Showing papers on "Sketch recognition published in 2005"
15 Oct 2005
TL;DR: It is shown that the direct 3D counterparts to commonly used 2D interest point detectors are inadequate, and an alternative is proposed, and a recognition algorithm based on spatio-temporally windowed data is devised.
Abstract: A common trend in object recognition is to detect and leverage the use of sparse, informative feature points. The use of such features makes the problem more manageable while providing increased robustness to noise and pose variation. In this work we develop an extension of these ideas to the spatio-temporal case. For this purpose, we show that the direct 3D counterparts to commonly used 2D interest point detectors are inadequate, and we propose an alternative. Anchoring off of these interest points, we devise a recognition algorithm based on spatio-temporally windowed data. We present recognition results on a variety of datasets including both human and rodent behavior.
2,699 citations
20 Jun 2005
TL;DR: This paper presents a face recognition system based on face sketches that is based on pseudo-sketch synthesis and sketch recognition, and experimental results show that the performance of the proposed method is encouraging.
Abstract: Most face recognition systems focus on photo-based face recognition. In this paper, we present a face recognition system based on face sketches. The proposed system contains two elements: pseudo-sketch synthesis and sketch recognition. The pseudo-sketch generation method is based on local linear preserving of geometry between photo and sketch images, which is inspired by the idea of locally linear embedding. The nonlinear discriminate analysis is used to recognize the probe sketch from the synthesized pseudo-sketches. Experimental results on over 600 photo-sketch pairs show that the performance of the proposed method is encouraging.
352 citations
TL;DR: In this paper, a language to describe how sketched diagrams in a domain are drawn, displayed, and edited is presented, which is then automatically transformed into domain specific shape recognizers, editing recognizers and shape exhibitors for use in conjunction with a domain independent sketch recognition system.
195 citations
10 Jan 2005
TL;DR: Results of a user study indicating that in certain domains people draw objects using consistent stroke orderings are reported, which enables this novel approach to have polynomial time algorithms for sketch recognition and segmentation, unlike conventional methods with exponential complexity.
Abstract: Current sketch recognition systems treat sketches as images or a collection of strokes, rather than viewing sketching as an interactive and incremental process. We show how viewing sketching as an interactive process allows us to recognize sketches using Hidden Markov Models. We report results of a user study indicating that in certain domains people draw objects using consistent stroke orderings. We show how this consistency, when present, can be used to perform sketch recognition efficiently. This novel approach enables us to have polynomial time algorithms for sketch recognition and segmentation, unlike conventional methods with exponential complexity.
163 citations
TL;DR: AC-SPARC as mentioned in this paper is a sketch-based interface for the SPICE electric circuit analysis program that automatically locates symbols by looking for areas of high ink density and for points at which the characteristics of the pen strokes change.
110 citations
31 Jul 2005
TL;DR: In this article, a sketch-recognition and modeling algorithm is proposed to match the points and curves of a set of given 2D templates to the sketch, and a 3D object is constructed using a series of measurements that are extracted from the labelled 2D sketch.
Abstract: Sketch-based modeling holds the promise of making 3D modeling accessib le to a significantly wider audience than current modeling tools. We present a modeling system that is capable of constructing 3D models of particular object classes from 2D sketches. The core of the system is a sketch reco gnition algorithm that seeks to match the points and curves of a set of given 2D templates to the sketch. The matching process employs an optimization metric that is based on curve feature vectors, and the search space of p ossible correspondences is restricted by encoding knowledge about relative part locations into the 2D template. Once a best-fit template is found, a 3D object is constructed using a series of measurements that are extracted from the labelled 2D sketch. We apply our sketch-recognition and modeling algorithms to sketches of cups and mugs , airplanes, and fish. The system allows non-experts to use drawings to quickly create 3D models of specific objec t classes.
74 citations
18 Oct 2005
TL;DR: In this paper, a probabilistic framework is proposed to extract information from hand activity without requiring that the wearer is explicit as in gesture-based interaction, and the results are used to build a visual summary of events.
Abstract: In this paper we develop a first step towards the recognition of hand activity by detecting objects subject to manipulation, and use the results to build a visual summary of events. The motivation is to extract information from hand activity without requiring that the wearer is explicit as in gesture-based interaction. Our method uses simple image measurements within a probabilistic framework and allows real-time implementation.
