[서평]「Algorithms on Strings, Trees, and Sequences」

We present a novel online signature verification system based on the Fast Fourier Transform. The advantage of using the Fourier domain is the ability to compactly represent an online signature using a fixed number of coefficients. The fixed-length representation leads to fast matching algorithms and is essential in certain applications. The challenge on the other hand is to find the right preprocessing steps and matching algorithm for this representation. We report on the effectiveness of the proposed method, along with the effects of individual preprocessing and normalization steps, based on comprehensive tests over two public signature databases. We also propose to use the pen-up duration information in identifying forgeries. The best results obtained on the SUSIG-Visual subcorpus and the MCYT-100 database are 6.2% and 12.1% error rate on skilled forgeries, respectively. The fusion of the proposed system with our state-of-the-art Dynamic Time Warping (DTW) system lowers the error rate of the DTW system by up to about 25%. While the current error rates are higher than state-of-the-art results for these databases, as an approach using global features, the system possesses many advantages. Considering also the suggested improvements, the FFT system shows promise both as a stand-alone system and especially in combination with approaches that are based on local features.

/pdf/online-signature-verification-using-fourier-descriptors-36bedzrysg.pdf

Online signature verification using Fourier descriptors

Biometrics and their use in e-passports

Sketching activities are widely adopted during early design phases of user interface development to convey informal specifications of the interface presentation and dialog. Designers or even end users can sketch some or all of the future interface they want. With the ever increasing availability of different computing platforms, a need arises to continuously support sketching across these platforms with their various programming languages, interface development environments and operating systems. To address needs along these dimensions, which pose new challenges to user interface sketching tools, Sketchi XML is a multi-platform multi-agent interactive application that enable designers and end users to sketch user interfaces with different levels of details and support for different contexts of use. The results of the sketching are then analyzed to produce interface specifications independently of any context, including user and platform. These specifications are exploited to progressively produce one or several interfaces, for one or many users, platforms, and environments.

/pdf/a-sketching-tool-for-designing-anyuser-anyplatform-anywhere-1zr31z6hjh.pdf

A sketching tool for designing anyuser, anyplatform, anywhere user interfaces

Dictionary lookup methods are popular in dealing with ambiguous letters which were not recognized by Optical Character Readers. However, a robust dictionary lookup method can be complex as apriori probability calculation or a large dictionary size increases the overhead and the cost of searching. In this context, Levenshtein distance is a simple metric which can be an effective string approximation tool. After observing the effectiveness of this method, an improvement has been made to this method by grouping some similar looking alphabets and reducing the weighted difference among members of the same group. The results showed marked improvement over the traditional Levenshtein distance technique.

Levenshtein Distance Technique in Dictionary Lookup Methods: An Improved Approach

A new method for on-line signature authentication is presented, which is based on a event-string modelling of features derived from pen-position and pressure signals of digitizer tablets. A distance measure well known from textual pattern recognition, the Levenshtein distance, is used for comparison of signatures and classification is carried out applying a nearest neighbor classifier. Results from a test set of 1376 signatures from 41 persons are presented, which have been conducted for four different feature sets. The results are rather encouraging, with correct identification rates of 96% at zero false classifications.

https://wwwiti.cs.uni-magdeburg.de/~vielhaue/jabreflib/[ScVD2004].pdf

Using adapted Levenshtein distance for on-line signature authentication

In this paper we present a new algorithm for automatic recognition of hand drawn sketches based on the Levenshtein distance. The purpose for drawing sketches in our application is to create graphical user interfaces in a similar manner as the well established paper sketching. The new algorithm is trainable by every user and improves the recognition performance of the techniques which were used before for widget recognition. In addition, this algorithm ay serve for recognizing other types of sketches, such as letters, figures, and commands. In this way, there is no modality disruption at sketching time.

/pdf/trainable-sketch-recognizer-for-graphical-user-interface-14jotivnen.pdf

Trainable sketch recognizer for graphical user interface design

In this paper we analyze chances and challenges with respect to the security of using biometrics in ID documents. We identify goals for ID documents, set by national and international authorities, and discuss the degree of security, which is obtainable with the inclusion of biometric into documents like passports. Starting from classical techniques for manual authentication of ID card holders, we expand our view towards automatic methods based on biometrics. We do so by reviewing different human biometric attributes by modality, as well as by discussing possible techniques for storing and handling the particular biometric data on the document. Further, we explore possible vulnerabilities of potential biometric passport systems. Based on the findings of that discussion we will expand upon two exemplary approaches for including digital biometric data in the context of ID documents and present potential risks attack scenarios along with technical aspects such as capacity and robustness.

/pdf/security-analysis-for-biometric-data-in-id-documents-ww1ymarmmi.pdf

Security Analysis for Biometric Data in ID Documents

In this paper we present a strategy for handling of multimodal signals from pen-based mobile devices for Human to Computer Interaction (HCI), where our focus is on the modalities of spoken and handwritten inputs. Each modality for itself is quite well understood, as the exhaustive literature demonstrates, although still a number of challenges exist, like recognition result improvements. Among the potentials in multimodal HCI are improvements in recognition and robustness as well as seamless men-machine communication based on fusion of different modalities by exploiting redundancies among these modalities. However, such valuable fusion of both modalities still poses some problems. Open problems today include design approaches for fusion strategies and with the increasing number of mobile and pen-based computers, particularly techniques for fusion of handwriting and speech appear to have a great potential. But today few publications can be found that addresses this potential. In this work we introduce a conceptional approach based on a model to describe a bimodal HCI process. We analyze four exemplary applications with respect to the structure of this model, and highlight the open problems within these applications. Further, we will outline possible solutions to these challenges. Having such fusion model for HCI may simplify the development of seamless and intuitive to user interfaces on pen-based mobile devices. For one of our application scenarios, a bimodal system for form data recording and recognition in medical or financial environment, we will present some first experimental results.

https://wwwiti.cs.uni-magdeburg.de/~vielhaue/jabreflib/[VSTS2005].pdf

Fusion strategies for speech and handwriting modalities in HCI

With the increasing pervasion of computers, the handwriting seemed to forfeit its position as the primary way of permanent expression of humans ideas; typed texts appeared as the new and better solution. However, with the today’s rise of modern pen based computer devices (e.g. TabletPC), we may see a renaissance of the traditional handwriting in the digital world. More and more electronic documents will be written with pens directly on the screen. One of the benefits of digital documents in comparison to paper, is the convenience of automated document management, including retrieval. For example, full text search in large amounts of sheets of papers is a time-consuming task in the analog world, while modern search engines demonstrate everyday the simplicity of the same task in the future digital world. The continuous breakthrough of digital handwriting will require search possibilities as they exist for typed documents. This paper discusses problems related to search in digital handwriting data and describes a novel approach to solve this searching problem.

Sascha Schimke

Papers

Using adapted Levenshtein distance for on-line signature authentication

Trainable sketch recognizer for graphical user interface design

Security Analysis for Biometric Data in ID Documents

Fusion strategies for speech and handwriting modalities in HCI

Similarity searching for on-line handwritten documents