Search

International audience; In the context of the ACTIF project that aims for active and collaborative learning promotion, this paper presents a pattern recognition and analysis system for Geometry learning in middle school. The goal is to allow students to draw geometric shapes on a touch-tablet, given a teacher's instruction. To make the student active, the system have to recognize and analyze on the fly the student's productions in order to produce real-time visual, corrective, and guidance feedback. We base our work on the visual grammar CD-CMG [1] (Context Driven Constraints Multi-set Grammar), to model the domain prior knowledge and interpret the hand-drawn sketches on the fly. Our first contribution lies in adapting this grammar to the Geometry domain to cover the geometric objects taught in middle school curriculum. Although being expressive enough to model this large scope, the formalism could not cope with the exigence of real-time analysis, given that the multiple interactions between geometric objects generate combinatorial issues. Our second contribution lies in extending the formalism which resulted in having an acceptable performance for a real-time user interaction system. The first experiments show that the proposed approach allows complexity and interpretation time reduction.

Many handwritten signature verification algorithms have been developed in order to distinguish between genuine signatures and forgeries. An important group of these methods is based on dynamic time warping (DTW). Traditional use of DTW for signature verification consists in forming a misalignment score between the verified signature and a set of template signatures. The right selection of template signatures has a big impact on that verification. In this article, we describe our proposition for replacing the template signatures with the hidden signature-an artificial signature which is created by minimizing the mean misalignment between itself and the signatures from the enrollment set. We present a few hidden signature estimation methods together with their comprehensive comparison. The hidden signature opens a number of new possibilities for signature analysis. We apply statistical properties of the hidden signature to normalize the error signal of the verified signature and to use the misalignment on the normalized errors as a verification basis. A result, we achieve satisfying error rates that allow creating an on-line system, ready for operating in a real-world environment

Conference of 12th IEEE International Conference on Automatic Face and Gesture Recognition, FG 2017 ; Conference Date: 30 May 2017 Through 3 June 2017; Conference Code:128713; International audience; In this article, we introduce the DAily Home LIfe Activity (DAHLIA) Dataset, a new dataset adapted to the context of smart-home or video-assistance. Videos were recorded in realistic conditions, with 3 KinectTMv2 sensors located as they would be in a real context. The very long-range activities were performed in an unconstrained way (participants received few instructions), and in a continuous (untrimmed) sequence, resulting in long videos (39 min in average per subject). Contrary to previously published databases, in which labeled actions are very short and with low-semantic level, this new database focuses on high-level semantic activities such as 'Preparing lunch' or 'House Working'. As a baseline, we evaluated several metrics on three different algorithms designed for online action recognition or detection.

In this thesis, we have investigated the human object interaction recognitionby using the skeleton data and local depth information provided byRGB-D sensors. There are two main applications we address in this thesis:human object interaction recognition and abnormal activity recognition.First, we propose a spatio-temporal modeling of human-object interactionvideos for on-line and off-line recognition. In the spatial modelingof human object interactions, we propose low-level feature and object relateddistance feature which adopted on on-line human object interactionrecognition and abnormal gait detection. Then, we propose object feature,a rough description of the object shape and size as new features tomodel human-object interactions. This object feature is fused with thelow-level feature for online human object interaction recognition. In thetemporal modeling of human object interactions, we proposed a shapeanalysis framework based on low-level feature and object related distancefeature for full sequence-based off-line recognition. Experiments carriedout on two representative benchmarks demonstrate the proposed methodare effective and discriminative for human object interaction analysis.Second, we extend the study to abnormal gait detection by using theon-line framework of human object interaction classification. The experimentsconducted following state-of-the-art settings on the benchmarkshows the effectiveness of proposed method.Finally, we collected a multi-view human object interaction dataset involvingabnormal and normal human behaviors by RGB-D sensors. Wetest our model on the new dataset and evaluate the potential of the proposedapproach.; Dans cette thèse, nous avons étudié la reconnaissance des actions quiincluent l’intéraction avec l’objet à partir des données du skeleton et desinformations de profondeur fournies par les capteurs RGB-D. Il existedeux principales applications que nous abordons dans cette thèse: lareconnaissance de l’interaction humaine avec l’objet et la reconnaissanced’une activité anormale.Nous proposons, dan un premier temps, une modélisation spatiotemporellepour la reconnaissance en ligne et hors ligne des intéractionsentre l’humain et l’objet. Dans la modélisation spatiale, nous proposonsdes caractéristiques de bas niveau liés à la distance entre les points duskeleton et la distance entre l’objet et les points du skeleton. Ces caractéristiquesont été adoptées pour la reconnaissance en ligne des intéractionshumaines avec l’objet et pour la détection de la démarche anormale.Ensuite, nous proposons des caractéristiques liées à d’objet qui décriventapproximativement la forme et la taille de l’objet. Ces caractéristiques sontfusionnées avec les caractéristiques bas-niveau pour la reconnaissance enligne des intéractions humaines avec l’objet. Dans la modélisation temporelle,nous avons proposé un framework élastique pour aligner les trajectoiresdes distances dans le temps afin de permettre une reconnaissancehors ligne invariante au taux d’exécution. Les expériences menées surdeux benchmarks démontrent l’efficacité de la méthode proposée. Dans ledeuxième volet de ce travail, nous étendons l’étude à la détection de la démarcheanormale en utilisant le cadre en ligne l’approche. Afin de validerla robustesse de l’approche à la pose, nous avons collecté une base multivuespour des intéractions humaines avec l’objet, de façon normale etanormale. Les résultats expérimentaux sur le benchmark des actions anormalesfrontales et sur la nouvelles base prouvent l’efficacité de l’approche.

