IET Biometrics最新文献

Human Activity Recognition (HAR) is the process of identifying and analysing activities performed by a person (or persons). This paper proposes an efficient HAR system based on wearable sensors that uses deep learning techniques. The proposed HAR takes the advantage of staking Convolutional Neural Network and Long Short-Term (LSTM), for extracting the high-level features of the sensors data and for learning the time-series behaviour of the abstracted data, respectively. This paper proposed a Fuzzy Soft-max classifier for the dense layer which classifies the output of LSTM Blocks to the associated activity classes. The authors’ decision for proposing this classifier was because sensor data related to the resembling human activities, such as walking and running or opening door and closing door, are often very similar to each other. For this reason, the authors expect that adding fuzzy inference power to the standard Soft-max classifier will increase its accuracy for distinguishing between similar activities. The authors were also interested in considering a post-processing module that considers activity classification over a longer period. Using the proposed Fuzzy Soft-max classifier and by the post-processing technique, the authors were able to reach the 97.03 and 85.1 rates of accuracy for the PAMAP2 and Opportunity dataset, respectively.

Due to the need for increased security measures for monitoring and safeguarding the activities, video anomaly detection is considered as one of the significant research aspects in the domain of computer vision. Assigning human personnel to continuously check the surveillance videos for finding suspicious activities such as violence, robbery, wrong U-turns, to mention a few, is a laborious and error-prone task. It gives rise to the need for devising automated video surveillance systems ensuring security. Motivated by the same, this paper addresses the problem of detection and localization of anomalies from surveillance videos using pipelined deep autoencoders and one-class learning. Specifically, we used a convolutional autoencoder and a sequence-to-sequence long short-term memory autoencoder in a pipelined fashion for spatial and temporal learning of the videos, respectively. The authors followed the principle of one-class classification for training the model on normal data and testing it on anomalous testing data. The authors achieved a reasonably significant performance in terms of an equal error rate and the time required for anomaly detection and localization comparable to standard benchmarked approaches, thus, qualifies to work in a near-real-time manner for anomaly detection and localization.

The growing scope, scale, and number of biometric deployments around the world emphasise the need for research into technologies facilitating efficient and reliable biometric identification queries. This work presents a method of indexing biometric databases, which relies on signal-level fusion of facial images (morphing) to create a multi-stage data structure and retrieval protocol. By successively pre-filtering the list of potential candidate identities, the proposed method makes it possible to reduce the necessary number of biometric template comparisons to complete a biometric identification transaction. The proposed method is extensively evaluated on publicly available databases using open-source and commercial off-the-shelf recognition systems. The results show that using the proposed method, the computational workload can be reduced down to around 30% while the biometric performance of a baseline exhaustive search-based retrieval is fully maintained, both in closed-set and open-set identification scenarios.

Diverse applications are used on mobile devices. Because of the increasing dependence on information systems, immense amounts of personal and sensitive data are stored on mobile devices. Thus, security or privacy breaches are a major challenge. To protect mobile systems and the private information on these systems from being accessed by adversaries, a framework for mobile user identification through the use of a multimodal behavioural biometrics scheme with a keystroke trajectory feature is presented herein. Conventionally, mobile devices have been protected by mechanisms such as PINs or passwords. However, these approaches have numerous disadvantages. Therefore, approaches that employ keystroke biometrics for secure and reliable mobile device identification have been proposed. Because unimodal behavioural biometrics identification mechanisms have limited accuracy and effectiveness, a multimodal scheme that includes different behavioural biometric traits, such as keystroke and swipe biometric traits, is examined. However, the information provided by the spatial and temporal features of keystroke biometrics is not comprehensive. Therefore, a trajectory model is derived to describe the behavioural biometric uniqueness of a user. In the user identification phase, a multistream recurrent neural network (RNN) is adopted. The results reveal that the proposed trajectory model performs well, and the multimodal scheme using an RNN with a late fusion method provides accurate identification results. The proposed system achieved an accuracy of 95.29%, F1 score of 94.64%, and equal error rate of 1.78%. Thus, the proposed mobile identification system is capable of resisting attacks that standard mechanisms may be vulnerable to and represents a valuable contribution to cyber security.

