Publications Académiques
Découvrez mes travaux de recherche, articles académiques et contributions scientifiques dans les domaines de l'Intelligence Artificielle et du d'éveloppement avancé.
Authors:
Time series data analysis is vital in numerous fields, driven by advancements in deep learning and machine learning. This paper presents a comprehensive overview of data augmentation techniques in time series analysis, with a specific focus on their applications within deep learning and machine learning. We commence with a systematic methodology for literature selection, curating 757 articles from prominent databases. Subsequent sections delve into various data augmentation techniques, encompassing traditional approaches like interpolation and advanced methods like Synthetic Data Generation, Generative Adversarial Networks (GANs), and Variational Autoencoders (VAEs). These techniques address complexities inherent in time series data. Moreover, we scrutinize limitations, including computational costs and overfitting risks. However, it's essential to note that our analysis does not end with limitations. We also comprehensively analyzed the advantages and applicability of the techniques under consideration. This holistic evaluation allows us to provide a balanced perspective. In summary, this overview illuminates data augmentation's role in time series analysis within deep and machine-learning contexts. It provides valuable insights for researchers and practitioners, advancing these fields and charting paths for future exploration.
International Journal of Advanced Computer Science and ApplicationsOpen AccessDOI: 10.14569/IJACSA.2024.01501118 data augmentationtime series
Authors:
Throughout this study, we employed unsupervised machine learning clustering algorithms, namely K-Means [1] and hierarchical agglomerative clustering (HAC) [2], to explore human locomotion and wayfinding using a VR Magic Carpet (VMC) [3], a table test version known as the Corsi Block Tapping task (CBT) [4]. This variation was carried out in the context of a virtual reality experimental setup. The participants were required to memorize a sequence of target positions projected on the rug and walk to each target figuring in the displayed sequence. the participant's trajectory was collected and analyzed from a kinematic perspective. An earlier study [5] identified three different categories, but the classification remained ambiguous, implying that they include both kinds of individuals (normal and patients with cognitive spatial impairments). On this basis, we utilized K-Means and HAC to distinguish the navigation behavior of patients from normal individuals, emphasizing the most important discrepancies and then delving deeper to gain more insights. © The Authors.
E3S Web of ConferencesOpen AccessDOI: 10.1051/e3sconf/202235101042 Authors:
In this study, we explore human navigation as evaluated by the VR Magic Carpet TM (VMC) [1], a variation of the Corsi Block Tapping task (CBT) [2, 3], employing Density-based spatial clustering of applications with noise (DBSCAN) [4]. As a result of the VMC, we acquired raw spatial data in 3D space, which we processed, analyzed, and turned into trajectories before evaluating them from a kinematic standpoint. Our previous research [5] revealed three unique groupings. However, the categorization remained ambiguous, showing clusters with diverse people (patients and healthy). We utilized DBSCAN to compare patients’ navigation behavior to healthy individuals, highlighting the most notable differences and assessing our existing classifiers. Our research aims to produce insights that may help clinicians and neuroscientists adopt machine learning, especially clustering algorithms, to identify cognitive impairments and analyze human navigation behavior. © 2022, The Author(s), under exclusive license to Springer Nature Switzerland AG.
Lecture Notes in Networks and SystemsDOI: 10.1007/978-3-031-02447-4_20 Authors:
In this study, We evaluated the efficiency of Multilayer perceptron for classification tasks related to cognitive impairments assessed in a virtual reality environment and on spatial data, “The VR Magic carpet” In our earlier work, we applied machine learning (ML) techniques for assessing and categorizing participants with cognitive impairments. The issue stems from the likelihood of not identifying the most relevant elements that will provide high accuracy in this navigation disorder detection. We used method multilayer perceptron (MLP) architectures to benefit from using layers for feature extraction on velocity time series and solve our classification problem. This navigation disorder identification model was prompt to develop a better understanding of targeting users with navigation disorders. The experimental results of the model in this study provide an enhancement because it can distinguish with more accuracy between healthy individuals and patients. © 2023, The Author(s), under exclusive license to Springer Nature Switzerland AG.
Lecture Notes in Networks and SystemsDOI: 10.1007/978-3-031-29857-8_6 Authors:
In this paper, we assess human navigation in a virtual reality neuropsychological test named the VR magic carpet; the task involves repeating sequences shown by the test administrator via a VR environment built in Unity. The objective of this assessment is to identify patients with vestibular dysfunction. In earlier work, we used unsupervised classical algorithms on an aggregated participant analysis based on kinematic characteristics, specifically head velocity global average. However, experts mentioned the necessity to analyze further and consider other parameters related to head velocity time series extracted during the clinical trials. Knowing that machine learning techniques are based on hand-crafted features may lead to forgetting essential features. The intuitive solution is embedding deep learning in the analysis process. For that reason, Experts launched clinical trials and labeled the extracted files containing raw 3D spatial data to approach the problem as a classification problem. Furthermore, we used residual neural network topologies, with layers used for feature extraction and classification, to establish an identification model for vestibular dysfunction. The experimental results of the model in this study demonstrate an improvement since they more accurately distinguish between healthy and patients. © 2023, The Author(s), under exclusive license to Springer Nature Switzerland AG.