In this thesis, we have investigated the human object interaction recognitionby using the skeleton data and local depth information provided byRGB-D sensors. There are two main applications we address in this thesis:human object interaction recognition and abnormal activity recognition.First, we propose a spatio-temporal modeling of human-object interactionvideos for on-line and off-line recognition. In the spatial modelingof human object interactions, we propose low-level feature and object relateddistance feature which adopted on on-line human object interactionrecognition and abnormal gait detection. Then, we propose object feature,a rough description of the object shape and size as new features tomodel human-object interactions. This object feature is fused with thelow-level feature for online human object interaction recognition. In thetemporal modeling of human object interactions, we proposed a shapeanalysis framework based on low-level feature and object related distancefeature for full sequence-based off-line recognition. Experiments carriedout on two representative benchmarks demonstrate the proposed methodare effective and discriminative for human object interaction analysis.Second, we extend the study to abnormal gait detection by using theon-line framework of human object interaction classification. The experimentsconducted following state-of-the-art settings on the benchmarkshows the effectiveness of proposed method.Finally, we collected a multi-view human object interaction dataset involvingabnormal and normal human behaviors by RGB-D sensors. Wetest our model on the new dataset and evaluate the potential of the proposedapproach.; Dans cette thèse, nous avons étudié la reconnaissance des actions quiincluent l’intéraction avec l’objet à partir des données du skeleton et desinformations de profondeur fournies par les capteurs RGB-D. Il existedeux principales applications que nous abordons dans cette thèse: lareconnaissance de l’interaction humaine avec l’objet et la reconnaissanced’une activité anormale.Nous proposons, dan un premier temps, une modélisation spatiotemporellepour la reconnaissance en ligne et hors ligne des intéractionsentre l’humain et l’objet. Dans la modélisation spatiale, nous proposonsdes caractéristiques de bas niveau liés à la distance entre les points duskeleton et la distance entre l’objet et les points du skeleton. Ces caractéristiquesont été adoptées pour la reconnaissance en ligne des intéractionshumaines avec l’objet et pour la détection de la démarche anormale.Ensuite, nous proposons des caractéristiques liées à d’objet qui décriventapproximativement la forme et la taille de l’objet. Ces caractéristiques sontfusionnées avec les caractéristiques bas-niveau pour la reconnaissance enligne des intéractions humaines avec l’objet. Dans la modélisation temporelle,nous avons proposé un framework élastique pour aligner les trajectoiresdes distances dans le temps afin de permettre une reconnaissancehors ligne invariante au taux d’exécution. Les expériences menées surdeux benchmarks démontrent l’efficacité de la méthode proposée. Dans ledeuxième volet de ce travail, nous étendons l’étude à la détection de la démarcheanormale en utilisant le cadre en ligne l’approche. Afin de validerla robustesse de l’approche à la pose, nous avons collecté une base multivuespour des intéractions humaines avec l’objet, de façon normale etanormale. Les résultats expérimentaux sur le benchmark des actions anormalesfrontales et sur la nouvelles base prouvent l’efficacité de l’approche.

