Catálogo de publicaciones - libros

A lot of printed documents and books has been published and saved as a form of images in digital libraries. Searching for a specified query word on document images is a challenging problem. The OCR software helps the images to be converted to the machine readable documents to search a full context [1]. Another approach [1, 2] is image-based one, in which both the document images and word information are saved in a database. The searching procedure is accomplished through comparing the features of query word image with the word images extracted from document images in the database. In this paper, we propose an accurate and fast keyword spotting system for searching user-specified keyword in Hangul document images by a two-level image-to-image matching method.

This paper proposes a character segmentation system for Korean printed postal images. The proposed method is composed of two main processes, which are robust skew correction and character segmentation. Experimental results on real postal images show that the proposed system effectively segments characters to be suitable for the input of OCR system.

The question of detecting pertinent information about an individual from properties of their speech is not a new one. Much research has been done to explore the manifestation of emotions in voice. The work presented in this paper proposes to apply these efforts in the domain of depression assessment using Case Based Reasoning. Cases are constructed using the recordings of responses to a questionnaire from English speaking Males and Females, and Spanish speaking Males and Females. We apply the exemplar and instance approach to classify new test cases. Experimental results show that the construction of cases using sound waveform statistics can be utilized by a case based reasoner to classify new instances correctly.

This paper presents a system for automatically generating linguistic questions based on a feature table. Such questions are an essential input for tree-based state tying, a technique which is widely used in speech recognition. In general, in order to utilize this technique, linguistic (or more accurately phonetic) questions have to be carefully defined. This may be extremely time consuming and require a considerable amount of resources. The system proposed in this paper provides a more elegant and efficient way to generate a set of questions from a simple feature table of the type employed in phonetic studies.

In speech recognition huge hypothesis spaces are generated. To overcome this problem dynamic programming can be used. In this paper we examine ways of speeding up this search process even more using heuristic search methods, multi-pass search and aggregation operators. The tests showed that these techniques can be applied together, and their combination could significantly speed up the recognition process. The run-times we obtained were 22 times faster than the basic dynamic search method, and 8 times faster than the multi-stack decoding method.

This paper presents a clear-cut definition of consciousness of humans, consciousness of self in particular. The definition “Consistency of cognition and behavior generates consciousness” explains almost all conscious behaviors of humans. A “consciousness system” was conceived based on this definition and actually constructed with recurrent neural networks. We succeeded in implementing imitation behavior, which we believe is closely related to consciousness, by applying the consciousness system to a robot.

Prediction is an important task in robot motor control where it is used to gain feedback for a controller. With such a self-generated feedback, which is available before sensor readings from an environment can be processed, a controller can be stabilized and thus the performance of a moving robot in a real-world environment is improved. So far, only experiments with artificially generated data have shown good results. In a sequence of experiments we evaluate whether a liquid state machine in combination with a supervised learning algorithm can be used to predict ball trajectories with input data coming from a video camera mounted on a robot participating in the RoboCup. This pre-processed video data is fed into a recurrent spiking neural network. Connections to some output neurons are trained by linear regression to predict the position of a ball in various time steps ahead. Our results support the idea that learning with a liquid state machine can be applied not only to designed data but also to real, noisy data.

This paper developed a robot competition system using a gesture based interface. The used interface recognizes a gesture as meaningful movements from a fixed camera and controls a robot by transforming gesture commands. In the experiment, the used robot is robot and the experimental results verify the feasibility and validity of the proposed system.

This paper presents an agent system ASGARD-0, that provides monitoring for the success or failure of Grid jobs in a High Energy Physics application. This application area is one where use of the Grid is extremely well motivated as processes are both data and computationally intensive. Currently however there is no mechanism for automated monitoring of jobs and physicists must manually check to see whether the job has completed and whether it has done so in a successful manner. ASGARD-0 provides some initial services in this area and is also a proof of concept for a much more ambitious agent support system.

Trust is an important issue in trade. For instance in food trade, market actors have to rely on their trade partner’s quality statements. The roles of trust and deception in supply networks in various cultural and organisational settings are subject of research in the social sciences. The Trust And Tracing game is an instrument for that type of study. It is a game for human players. Conducting experiments is time-consuming and expensive. Furthermore, it is hard to formulate hypotheses and to test effects of parameter changes, as this requires many participants. For these reasons the project reported in this paper investigated the feasibility of multi-agent simulation of the game and delivered a prototype. This paper briefly describes the game and introduces the process composition of the agents. The prototype uses simple, but effective models. The paper concludes with directions for refinement of models for agent behaviour.