Alessandro Zimmer

A hybrid on/off line handwritten signature verification system

This paper proposes a new hybrid handwritten signature verification system where the on-line reference data acquired through a digitizing tablet serves as the basis for the segmentation process of the corresponding scanned off-line data. Local foci of attention over the image are determined through a self-adjustable learning process in order to pinpoint the feature extraction process. Both local and global primitives are processed and the decision about the authenticity of the specimen is defined through similarity measurements. The global performance of the system is measured using two different classifiers.

Offline signature verification system based on the online data

Most of the signature verification work done in the past years focused either on offline or online approaches. In this paper, a different methodology is proposed, where the online reference data acquired through a digitizing tablet serves as the basis for the segmentation process of the corresponding scanned offline data. Local windows are built over the image through a self-adjustable learning process and are used to focus on the feature extraction step. The windows positions are determined according to the complexity of the underlying strokes based on the observation of a delta-lognormal handwritten reproduction model. Local features extraction that takes place focused on the windows formed, and it is used in conjunction with the global primitives to feed the classifier. The overall performance of the system is then measured with three different classification schemes.

Automatic pupil size measurement based on region growth

Dynamic pupillometry, a method that consists of measuring changes in pupil size in response to a light stimulus, can provide information about the autonomous nervous system. Therefore, it is used in the assessment of some diseases, such as diabetes. In biometric systems, it has also been used as aliveness test and, more recently to extract features to identify subjects by dynamic characteristics. This paper presents some automatic methods that can be used for this purpose and proposes a new approach, based on region grow, comparing the results of each one.

A novel approach for intelligent pre-crash threat assessment systems

Compared to the state-of-the-art on integrated safety systems, earlier activated safety systems can further reduce the risk of suffering a major injury. Activation of such systems prior to a collision can be realized by analysing measurements of exteroceptive sensors (pre-crash data). An algorithm for estimating collisions in real-time using fused measurements of a video camera, a laser range finder (LRF), and ego vehicle motion sensors is presented. The threat posed by the actual driving situation is assessed by calculating a certain risk value, which is determined by combining the collision probability and crash severity estimations in a comprehensive way. A scale model vehicle is introduced to capture characteristics of the proposed system experimentally. First test runs show that the object width measurement is very accurate (absolute error of 5%) and the maximum time to collision (TTC) estimation error is around 17% about 300ms before the impact. Comparing different obstacles and impact scenarios (e.g. small overlap vs. full frontal collision), the calculated risk is a promising new measure to early discriminate crash types.

A Window-Based Hybrid Signature Verification System

In this work we present a hybrid handwritten signature verification system where the on-line reference data acquired through a digitizing tablet serves as the basis for the segmentation process of the corresponding scanned off-line data. Local windows are determined over the image through a self-adjustable learning process and are used to focus the feature extraction step. The positions of the windows are determined according to the complexity of the underlying strokes given by the observation of a handwritten reproduction model. Local feature extraction is bounded by the windows formed and it is used with global primitives to feed the classifier. The overall performance of the system is then measured.

Advantages in Crash Severity Prediction Using Vehicle to Vehicle Communication

The paper discusses a new approach in contactless crash detection combining measurements of vehicle dynamics, exteroceptive sensors and vehicle-to-vehicle (V2V) communication data. The proposed architecture aims to activate vehicle safety functions prior an imminent collision to minimize the risk of suffering a major injury. An activation needs a precise prediction of time to collision (TTC), the crash severity (Cs) and other relevant crash parameters. This paper studies the contribution of V2V communication data to predict potential collisions and to realize a reliable activation. An algorithm is presented, that merges fused measurements of a video camera, a laser range finder (LRF) and ego vehicle motion sensors with V2V communication data to predict collisions. The benefit using V2V communication is demonstrated by evaluating collision prediction errors. This analysis is carried out based on experimental data produced by two scale model vehicles.

Biometric Authentication Based on Pupillary Light Reflex Using Neural Networks

Many different biometric traits can be used for human identification, and each one has specific advantages over the others. However, there is a common problem for most of them: the spoofing vulnerability of the features. The authentication based on human reflexes is a recent proposal against frauds at the biometric level. The involuntary responses of the body cannot be removed from the person, naturally mimicked or reproduced by artificial means. This paper presents a proposal for the use of the features of the Pupillary Light Reflex for user authentication, using artificial neural networks for classification.

Using the pupillary reflex as a diabetes occurrence screening aid tool through neural networks

Diabetes mellitus is a disease that may cause dysfunctions in the sympathetic and parasympathetic nervous system. Therefore, the pupillary reflex of diabetic patients shows characteristics that distinguish them from healthy people, such as pupil radius and contraction time. These features can be measured by the noninvasive way of dynamic pupillometry, and an analysis of the data can be used to check the existence of a neuropathy. In this paper, it is proposed the use of artificial neural networks for helping screening the diabetes occurrence through the dynamic characteristics of the pupil, with successful results.

A Model-Based Signature Verification System

In this work a hybrid handwritten signature verification system architecture is built where the segmentation of the signature is done inspired by a vectorial delta-lognormal kinematic handwritten reproduction model. The acquisition module processes dynamic data sampled from a digitizer tablet creating a prototype signature that is saved in a database with the correspondent features thresholds. The verification is done over the image of the test signature without the presence of the author. Details about the segmentation methodology are presented along with the systems architecture and performance results.

Using Dynamic Pupillometry Into A Distributed Healthcare System

Reducing the complexity of diagnostic systems and methods allows a larger number of medical centers, mainly those in small cities and in isolated regions, due to the easy of use and the smaller complexity for adoption. On the other hand, centralizing the database and diagnostic evaluation of the results allows for traditional and newly diagnostic methods to be applied analyzing a larger set of patients and clinical conditions that can lead to new clinical findings. While dynamic pupillometry has been proposed as a simpler and more sensitive tool to detect subclinical autonomic dysfunction, this paper proposes using it integrate to a distributed medical healthcare system that integrates several small modules via internet, while allowing enabling medical centers to have access to data acquired in that distributed manner with the potential to apply advanced diagnostic methods.