IET Biometrics最新文献

Due to the abuse of deep forgery technology, the research on forgery detection methods has become increasingly urgent. The corresponding relationship between the frequency spectrum information and the spatial clues, which is often neglected by current methods, could be conducive to a more accurate and generalized forgery detection. Motivated by this inspiration, we propose a wavelet-based texture mining and enhancement framework for face forgery detection. First, we introduce a frequency-guided texture enhancement (FGTE) module that mining the high-frequency information to improve the network’s extraction of effective texture features. Next, we propose a global–local feature refinement (GLFR) module to enhance the model’s leverage of both global semantic features and local texture features. Moreover, the interactive fusion module (IFM) is designed to fully incorporate the enhanced texture clues with spatial features. The proposed method has been extensively evaluated on five public datasets, such as FaceForensics++ (FF++), deepfake (DF) detection (DFD) challenge (DFDC), Celeb-DFv2, DFDC preview (DFDC-P), and DFD, for face forgery detection, yielding promising performance within and cross dataset experiments.

Emotions play a significant role in how we perceive and interact with products. Thoughtfully designed emotionally appealing products can evoke strong user responses, making them more attractive. Color, as a crucial attribute of products, is a significant aspect to consider in the process of emotional product design. However, users’ emotional perception of product colors is highly intricate and challenging to define. To address this, this research proposes a product color design concept that considers human emotion perception based on deep learning and cluster analysis. First, for a given product, a color style is chosen for rerendering, which is an emotional color image. Different emotional color images have distinct RGB color representations. Second, clustering methods are employed to establish relationships between various emotional color images and different colors, selecting emotionally close style images. Subsequently, through transfer learning techniques, specific grid structures are used to retrain network weights, allowing for the fusion design of style and content images. This process ultimately achieves emotional color rendering design based on emotional color clustering and transfer learning. Multiple sets of emotional color design examples demonstrate that the method proposed in this study can accurately fulfill the emotional color design requirements of products, thereby, offering practical applicability. The satisfaction survey shows that the proposed method has certain guiding significance for clothing emotional color design.

Deepfake (DF) involves utilizing artificial intelligence (AI) technology to synthesize or manipulate images, voices, and other human or object data. However, recent times have seen a surge in instances of DF technology misuse, raising concerns about cybercrime and the credibility of manipulated information. The objective of this study is to devise a method that employs remote photoplethysmography (rPPG) biosignals for DF detection. The face was divided into five regions based on landmarks, with automatic extraction performed on the neck region. We conducted rPPG signal extraction from each facial area and the neck region was defined as the ground truth. The five signals extracted from the face were used as inputs to an support vector machine (SVM) model by calculating the euclidean distance between each signal and the signal extracted from the neck region, measuring rPPG signal similarity with five features. Our approach demonstrated robust performance with an area under the curve (AUC) score of 91.2% on the audio-driven dataset and 99.7% on the face swapping generative adversarial network (FSGAN) dataset, even though we only used datasets excluding DF techniques that can be visually identified in Korean DF Detection Dataset (KoDF). Therefore, our research findings demonstrate that similarity features of rPPG signals can be utilized as key features for detecting DFs.

Recently, multimodal authentication methods based on deep learning have been widely explored in biometrics. Nevertheless, the contradiction between the data privacy protection and the requirement of sufficient data when model optimizing has become increasingly prominent. To this end, we proposes a multimodal biometric federated learning framework (FedMB) to realize the multiparty joint training of identity authentication models with different modal data while protecting the users’ data privacy. Specifically, a personalized multimodal biometric recognition model fully trained by each participant is first obtained to improve the authentication performance, using modal point clustering with class-first federated learning methods on the service side with the modal. Then a complementary multimodal biometric recognition strategy is implemented to build a complementary modal model. Finally, the fusion participant local model, with the modal model and complementary modal model, is trained by all participants again to obtain a more personalized modal model. The experimental results have demonstrated that the proposed FedMB can either protect the data privacy or utilize the data from all participants to train the personalized biometric recognition model to improve identity authentication performance.

