Mary Ann F Harrison

A survey on ear biometrics

Recognizing people by their ear has recently received significant attention in the literature. Several reasons account for this trend: first, ear recognition does not suffer from some problems associated with other non-contact biometrics, such as face recognition; second, it is the most promising candidate for combination with the face in the context of multi-pose face recognition; and third, the ear can be used for human recognition in surveillance videos where the face may be occluded completely or in part. Further, the ear appears to degrade little with age. Even though current ear detection and recognition systems have reached a certain level of maturity, their success is limited to controlled indoor conditions. In addition to variation in illumination, other open research problems include hair occlusion, earprint forensics, ear symmetry, ear classification, and ear individuality. This article provides a detailed survey of research conducted in ear detection and recognition. It provides an up-to-date review of the existing literature revealing the current state-of-art for not only those who are working in this area but also for those who might exploit this new approach. Furthermore, it offers insights into some unsolved ear recognition problems as well as ear databases available for researchers.

Correlation dimension and integral do not predict epileptic seizures

Reports in the literature have indicated potential value of the correlation integral and dimension for prediction of epileptic seizures up to several minutes before electrographic onset. We apply these measures to over 2000 total hours of continuous electrocortiogram, taken from 20 patients with epilepsy, examine their sensitivity to quantifiable properties such as the signal amplitude and autocorrelation, and investigate the influence of embedding and filtering strategies on their performance. The results are compared against those obtained from surrogate time series. Our conclusion is that neither the correlation dimension nor the correlation integral has predictive power for seizures.

Time-series-based prediction of complex oscillator networks via compressive sensing

Complex dynamical networks consisting of a large number of interacting units are ubiquitous in nature and society. There are situations where the interactions in a network of interest are unknown and one wishes to reconstruct the full topology of the network through measured time series. We present a general method based on compressive sensing. In particular, by using power series expansions to arbitrary order, we demonstrate that the network-reconstruction problem can be casted into the form X=Ga, where the vector X and matrix G are determined by the time series and a is a sparse vector to be estimated that contains all nonzero power series coefficients in the mathematical functions of all existing couplings among the nodes. Since a is sparse, it can be solved by the standard L1-norm technique in compressive sensing. The main advantages of our approach include sparse data requirement and broad applicability to a variety of complex networked dynamical systems, and these are illustrated by concrete examples of model and real-world complex networks.

Observations on the application of the correlation dimension and correlation integral to the prediction of seizures.

Summary: The authors reexamine the correlation integral and the related correlation dimension in the context of EEG analysis with application to seizure prediction. They identify dependencies of the correlation integral and the correlation dimension on frequency and amplitude of the signal, which may result in a reinterpretation of the dynamic importance of these measures and may cast doubts on their predictive abilities for certain classes of seizures. The relevance, for clinical and research purposes, of the distinction between retrospective and prospective inference (prediction) is addressed briefly. The authors point to the need for further research, consisting of long time series, containing multiple seizures, and for the development of objective prediction criteria.

Accumulated energy revisited.

OBJECTIVE To examine the seizure prediction and detection abilities of the accumulated energy on multi-center data submitted to the First International Collaborative Workshop on Seizure Prediction. METHODS The accumulated energy (AE), windowed average power, and FHS seizure detection algorithm were applied to a single channel of ECoG data taken from the data sets contributed to the workshop. The FHS seizure detection algorithm was used to perform automated scoring of the data in order to locate subclinical events not picked up by the centers where the data was collected. The results were analyzed retrospectively, comparing the behavior of the accumulated energy and windowed average power on segments containing seizures to interictal segments. RESULTS Accumulated energy curves showed no divergence from interictal curves prior to seizure. Distinctive or clear increases in the AE slope occurred sometime at or after electrographic seizure onset for some seizures. Similarly, the windowed average power showed no consistent increases in broadband energy prior to seizures. However, both methods may have detection ability for some seizures. CONCLUSIONS The accumulated energy did not appear to have predictive abilities for these data sets. Some detection ability was apparent. SIGNIFICANCE In data unsorted by sleep/wake state, no seizure prediction was evident. The lack of prediction calls into question the existence of a preictal state as previously claimed in the literature using this method.