With the widespread of computing and mobile devices, authentication using biometrics has received greater attention. Although biometric systems usually provide efficient solutions, the recognition performance tends to be affected over time due to changing conditions and the ageing of biometric data, which results in intra-class variability. This issue is one of the leading causes of the high false rejection rate in biometric authentication systems. Fortunately, this issue has been addressed by using adaptive biometric solutions in which the system gradually adapts to new changes in user biometrics. However, their adaptability to changes may be exploited by an attacker to compromise the stored templates, either to impersonate a specific client or to deny access to him/her. In this work, the authors will carry out a systematic literature review by conducting a comparative study on state-of-the-art solutions for spoofing detection on adaptive authentication systems. This paper will identify the main issues that need to be addressed in adaptive authentication systems. Thus, the authors aim to encourage researchers to develop more robust adaptive solutions to overcome the identified gaps in this research.

The vast advances in biometrics over the past several decades have brought with them a host of pressing concerns. Philosophical scrutiny has already been devoted to many of the relevant ethical and political issues, especially ones arising from matters of privacy, bias, and security in data collection. But philosophers have devoted surprisingly little attention to the relevant metaphysical issues, in particular, ones concerning matters of personal identity. This paper aims to take some initial steps to correct this oversight. After discussing the philosophical problem of personal identity, the ways in which the notion of biometric identity connects with, or fails to connect with, the philosophical notion of personal identity is explored. Though there may be some good reasons to use biometric identity to track personal identity, it is contended that biometric identity is not the same thing as personal identity and thus that biometrics researchers should stop talking as if it were.

Head pose estimation (HPE) notoriously represents a crucial task for many computer vision applications in robotics, biometry and video surveillance. While, in general, HPE can be performed on both still images and frames extracted from live video or captured footage, its functional approach and the related processing pipeline may have a significant impact on suitability to different application contexts. This implies that, for any real-time application in which HPE is required, this information, namely the angular value of yaw, pitch and roll axes, should be provided in real-time as well. Since, so far, the primary aim in HPE research has been on improving estimation accuracy, there are only a few works reporting the computing time of the proposed HPE method and even less explicitly addressing it. The present work stems from a previous Partitioned Iterated Function Systems-based approach providing state-of-the-art accuracy with high computing cost, and improve it by means of two regression models, namely Gradient Boosting Regressor and Extreme Gradient Boosting Regressor, achieving much faster response and an even lower mean absolute error on the yaw and roll axis, as shown by experiments conducted on the BIWI and AFLW2000 datasets.

Extracting a palm region with fixed location from an input hand image is a crucial task for palmprint recognition to realise reliable person authentication under contactless and unconstrained conditions. A palm region can be extracted from the fixed location using the gaps between fingers. An accurate and robust hand segmentation method is indispensable to extract a palm region from an image with complex background taken under various environments. In this study, HandSegNet, which is a hand segmentation method using Convolutional Neural Network (CNN) for contactless palmprint recognition, is proposed. HandSegNet employs a new CNN architecture consisting of an encoder–decoder model with a pyramid pooling module. Through performance evaluation using a set of synthesised hand images, HandSegNet exhibited the best segmentation results of 98.90% and 93.20% for accuracy and intersection over union, respectively. The effectiveness of HandSegNet in contactless palmprint recognition through experiments using a set of synthesised images of hand images is also demonstrated. Comparing the performance of palmprint recognition using three conventional methods and HandSegNet for palm region extraction, the proposed method has the lowest equal error rate of 4.995%, demonstrating its effectiveness in palm region extraction for contactless palmprint recognition.