Contactless palmprint recognition offers friendly customer experience due to its ability to operate without touching the recognition device under rigid constrained conditions. Recent palmprint recognition methods have shown promising accuracy; however, there still exist some issues that need to be further studied such as the limited discrimination of the single feature and how to effectively fuse deep features and shallow features. In this paper, deep features and shallow features are integrated into a unified framework using feature-level and score-level fusion methods. Specifically, deep feature is extracted by residual neural network (ResNet), and shallow features are extracted by principal component analysis (PCA), linear discriminant analysis (LDA), and competitive coding (CompCode). In feature-level fusion stage, ResNet feature and PCA feature are dimensionally reduced and fused by canonical correlation analysis technique to achieve the fused feature for the next stage. In score-level fusion stage, score information is embedded in the fused feature, LDA feature, and CompCode feature to obtain a more reliable and robust recognition performance. The proposed method achieves competitive performance on Tongji dataset and demonstrates more satisfying generalization capabilities on IITD and CASIA datasets. Comprehensive validation across three palmprint datasets confirms the effectiveness of our proposed deep and shallow feature fusion approach.

Human behavior recognition is the process of automatically identifying and analyzing multiple human behaviors using modern technology. From previous studies, we find that redundant features not only slow down the model training process and increase the structural complexity but also degrade the overall performance of the model. To overcome this problem, this paper investigates a temporal convolutional neural network (TCN) model based on improved sparrow search algorithm random forest (SSARF) feature selection to accurately identify human behavioral traits based on wearable devices. The model is based on the TCN classification model and incorporates a random forest with the sparrow optimization algorithm to perform dimensionality reduction on the original features, which is used to remove poorly correlated and unimportant features and retain effective features with a certain contribution rate to generate the optimal feature subset. In order to verify the reliability of the method, the performance of the model was evaluated on two public datasets, UCI Human Activity Recognition and WISDM, respectively, and 98.54% and 97.83% recognition accuracies were obtained, which were improved by 0.47% and 1.04% compared to the prefeature selection, but the number of features was reduced by 84.31% and 32.50% compared to the original feature set. In addition, we compared the TCN classification model with other deep learning models in terms of evaluation metrics such as F₁ score, recall, precision, and accuracy, and the results showed that the TCN model outperformed the other control models in all four metrics. Meanwhile, it also outperforms the existing recognition methods in terms of accuracy and other aspects, which have some practical application value.

Structural features are capable of effectively capturing the overall texture variations in images. However, in locally prominent areas with visible veins, other characteristics such as directionality, convexity–concavity, and curvature also play a crucial role in recognition, and their impact cannot be overlooked. This paper introduces a novel approach, the histogram of variable curvature directional binary statistical (HVCDBS), which combines the structural and directional features of images. The proposed method is designed for extracting discriminative multifeature information in vein recognition. First, a multidirection and multicurvature Gabor filter is introduced for convolution with vein images, yielding directional and convexity–concavity information at each pixel, along with curvature information for the corresponding curve. Simultaneously incorporating the original image feature information, these four aspects of information are fused and encoded to construct a variable curvature binary pattern (VCBP) with multifeatures. Second, the feature map containing multifeature information is blockwise processed to build variable curvature binary statistical features. Finally, competitive Gabor directional binary statistical features are combined, and a matching score-level fusion scheme is employed based on maximizing the interclass distance and minimizing the intraclass distance to determine the optimal weights. This process fuses the two feature maps into a one-dimensional feature vector, achieving an effective representation of vein images. Extensive experiments were conducted on four widely utilized vein databases, and the results indicate that the proposed algorithm, compared with solely extraction of structural features, achieved higher recognition rates and lower equal error rates.