Cyber-Physical Systems (CPS) are tightly coupled with the environment, and therefore it is important that interactions with the surroundings like Human-Computer-Interactions are performed very responsive. Since CPS are often embedded without traditional input devices, like in medical or automotive contexts, gesture recognition approaches are emerging. As those algorithms are computationally complex especially when implemented on multi-core architectures, design decisions have to be taken carefully in order to meet performance and energy constraints. In this paper, we present a Scenario-Aware Dataflow model to estimate the performance of a template-based hand gesture recognition system based on Dynamic Time Warping (DTW). Our model enables us to estimate the important characteristics like real-time capabilities for online recognition and response time of the system when implemented on a multi-core architecture. Moreover, we introduce an extension to existing SADF performance analysis tools, which enables us to acquire processor utilization from our model. Based on the performance estimations the real-time capability for online recognition was validated for different configurations and verified in our experiments.

Cyber-Physical Systems (CPS) are tightly coupled with the environment, and therefore it is important that interactions with the surroundings like Human-Computer-Interactions are performed very responsive. Since CPS are often embedded without traditional input devices, like in medical or automotive contexts, gesture recognition approaches are emerging. As those algorithms are computationally complex especially when implemented on multi-core architectures, design decisions have to be taken carefully in order to meet performance and energy constraints. In this paper, we present a Scenario-Aware Dataflow model to estimate the performance of a template-based hand gesture recognition system based on Dynamic Time Warping (DTW). Our model enables us to estimate the important characteristics like real-time capabilities for online recognition and response time of the system when implemented on a multi-core architecture. Moreover, we introduce an extension to existing SADF performance analysis tools, which enables us to acquire processor utilization from our model. Based on the performance estimations the real-time capability for online recognition was validated for different configurations and verified in our experiments.

Cataloged from PDF version of article. In this paper we present a video based text and equation editor for LaTeX. The system recognizes what is written onto paper and generates the LaTeX code. Text and equations are written on a regular paper using a board marker, and a USB camera attached to a computer is used to capture and record the pen-tip positions in each consecutive image frame. Characters and symbols are represented as separate finite state machines (FSMs). They are written in an isolated manner and they are recognized on-line using the FSMs. In the last step, LaTeX code corresponding to recognized characters and symbols is generated. (c) 2007 Elsevier Ltd. All rights reserved.

International audience; This paper presents a method to automatically locate and recognize erasures in on-line captured handwritten documents in order to avoid a subsequent misrecognition of characters and words. We offer a comprehensive definition of the ambiguous concept of erasure in handwriting that results in a more accurate characterization of the different types of erasures. Thanks to this characterization, a preprocessing step, placed upstream of the word recognition engine, enables to classify through an MLP each couple of connected strokes as being an erasure or not using a low-level feature set. We evaluate our system on a real handwritten document database and show how our system can be tuned to operate in accordance with various recognition engines thus leading to high performance in erasure detection and recognition.

Given a language L that is online recognizable in linear time and space, we construct a linear time and space online recognition algorithm for the language L・Pal, where Pal is the language of all nonempty palindromes. Hence for every fixed positive k, Palk is online recognizable in linear time and space. Thus we solve an open problem posed by Galil and Seiferas in 1978. © Springer-Verlag Berlin Heidelberg 2015.

Personal use of this material is permitted. Permission from IEEE must be obtained for all other uses, in any current or future media, including reprinting/republishing this material for advertising or promotional purposes, creating new collective works, for resale or redistribution to servers or lists, or reuse of any copyrighted component of this work in other works. F. Alonso-Fernández, J. Fiérrez, M. Martínez-Diaz, J. Ortega-García, "Fusion of static image and dynamic information for signature verification" in 16th Conference on Image Processing (ICIP), Cairo (Egypt), 2009, pp. 2725 - 2728, This paper evaluates the combination of static image (off-line) and dynamic information (on-line) for signature verification. Two off-line and two on-line recognition approaches exploiting information at the global and local levels are used. Experimental results are given using the BiosecurID database (130 signers, 3,640 signatures). Fusion experiments are done using a trained fusion approach based on linear logistic regression. It is shown experimentally that the local systems outperform the global ones, both in the on-line and in the off-line case. We also observe a considerable improvement when combining the two on-line systems, which is not the case with the off-line systems. The best performance is obtained when fusing all the systems together, which is specially evident for skilled forgeries when enough training data is available., This work has been supported by Spanish MCYT TEC2006-13141-C03-03 project.