67 citations
Proceedings Article•
30 Jul 2005TL;DR: A novel form of dynamically constructed Bayes net, developed for multi-domain sketch recognition, integrating the influence of stroke data and domain-specific context in recognition, enabling the recognition engine to handle noisy input.
Abstract: This paper presents a novel form of dynamically constructed Bayes net, developed for multidomain sketch recognition. Our sketch recognition engine integrates shape information and domain knowledge to improve recognition accuracy across a variety of domains using an extendible, hierarchical approach. Our Bayes net framework integrates the influence of stroke data and domain-specific context in recognition, enabling our recognition engine to handle noisy input. We illustrate this behavior with qualitative and quantitative results in two domains: hand-drawn family trees and circuits.
48 citations
31 Aug 2005
TL;DR: This paper proposes a grammar formalism, namely Sketch Grammars (SkGs), for describing both the shape of the symbols' language and the syntax of sketch languages, which are automatically generated from the sketch grammar descriptions.
Abstract: Sketch-based user interfaces are increasingly common and are being built for a variety of different disciplines. However, at present the implementation of sketch recognizers is quite time consuming since they are mostly based on specific techniques, as opposed to several other fields such as textual/visual languages and speech recognition, which benefit from the availability of compiler generation techniques and tools. This paper proposes a grammar formalism, namely Sketch Grammars (SkGs), for describing both the shape of the symbols' language and the syntax of sketch languages. Recognizers are automatically generated from the sketch grammar descriptions.
39 citations
01 Jan 2005
TL;DR: A viewpoint independent method for sign recognition that can reach both temporal and viewpoint invariance and can be easily extended to other recognition tasks, such as gait recognition and lip-reading recognition.
Abstract: Sign language is the primary modality of communication among deaf and mute society all over the world. This paper proposes a viewpoint independent method for sign recognition. Considering that two sequences of the same sign can be roughly considered as the input of a stereo vision system after time-warping, and the fundamental matrix associated with two views should be unique, we can convert the temporal-spatial recognition task as a verification task within a stereo vision framework. After time-warping of the input sequences, the proposed framework can reach both temporal and viewpoint invariance. We demonstrate the efficiency of the proposed framework by recognizing a vocabulary of 100 words of Chinese sign language. The recognition rate is up to 97% at rank 3. Furthermore, the proposed framework can be easily extended to other recognition tasks, such as gait recognition and lip-reading recognition.
33 citations
Posted Content•
TL;DR: A simplified neural approach to recognition of optical or visual characters is portrayed and discussed and is expected to serve as a resource for learners and amateur investigators in pattern recognition, neural networking and related disciplines.
Abstract: The recognition of optical characters is known to be one of the earliest applications of Artificial Neural Networks, which partially emulate human thinking in the domain of artificial intelligence. In this paper, a simplified neural approach to recognition of optical or visual characters is portrayed and discussed. The document is expected to serve as a resource for learners and amateur investigators in pattern recognition, neural networking and related disciplines.
03 Oct 2005
TL;DR: An approach to automatic visual recognition of expressive face and upper body action units (FAUs and BAUs) suitable for use in a vision-based affective multimodal framework and fuse face and body information for classification into combined emotion categories is presented.
Abstract: Research shows that humans are more likely to consider computers to be human-like when those computers understand and display appropriate nonverbal communicative behavior. Most of the existing systems attempting to analyze the human nonverbal behavior focus only on the face; research that aims to integrate gesture as an expression mean has only recently emerged. This paper presents an approach to automatic visual recognition of expressive face and upper body action units (FAUs and BAUs) suitable for use in a vision-based affective multimodal framework. After describing the feature extraction techniques, classification results from three subjects are presented. Firstly, individual classifiers are trained separately with face and body features for classification into FAU and BAU categories. Secondly, the same procedure is applied for classification into labeled emotion categories. Finally, we fuse face and body information for classification into combined emotion categories. In our experiments, the emotion classification using the two modalities achieved a better recognition accuracy outperforming the classification using the individual face modality.
10 Jan 2005
TL;DR: A sketch recognition algorithm based on incremental intention extraction that can recognize kinds of sketches in real time by defining the lag window and updating the existing intention sections according to the latest information is presented.
