself uses a simple design inspired by computer keyboards and is made from low cost, off the shelf materials. Application of the sensor floor to a small database of 15 individuals was performed. Three features were extracted : stride length, stride cadence, and time on toe to time on heel ratio. Two of these measures have been used in video based gait recognition while the third is new to this analysis. These features proved sufficient to achieve an 80% recognition rate.

Aligning Text and Phonemes for Speech Technology Applications Using an EM-Like Algorithm

A common requirement in speech technology is to align two different symbolic representations of the same linguistic ‘message’. For instance, we often need to align letters of words listed in a dictionary with the corresponding phonemes specifying their pronunciation. As dictionaries become ever bigger, manual alignment becomes less and less tenable yet automatic alignment is a hard problem for a language like English. In this paper, we describe the use of a form of the expectation-maximization (EM) algorithm to learn alignments of English text and phonemes, starting from a variety of initializations. We use the British English Example Pronunciation (BEEP) dictionary of almost 200,000 words in this work. The quality of alignment is difficult to determine quantitatively since no ‘gold standard’ correct alignment exists. We evaluate the success of our algorithm indirectly from the performance of a pronunciation by analogy system using the aligned dictionary data as a knowledge base for inferring pronunciations. We find excellent performance—the best so far reported in the literature. There is very little dependence on the start point for alignment, indicating that the EM search space is strongly convex. Since the aligned BEEP dictionary is a potentially valuable resource, it is made freely available for research use.

Gait Verification Using Probabilistic Methods

In this paper we describe a novel method for gait based identity verification based on Bayesian classification. The verification task is reduced to a two class problem (Client or Impostor) with logistic functions constructed to provide probability estimates of intra-class (Client) and inter-class (Impostor) likelihoods. These likelihoods are combined using Bayes rule and thresholded to provide a decision boundary. Since the outputs of the classifier are probabilities they are particularly well suited for use without modification in classifier fusion schemes. On tests using 1664 examples from 100 clients and 100 impostors the Bayesian method achieved an equal error rate of 7.3%. The improvement over a Euclidean distance classifier was shown to be statistically significant at the 5% level using McNemars test.

Developing a non-intrusive biometric environment

The development of large scale biometric systems requires experiments to be performed on large amounts of data. Existing capture systems are designed for fixed experiments and are not easily scalable. In this scenario even the addition of extra data is difficult. We developed a prototype biometric tunnel for the capture of non-contact biometrics. It is self contained and autonomous. Such a configuration is ideal for building access or deployment in secure environments. The tunnel captures cropped images of the subjects face and performs a 3D reconstruction of the persons motion which is used to extract gait information. Interaction between the various parts of the system is performed via the use of an agent framework. The design of this system is a trade-off between parallel and serial processing due to various hardware bottlenecks. When tested on a small population the extracted features have been shown to be potent for recognition. We currently achieve a moderate throughput of approximate 15 subjects an hour and hope to improve this in the future as the prototype becomes more complete

Probabilistic combination of static and dynamic gait features for verification

This paper describes a novel probabilistic framework for biometric identification and data fusion. Based on intra and inter-class variation extracted from training data, posterior probabilities describing the similarity between two feature vectors may be directly calculated from the data using the logistic function and Bayes rule. Using a large publicly available database we show the two imbalanced gait modalities may be fused using this framework. All fusion methods tested provide an improvement over the best modality, with the weighted sum rule giving the best performance, hence showing that highly imbalanced classifiers may be fused in a probabilistic setting; improving not only the performance, but also generalized application capability.

A smart environment for biometric capture

Current biometric capture methodologies were born in a laboratory environment. In this scenario you have cooperative subjects, large time capture windows, and staff to edit and mark up data as necessary. However, as biometrics moves from the laboratory these factors impinge upon the scalability of the system. In this work we developed a prototype biometric tunnel for the capture of non-contact biometrics. The system is autonomous to maximise subject throughput and self-contained to allow flexible deployment and user friendliness. Currently we deploy 8 cameras to capture the 3D motion (specifically gait) and 1 camera to capture the face of a subject. The gait and face information thus extracted can be used for subsequent biometric analysis. Interaction between the various system components is performed via the use of an agent framework. Performance analysis of the current system shows that we can currently achieve a moderate throughput of 15 subjects per hour. Additionally, analysis performed upon the biometric features extracted from a small population show them to be potent for recognition

An automated system for contact lens inspection

This paper describes a novel method for the industrial inspection of ophthalmic contact lenses in a time constrained production line environment. We discuss the background to this problem, look at previous solutions and relevant allied work before describing our system. An overview of the system is given together with detailed descriptions of the algorithms used to perform the image processing, classification and inspection system. We conclude with a preliminary assessment of the system performance and discuss future work needed to complete the system.