The last two decades witnessed some advances in the development of an Arabic character recognition (CR) system. Arabic CR faces technical problems not encountered in any other language that make Arabic CR systems achieve relatively low accuracy and retards establishing them as market products. We propose the basic stages towards a system that attacks the problem of recognizing online Arabic cursive handwriting. Rule-based methods are used to perform simultaneous segmentation and recognition of word portions in an unconstrained cursively handwritten document using dynamic programming. The output of these stages is in the form of a ranked list of the possible decisions. A new technique for text line separation is also used., {"references":["Re-jean Plamondon and N. Srihari, \"On-line and off-line handwriting\nrecognition: a comprehensive survey,\" IEEE transactions on pattern\nanalysis and machine intelligence, vol. 22, No. 1, January 2000.","Yasser Hifny, \"On-line Arabic handwriting recognition,\" M.S. thesis,\nDept. Communication and Electronics Eng., Cairo Univ., Egypt, 2000.","Hazem Y. Abdelazim, \"Recent trends in Arabic OCR,\" in Proc. 5th\nConference of Engineering Language, Ain Shams University, 2005.","R. A. Cole, J. Mariani, H. Uszkoreit, A. Zaenen, and V. Zue, Survey of\nthe State of the Art in Human Language Technology. Center for Spoken\nLanguage Understanding CSLU, Carnegie Mellon University,\nPittsburgh, PA, 1995.","E.H. Ratzlaff, \"Inter-line Distance Estimation and Text Line Extraction\nFor Unconstrained Online Handwriting,\" in Proc. 7th International\nWorkshop on Frontiers in Handwriting Recognition, pp. 33-42, 2000.","Gareth Loudon, Olle Pellijeff, and LI Zhong-Wei, \"A Method for\nHandwriting Input and Correction on Smartphones,\" in Proc. 7th\nInternational Workshop on Frontiers in Handwriting Recognition, pp.\n481-485, 2000.","H. Shimodaira, T. Sudo, M. Nakai, and S. Sagayama, \"On-line Overlaid-\nHandwriting Recognition Based on Substroke HMMs,\" in Proc. ICDAR,\n2003.","J. Lee, J. Kim, and J. H. Kim, \"Data driven design of HMM topology for\non-line handwriting recognition,\" in Proc. 7th International Workshop\non Frontiers in Handwriting Recognition, pp. 239-248, 2000.","Han Shu, \"On-Line Handwriting Recognition Using Hidden Markov\nModels,\" M.S. thesis, Massachusetts Institute of Technology, 1997.\n[10] Daniel Jurafski and James H. Martin, \"Speech and Language Processing:\nAn introduction to Natural Language processing, computational\nLinguistic and Speech recognition,\" Prentice-Hall, 2000, pp. 156."]}

http://www.suvisoft.com; In this paper, we describe a preprocessing system which locates erasures in on-line captured handwritten documents. Our approach is conceived so as to be placed upstream of the handwritten recognition engine. This system classifies each couple of connected strokes using a low level feature set and a multi layer perceptron classifier. One part of this study gives an efficient definition of erasure, which results in splitting the two original classes of the problem into nineteen more accurate sub-classes. The tunable tolerance level of the system provides a good flexibility to operate in accordance with various recognition engines. We evaluate our system on a real document database and present encouraging performance results.

This paper describes a compact on-line/off-line combined handwriting recognizer for Japanese characters. Conventional combined recognizers mainly consider the recognition accuracy, though recognition speed and memory size are important as well. Especially, the off-line method requires a large prototype dictionary, and therefore the on-line/off-line combined recognizers are difficult to use practically in a small computer. In order to tackle this problem, we propose an on-line/offline combined recognizer where an off-line recognizer is composed of the Modified Quadratic Discriminant Function whose dictionary size is significantly reduced. Moreover, its on-line recognizer is composed of a structured character pattern representation (SCPR) dictionary which reduces the total size of memory and Linear-time Elastic Matching (LTM) which reduces the computation time. Experimental results show that the cumulative recognition rate of top 5 candidates of a proposed 1MB dictionary is 97.3% (almost the same as that of a conventional 90MB dictionary) and that the recognition speed of the 1MB dictionary is 1.75 times faster then that of the 90MB dictionary.