Abstract: On-line synchronous sketch recognition has the advantages of convenient input and natural interaction. But among the existing algorithms, some are just able to process simple sketches, and some have so high computational complexity as not to satisfy the real-time demand. In order to solve the problem of efficiency and coverage, a sketch recognition algorithm based on incremental intention extraction is presented. By defining the lag window, the algorithm understands the sketch intention of users on the base of incremental intention extraction. Moreover, the algorithm can update the existing intention sections according to the latest information in order that the recognition results are in line with the sketch intention of users. Experiments show that, the algorithm can recognize kinds of sketches in real time.
01 Jan 2005
TL;DR: An overview of recent methodological advances in developing nearest-neighbor-based recommender systems that have substantially improved their performance and the use of statistical learning and lower-dimensional representations to handle issues associated with data sparsity is presented.
Abstract: This article presents an overview of recent methodological advances in developing nearest-neighbor-based recommender systems that have substantially improved their performance. The key components in these methods are: (i) the use of statistical learning to estimate from the data the desired user-user and item-item similarity matrices, (ii) the use of lower-dimensional representations to handle issues associated with data sparsity, (iii) the combination of neighborhood and latent space models, and (iv) the direct incorporation of auxiliary information during model estimation. The article will also provide illustrative examples for these methods in the context of item-item nearest-neighbor methods for rating prediction and Top-N recommendation. In addition, the article will present an overview of exciting new application areas of recommender systems along with the challenges and opportunities associated with them.
Dissertation•
01 Jan 2005
TL;DR: Analysis is presented that shows that the data streams captured can be readily processed to detect gestures and coordinated activity and some pertinent research that can be pursued with these nodes in the areas of biomotion analysis and interactive entertainment are introduced.
Abstract: This thesis describes the design and implementation of a wireless 6 degree-of-freedom inertial sensor system to be used for multiple-user, real-time gesture recognition and coordinated activity detection. Analysis is presented that shows that the data streams captured can be readily processed to detect gestures and coordinated activity. Finally, some pertinent research that can be pursued with these nodes in the areas of biomotion analysis and interactive entertainment are introduced. Thesis Supervisor: Joseph A. Paradiso Title: Sony Career Development Prof. of Media Arts and Sciences Associate Professor, MIT Media Lab
31 Aug 2005
TL;DR: A pre-classification strategy is used, in combination with elastic matching, to improve recognition speed and prune prototypes by examining character features in a model-based method.
Abstract: Natural and convenient mathematical handwriting recognition requires recognizers for large sets of handwritten symbols. This paper presents a recognition system for such handwritten mathematical symbols. We use a pre-classification strategy, in combination with elastic matching, to improve recognition speed. Elastic matching is a model-based method that involves computation proportional to the set of candidate models. To solve this problem, we prune prototypes by examining character features. To this end, we have defined and analyzed different features. By applying these features into an elastic recognition system, the recognition speed is improved while maintaining high recognition accuracy.
22 Oct 2005
TL;DR: A sketch-based graphics input prototype system designed for creative brainstorming in conceptual design is introduced and two core technologies for implementing such a system, adaptive sketch recognition and dynamic user modeling are outlined.
Abstract: This paper explores a concept of sketch-based informal user interface for graphic computing, which can be characterized by two properties: stroke-based input and perceptual processing of strokes. A sketch-based graphics input prototype system designed for creative brainstorming in conceptual design is introduced. Two core technologies for implementing such a system, adaptive sketch recognition and dynamic user modeling, are also outlined.
01 Jan 2005
TL;DR: The proposed method, based on adaptive threshold and fuzzy knowledge with respect to curve's linearity and convexity, can identify sketch strokes (curves) into lines, circles, arcs, ellipses, elliptical arcs, loop lines, spring lines and free-form B-spline curves.
Abstract: This paper presents an intelligent method for classifying pen strokes in an on-line sketching system The method, based on adaptive threshold and fuzzy knowledge with respect to curve's linearity and convexity, can identify sketch strokes (curves) into lines, circles, arcs, ellipses, elliptical arcs, loop lines, spring lines and free-form B-spline curves The proposed method has proven to be fast, suitable for real-time classification and identification
18 Aug 2005
TL;DR: This paper presents a novel approach for adaptive online multi-stroke sketch recognition based on Hidden Markov Model that views the drawing sketch as the result of a stochastic process that is governed by a hidden Stochastic model and identified according to its probability of generating the output.