Lecture Notes in Networks and SystemsDOI: 10.1007/978-3-031-29860-8_94 Authors:
Human behavior modeling aims to extract and understand patterns of behavior in one's daily life. A seemingly intuitive walk to a destination following a specific path raises questions about the existence of a model that can explain the choices and strategies predetermined by the brain through physical behavior. Tasks such as finding objects, achieving goals, avoiding obstacles, and returning home can be challenging when one lacks navigational abilities. In our previous work. We employed machine learning (ML) approaches to evaluate and classify cognitively challenged individuals to identify crucial markers that could be used to simulate human navigation behavior. However, the challenge lies in the possibility of missing key elements. Our proposed method utilizes deep time series clustering architectures, in which we employ auto-encoders to extract features and then cluster our data using unsupervised machine learning techniques. This approach will provide us with initial insights into human navigation patterns and meaningful clustering. Our developed framework combines unsupervised learning with a dataset obtained from a neuropsychological evaluation in a virtual reality setting called “The VR Magic Carpet”, a variant of “the Corsi block-tapping task”. © 2023, The Author(s), under exclusive license to Springer Nature Switzerland AG.
Lecture Notes in Networks and SystemsDOI: 10.1007/978-3-031-29860-8_96 Authors:
In this study, we implemented a spatiotemporal clustering approach to analyze the outcome of a virtual reality human navigation neuropsychological assessment (the VR Magic Carpet). Our main objective was to establish a clustering of participants using a deep multi-step clustering model on velocity signals extracted during clinical trials. We used a multi-step neural network architecture to analyze the feature extraction and the clustering stage separately. In the feature extraction stage, we adopted a 1D-DCAE autoencoder, and for the clustering, we used a soft temporal clustering layer, a combination of similarity metrics, K-means, and probability. This method enabled us to comprehend, to a certain extent, the clustering results in contrast to the joint architecture we had been using before. We obtained five significant clusters that are associated with specific clinical groups. © 2023, The Author(s), under exclusive license to Springer Nature Switzerland AG.
Lecture Notes in Networks and SystemsDOI: 10.1007/978-3-031-29860-8_98 Authors:
In this research, we use unsupervised machine learning clustering techniques, notably K-means (Jain in Pattern Recogn Lett 31:651–666, 2010 [1]), to explore human navigation using the VR Magic Carpet (Berthoz and Zaoui in Dev Med Child Neurol 57:15–20, 2015 [2]). This is a variant of the Corsi Block Tapping task (CBT) (Corsi in Human memory and the medial temporal region of the brain. McGill University, 1972 [3]) that was carried out within the experimental framework of virtual reality. The participant’s trajectory was captured as raw spatial data and afterward kinematically evaluated. Our previous research (Annaki et al. in Digital technologies and applications. ICDTA 2021. Lecture notes in networks and systems, vol 211. Springer, Cham, 2021 [4]) found three distinct groups. However, the classification remained unclear, suggesting that they include both types of people (ordinary and patients with cognitive spatial impairments). Based on this premise, we used K-means to distinguish patients’ navigation behavior from that of healthy people, highlighting the most significant differences and validating the feature on which our previous analysis was based. © 2023, The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd.
Lecture Notes in Electrical EngineeringDOI: 10.1007/978-981-19-6223-3_9 Authors:
In this paper, we use computational tools (Cipresso P, Matic, A, Giakoumis D, Ostrovsky Y (2015) Advances in computational psychometrics. Comput Math Methods Med. Article ID 418683. https://doi.org/10.1155/2015/418683.5 ) to explore human navigation through an example of a visuomotor spatial memory locomotor task, the Walking Corsi task (WCT) variant from a well-known table test known as the Corsi Block Tapping task [(CBT) [2] and [15]. This variant was performed using the “Virtual Carpet” ™ experimental setup. The subjects had to memorize a succession of the position of targets projected on the ground and reproduce sequences of 2 to 9 targets by walking to each. The trajectory of the head was recorded and processed from a kinematic point of view. Generic tools that computational data analytics provides and through computer simulations by replicating visually this data allowed categorization of the different features of the behavior of the subjects providing a new powerful tool for both normal and pathological behavior characterization. © 2021, The Author(s), under exclusive license to Springer Nature Switzerland AG.
Lecture Notes in Networks and SystemsDOI: 10.1007/978-3-030-73882-2_22