Yun Koo Chung

Using geometric extrema for segment-to-segment characteristics comparison in online signature verification

Abstract A signature verification algorithm based on a segment-to-segment matching is described, in which a new approach for finding segment-to-segment correspondence is proposed. In the proposed method, geometric extrema are used as segmenting boundaries with two reasons: (1) extrema are reproduced with high stability and (2) the properties of extrema are useful in finding segment-to-segment correspondence. In this paper, a set of rules for legitimate correspondence between extrema is defined based on their properties, which is utilized for the similarity evaluation between segments. Dynamic programming is then applied to find an optimal correspondence map. Through experiments, the proposed algorithm is assessed to be effective in improving discriminative ability. We also showed that the overall performance is further improved by combining the proposed method with the traditional global parametric algorithm.

Legal amount recognition based on the segmentation hypotheses for bank check processing

A sophisticated methodology of legal amount recognition based on the word segmentation hypotheses is introduced for automatic bank check processing. Word segmentation hypotheses are derived according to the grapheme level segmentation results of the legal amount. Novel hybrid schemes of HMM-MLP classifiers are also introduced for producing the ordered legal word recognition results with reliable decision values. These values can be used for obtaining an optimal word segmentation path of over-segmentation hypotheses as well as an efficient rejection criterion of word recognition result. Simulation was performed with CENPARMI bank check database and shows quite encouraging results.

Recognition of unconstrained handwritten numeral strings using decision value generator

This paper presents recognition of unconstrained handwritten numeral strings using a decision value generator. The numeral string recognition system is composed of three modules: pre-segmentation, segmentation and recognition. The pre-segmentation module classifies a numeral string into sub-images, such as isolated digits, touching digits or broken digits, based on the confidence value of decision value generator. The segmentation module splits the touching digits using the reliability value of decision value generator. Both segmentation-based and segmentation free methods are used in classification and segmentation. To evaluate the proposed method, experiments were conducted using the handwritten numeral strings of NIST SD19 and a higher recognition performance than previous works was obtained.

An Adaptive User Interface in Smart Environment exploiting Semantic Descriptions

In ubiquitous environment, various kinds of devices such as automated appliances and sensing devices exist for providing user with convenient services. However, in order to use such devices with felicity, people should have proper knowledge how to exploit and operate it through heterogeneous interfaces. Then, to enable human user to utilize the devices easily, the integrated interface, employing them with knowledge of them at ad-hoc manner, is necessarily needed. Therefore, to generate the adaptive interface for user in the environment, we propose an adaptive user controller exploiting the representation of devices through semantic descriptions. The devices in the environment are marked up with semantic description which can offer the logic describing information resources with semantics. The controller discovers the service devices, and it constructs the interfaces dynamically using knowledge for effective arrangement and user preference adaptation. We expect the system will dynamically provide user interfaces which make users control environment continuously without spatial restrictions. In addition, it can offer complex services with multiple devices using knowledge and defined rule through simple visualized interface

Post-processing scheme for improving recognition performance of touching handwritten numeral strings

This paper describes a post-processing scheme to improve recognition performance of touching handwritten numeral strings. Verification factors are defined to rectify the recognition results of digits that are segmented by the highest reliability values. Three kinds of verification factors from structural features and recognition probability are used to determine mis-segmented digits. Additional segment digit set is included in valid digit sets when it satisfies the reconsideration condition. A final optimal digit set is selected as the highest ranked segment digit set among all candidate segmented digit sets. Introduction of reconsideration condition and additional segment digit set improves the reliability of the recognition result as well as recognition performance. Experiments were carried out with touching handwritten numeral strings of NIST SD19 database and an encouraging recognition result was obtained.