Abstract: This paper presents a novel approach for adaptive online multi-stroke sketch recognition based on Hidden Markov Model (HMM). The method views the drawing sketch as the result of a stochastic process that is governed by a hidden stochastic model and identified according to its probability of generating the output. To capture a user’s drawing habits, a composite feature combining both geometric and dynamic characteristics of sketching is defined for sketch representation. To implement the stochastic process of online multi-stroke sketch recognition, multi-stroke sketching is modeled as an HMM chain while the strokes are mapped as different HMM states. To fit the requirement of adaptive online sketch recognition, a variable state-number determining method for HMM is also proposed. The experiments prove both the effectiveness and efficiency of the proposed method.
07 Nov 2005
TL;DR: A new method to determine HMM state-number is introduced, based on which an adaptive HMM sketch recognizer is constructed, and a combined feature based on curvature, velocity and geometrical character of stroke for sketch recognition is proposed to improve recognition accuracy.
Abstract: This paper describes a new approach for on-line multi-stroke sketch recognition. The approach is based on hidden Markov model (HMM). Sketches are modeled to HMM chains, and strokes are mapped to different HMM states. The proposed approach introduces a new method to determine HMM state-number, based on which an adaptive HMM sketch recognizer is constructed. A combined feature based on curvature, velocity and geometrical character of stroke for sketch recognition is also proposed to improve recognition accuracy. Finally, the experiments prove the effectiveness and efficiency of the proposed approach.
03 Oct 2005
TL;DR: The results obtained from applying well-known techniques for pattern recognition to emotion recognition in speech are described and the novelty of the paper relies on testing these techniques over non-structured utterances.
Abstract: The aim of our research is to provide robots the capability to interpret the emotions of their counterparts. The lack of automatic recognition of emotional expression in speech presents an important challenge to the pattern analysis and human-robot interaction research community. Emotion recognition in speech provides useful information of the state of the speaker which enhances communication between humans and robots. In this paper, the results obtained from applying well-known techniques for pattern recognition to emotion recognition in speech are described. The novelty of the paper relies on testing these techniques over non-structured utterances. While classifiers have been trained with synthetic data, the validation set has been created using utterances extracted from movies. Classifiers' performance has been measured against human classification for the same validation set.
01 Jan 2005
TL;DR: This work presents a modeling system that is capable of constructing 3D models of particular object classes from 2D sketches, and applies its sketch-recognition and modeling algorithms to sketches of cups and mugs, airplanes, and fish.
Abstract: Sketch-based modeling holds the promise of making 3D modeling accessib le to a significantly wider audience than current modeling tools. We present a modeling system that is capable of constructing 3D models of particular object classes from 2D sketches. The core of the system is a sketch reco gnition algorithm that seeks to match the points and curves of a set of given 2D templates to the sketch. The matching process employs an optimization metric that is based on curve feature vectors, and the search space of p ossible correspondences is restricted by encoding knowledge about relative part locations into the 2D template. Once a best-fit template is found, a 3D object is constructed using a series of measurements that are extracted from the labelled 2D sketch. We apply our sketch-recognition and modeling algorithms to sketches of cups and mugs , airplanes, and fish. The system allows non-experts to use drawings to quickly create 3D models of specific objec t classes.
27 Aug 2005
TL;DR: A three-dimensional gesture recognition algorithm and a system that adopts the algorithm for non-contact human-computer interaction and introduces principal component analysis method to get more robust gesture recognition results.
Abstract: User-friendly Human-Computer interaction becomes more important accordance with rapid development of various information systems. In this paper we describe a three-dimensional gesture recognition algorithm and a system that adopts the algorithm for non-contact human-computer interaction. From sequence of stereo images, five feature regions are extracted with simple color segmentation algorithm and then those are used for three dimensional locus calculation processing. However, the result is not so stable, noisy, that we introduce principal component analysis method to get more robust gesture recognition results. This method can overcome the weakness of conventional algorithms since it directly uses three-dimensional information for human gesture recognition.
31 Aug 2005
TL;DR: This paper investigates the generation and use of multiple recognition results to improve the performance of an offline handwritten text line recognition system and the ROVER algorithm is applied to combine the multiple results.
Abstract: This paper investigates the generation and use of multiple recognition results to improve the performance of an offline handwritten text line recognition system. Multiple recognition results are created by specific integration of a language model in the hidden Markov model based recognition system. The ROVER algorithm is applied to combine the multiple results. Experiments conducted on the IAM database show that the proposed system is able to produce statistically significant improvements in the recognition rate compared to the original system.