Face recognition from side-view positions poses a considerable challenge in automatic face recognition tasks. Pose variation up to the side-view is an issue of difference in appearance and visibility since only one eye is visible at the side-view poses. Traditionally overlooked, recent advancements in deep learning have brought side-view poses to the forefront of research attention. This survey comprehensively investigates methods addressing pose variations up to side-view and categorizes research efforts into feature-based, image-based, and set-based pose handling. Unlike existing surveys addressing pose variations, our emphasis is specifically on extreme poses. We report numerous promising innovations in each category and contemplate the utilization and challenges associated with side-view. Furthermore, we introduce current datasets and benchmarks, conduct performance evaluations across diverse methods, and explore their unique constraints. Notably, while feature-based methods currently stand as the state-of-the-art, our observations suggest that cross-dataset evaluations, attempted by only a few researchers, produce worse results. Consequently, the challenge of matching arbitrary poses in uncontrolled settings persists.

In recent years, fingerprint authentication has gained widespread adoption in diverse identification systems, including smartphones, wearable devices, and attendance machines, etc. Nonetheless, these systems are vulnerable to spoofing attacks from suspicious fingerprints, posing significant risks to privacy. Consequently, a fingerprint presentation attack detection (PAD) strategy is proposed to ensure the security of these systems. Most of the previous work concentrated on how to build a deep learning framework to improve the PAD performance by augmenting fingerprint samples, and little attention has been paid to the fundamental difference between live and fake fingerprints to optimize feature extractors. This paper proposes a new fingerprint liveness detection method based on Siamese attention residual convolutional neural network (Res-CNN) that offers an interpretative perspective to this challenge. To leverage the variance in ridge continuity features (RCFs) between live and fake fingerprints, a Gabor filter is utilized to enhance the texture details of the fingerprint ridges, followed by the construction of an attention Res-CNN model to extract RCF between the live and fake fingerprints. The model mitigates the performance deterioration caused by gradient disappearance. Furthermore, to highlight the difference in RCF, a Siamese attention residual network is devised, and the ridge continuity amplification loss function is designed to optimize the training process. Ultimately, the RCF parameters are transferred to the model, and transfer learning is utilized to aid its acquisition, thereby assuring the model’s interpretability. The experimental outcomes conducted on three publicly accessible fingerprint datasets demonstrate the superiority of the proposed method, exhibiting remarkable performance in both true detection rate and average classification error rate. Moreover, our method exhibits remarkable capabilities in PAD tasks, including cross-material experiments and cross-sensor experiments. Additionally, we leverage Gradient-weighted Class Activation Mapping to generate a heatmap that visualizes the interpretability of our model, offering a compelling visual validation.

We propose original semantic labels for detailed face parsing to improve the accuracy of face recognition by focusing on parts in a face. The part labels used in conventional face parsing are defined based on biological features, and thus, one label is given to a large region, such as skin. Our semantic labels are defined by separating parts with large areas based on the structure of the face and considering the left and right sides for all parts to consider head pose changes, occlusion, and other factors. By utilizing the capability of assigning detailed part labels to face images, we propose a novel data augmentation method based on detailed face parsing called Face Semantic Erasing (FSErasing) to improve the performance of face recognition. FSErasing is to randomly mask a part of the face image based on the detailed part labels, and therefore, we can apply erasing-type data augmentation to the face image that considers the characteristics of the face. Through experiments using public face image datasets, we demonstrate that FSErasing is effective for improving the performance of face recognition and face attribute estimation. In face recognition, adding FSErasing in training ResNet-34 with Softmax using CelebA improves the average accuracy by 0.354 points and the average equal error rate (EER) by 0.312 points, and with ArcFace, the average accuracy and EER improve by 0.752 and 0.802 points, respectively. ResNet-50 with Softmax using CASIA-WebFace improves the average accuracy by 0.442 points and the average EER by 0.452 points, and with ArcFace, the average accuracy and EER improve by 0.228 points and 0.500 points, respectively. In face attribute estimation, adding FSErasing as a data augmentation method in training with CelebA improves the estimation accuracy by 0.54 points. We also apply our detailed face parsing model to visualize face recognition models and demonstrate its higher explainability than general visualization methods.