Design and evaluation of photometric image quality measures for effective face recognition

The performance of an automated face recognition system can be significantly influenced by face image quality. Designing effective image quality index is necessary in order to provide real-time feedback for reducing the number of poor quality face images acquired during enrollment and authentication, thereby improving matching performance. In this study, the authors first evaluate techniques that can measure image quality factors such as contrast, brightness, sharpness, focus and illumination in the context of face recognition. Second, they determine whether using a combination of techniques for measuring each quality factor is more beneficial, in terms of face recognition performance, than using a single independent technique. Third, they propose a new face image quality index (FQI) that combines multiple quality measures, and classifies a face image based on this index. In the authors studies, they evaluate the benefit of using FQI as an alternative index to independent measures. Finally, they conduct statistical significance Z-tests that demonstrate the advantages of the proposed FQI in face recognition applications.

Fast learning ear detection for real-time surveillance

Fully automated image segmentation is an essential step for designing automated identification systems. This paper investigates the problem of real-time image segmentation in the context of ear biometrics. The proposed approach is based on Haar features arranged in a cascaded Adaboost classifier. This method, widely known as Viola-Jones in the context of face detection, has a limitation of an extremely long training time, approximately a month. We efficiently implement a modified training / learning method, which significantly reduces training time. This approach is trained about 80 times faster than the original method, and achieves ~ 95% accuracy based on four different test sets (> 2000 profile images for app. 450 persons). The developed ear detection system is very fast and can be used in a real-time surveillance scenario. Experimental results show that the proposed ear detection is robust in the presence of partial occlusion, noise and multiple ears with various resolutions.

BIFURCATION TO HIGH-DIMENSIONAL CHAOS

High-dimensional chaos has been an area of growing recent investigation. The questions of how dynamical systems become high-dimensionally chaotic with multiple positive Lyapunov exponents, and what the characteristic features associated with the transition are, remain less investigated. In this paper, we present one possible route to high-dimensional chaos. By this route, a subsystem becomes chaotic with one positive Lyapunov exponent via one of the known routes to low-dimensional chaos, after which the complementary subsystem becomes chaotic, leading to additional positive Lyapunov exponents for the whole system. A characteristic feature of this route is that the additional Lyapunov exponents pass through zero smoothly. As a consequence, the fractal dimension of the chaotic attractor changes continuously through the transition, in contrast to the transition to low-dimensional chaos at which the fractal dimension changes abruptly. We present a heuristic theory and numerical examples to illustrate this route to high-dimensional chaos.

PROBING COMPLEX NETWORKS FROM MEASURED TIME SERIES

We propose a method to detect nodes of relative importance, e.g. hubs, in an unknown network based on a set of measured time series. The idea is to construct a matrix characterizing the synchronization probabilities between various pairs of time series and examine the components of the principal eigenvector. We provide a heuristic argument indicating the existence of an approximate one-to-one correspondence between the components and the degrees of the nodes from which measurements are obtained. The striking finding is that such a correspondence appears to be quite robust, which holds regardless of the detailed node dynamics and of the network topology. Our computationally efficient method thus provides a general means to address the important problem of network detection, with potential applications in a number of fields.

Detection of seizure rhythmicity by recurrences.

Epileptic seizures show a certain degree of rhythmicity, a feature of heuristic and practical interest. In this paper, we introduce a simple model of this type of behavior, and suggest a measure for detecting and quantifying it. To evaluate our method, we develop a set of test segments that incorporate rhythmicity features, and present results from the application of this measure to test segments. We then analyze electrocorticogram segments containing seizures, and present two examples. Finally, we discuss the similarity of our method to techniques for detecting unstable periodic orbits in chaotic time series.