22 Oct 2005
TL;DR: A novel multi-stroke sketch recognition method is presented, in which both stroke segmentation and sketch recognition are uniformed as a problem of “fitting to a template” with a minimal fitting error, and a nesting Dynamic Programming algorithm is designed to accelerate the optimizing approach.
Abstract: In this paper a novel multi-stroke sketch recognition method is presented. This method integrates the stroke segmentation and sketch recognition into a single approach, in which both stroke segmentation and sketch recognition are uniformed as a problem of “fitting to a template” with a minimal fitting error, and a nesting Dynamic Programming algorithm is designed to accelerate the optimizing approach.
10 Oct 2005
TL;DR: A 2-staged recognition process which gives robustness to the hand-posture recognition against view-point variation and the composition methods of multiple recognition results from each camera to get the final decision are shown.
Abstract: This paper deals with a method to recognize hand-postures in Softremocon system. The system uses multiple cameras to recognize the user's hand-posture. It is hard to recognize hand-posture since a human-hand is the object with high degree of freedom and there follows the self-occlusion problem, the well-known problem in vision-based recognition area. It can be a possible solution to use multiple cameras. However, when we use multiple images, another problem is arisen, that is, the composition methods of multiple recognition results from each camera to get the final decision. We show 2-staged recognition process. The first stage gets the results which represent each camera viewpoint. And the second one results out the final decision. This structure gives robustness to the hand-posture recognition against view-point variation.
22 Aug 2005
TL;DR: Geographical Sketch Query Language (GSQL) as discussed by the authors is a sketch-based approach for querying geographical databases based on the GeoPQL query language, which allows users to easily sketch their/her geographical query by drawing it, erasing and modifying its parts and highlighting the query target.
Abstract: This paper presents GSQL (Geographical Sketch Query Language), a sketch-based approach for querying geographical databases based on the GeoPQL query language. Geographic information is intrinsically spatial and can be conveniently represented using a bi-dimensional space. Graphical User Interfaces and Visual Languages can be used to satisfy this need. However, the growing availability of sketching tools (PDA, digital pens, etc.) enables a more informal and natural user interaction. Sketch based interaction is very effective. Each user can easily sketch his/her geographical query by drawing it, erasing and modifying its parts and highlighting the query target. A query is the expression of the configuration of the expected result. Sketch recognition and query interpretation (and solution of their ambiguities) starts from a context- independent approach and uses the characteristic application domain information. Context-independent sketch interpretation uses spatial and temporal information related to the sketching process. Context-dependent sketch interpretation uses geographic domain information to solve the remaining ambiguities and correctly interpret the drawing and query. An analysis of the ambiguities characterising object sketching in the geographic application domain and their possible solutions are presented herein. Each query identifies the set of geographical objects involved and the target; the query interpretation must unambiguously identify the set of its results.
01 Jan 2005
TL;DR: The main objective of this research was to study freehand architectural sketches and computer algorithms to develop a sketch recognition software and to apply software with basic 3D modelling.
Abstract: The main objective of this research was to study freehand architectural sketches and computer algorithms to develop a sketch recognition software and to apply software with basic 3D modelling. This research was conducted by collecting the data of sketch recognition and related software to analyst relevant information and to develop new software that would be practical for architects.
23 Jul 2005
TL;DR: This paper presents novel algorithms for sketch recognition and for part identification, and evaluates the accuracy of the recognition algorithm on sketches obtained from medical students.
Abstract: Sketching is ubiquitous in medicine. Physicians commonly use sketches as part of their note taking in patient records and to help convey diagnoses and treatments to patients. Medical students frequently use sketches to help them think through clinical problems in individual and group problem solving. Applications ranging from automated patient records to medical education software could benefit greatly from the richer and more natural interfaces that would be enabled by the ability to understand sketches. In this paper we take the first steps toward developing a system that can understand anatomical sketches. Understanding an anatomical sketch requires the ability to recognize what anatomical structure has been sketched and from what view (e.g. parietal view of the brain), as well as to identify the anatomical parts and their locations in the sketch (e.g. parts of the brain), even if they have not been explicitly drawn. We present novel algorithms for sketch recognition and for part identification. We evaluate the accuracy of the recognition algorithm on sketches obtained from medical students. We evaluate the part identification algorithm by comparing its results to the judgment of an experienced physician.