Stroke-Based Intelligent Word Recognition Using a Formal Language

We have developed a method that makes it easier for language beginners to look up Japanese kanji characters. Instead of using the arbitrary conventions of kanjis, this method is based on three simple prototypes: horizontal, vertical, and other strokes. For example, the code for the kanji \(\boxplus\) (ta, meaning rice field) is ‘3-3-0’, indicating the kanji consists of three horizontal strokes and three vertical strokes. Such codes allow a beginner to look up kanjis even with no knowledge of the ideographic conventions used by native speakers. To make the search easier, a complex kanji can be looked up via the components making up the kanji. We conducted a user evaluation of this system and found that non-native speakers could look up kanjis more quickly and reliably, and with fewer failures, with our system than with conventional methods.

In this paper, a bilingual script recognition system is developed to identify and separate out texts written in Devnagari and Gurumukhi scripts. It is observed that vertical half strokes and horizontal strokes are predominantly used in Gurumukhi script compared to Devnagari. On the basis of this observation, we develop a morphological image processing based scheme to extract vertical and horizontal strokes in a document image. The two script regions are subsequently separated out on the basis of the density of these strokes in different regions of the document. Experimental results demonstrate the effectiveness of our proposed scheme. However, the proposed scheme works well only with machine printed documents.

This work proposes a new stroke based methodology for handwritten character recognition. After the pre-processing, several steps are involved to achieve the recognition. First, the character is segmented into its strokes. Then, we determine the maximum length of the longest horizontal segment that can be inscribed on a stroke. We also compute that for the vertical direction. So we decide whether the stroke must be tagged as horizontal or vertical. After that, we represent by a string the vertical stroke position and its relational-ship with its adjacent horizontal strokes. In that string, each vertical stroke is represented by a character followed by a set of numbers which means where the adjacent horizontal strokes join the vertical one. A formal language grammar has been set and a knowledge base developed with known characters written by a single writer. Finally, an inference engine allow us to recognize unknown characters written by that single user. This algorithm has been tested on different writers and provides a hit rate of 87,13%.

A trial approach is presented for the realization of an optical character reader for printed Chinese characters (kanji). This approach is different from conventional ones as far as kanji features and recognition are concerned. Use is made of structural features based on only the horizontal and vertical strokes in a kanji, taking advantage of optical/digital hybrid processing. The recognition experimental results are promising: 99.6%, 0.3%, and 0.1% for the correct, rejected, and error rates, respectively.< <ETX xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">&gt;</ETX>

We developed a method to enter Hiragana (cursive in Japanese) characters and alphanumeric characters by thumb gestures for Japanese people. This method is optimized to be used while holding the left and right ends of a tablet PC with hands. Each gesture is designed by concatenating a few horizontal and/or vertical strokes, but path of the gesture remains inside the one-stroke square centering its starting position. A Hiragana character is input by selecting a row of the table of Japanese syllabary and then a column of the table. The alphabet characters are segmented into 6 groups in order of letters. Just like Hiragana, an alphabet character is entered by selecting a group and then selecting a member. However, in the alphanumeric mode, to reduce the cost of modifying the mistakenly selected group, an operation to move to the previous or next group have been added.

Recently, some models of neural networks, recurrent neural networks, have been used in conjunction with their associated neural learning schemes to infer regular grammars from a set of sample strings. The representation of the inferred automata is hidden in the weights and connections of the net, this being a common feature in emergent subsymbolic representations. In order to relate the symbolic and connectionist approaches to the tasks of grammatical inference and recognition, we address and solve a basic problem, which is, how to build a neural net recognizer for a given regular language specified by a deterministic finite-state automaton. A second-order recurrent network model is employed, which allows to formulate the problem as one of solving a linear system of equations. These equations directly represent the automaton transitions in terms of static linear approximations of the network running equations, and can be viewed as constraints to be satisfied by the network weights. A description is given both for the weight computation step and the string recognition procedure.

In this paper, the similarity between a pair of text documents is measured by identifying three sets of words: words that exist in the first but not in the second, in the second but not in the first and those words that exist in both. Two deterministic finite automata are built to recognize the words of the two documents. These finite state automata are used to identify these three sets of words. An experimental study was conducted to present the proposed approach and to compare it with other peer text similarity measures: Jaccard and cosine distance. The proposed approach gave more reasonable results than Jaccard but more inflated than cosine distance approach.

The recognition rate of OCR (optical character recognition) systems often decreases for degraded text images, such as stroke-connected, stroke-broken and noisy text images. In this paper a unified method for preprocessing binary text images based on mathematical morphology is presented. Simple but effective morphological filters are designed to process stroke-connected and stroke-broken images separately. A method to classify stroke-connected and stroke-broken text images based on the horizontal and vertical average stroke width is also presented. A morphological contour filter is used for removing noise. Experimental results show that the result of the preprocessing method are satisfactory.

Characterization of the fine motor problems in patients with cognitive dysfunction – A computerized handwriting analysis

On-line size control in handwriting demonstrates the continuous nature of motor programs

Total alloplastic temporomandibular joint (TMJ) replacement has become common. This in vitro study aimed to evaluate wear response of custom-fabricated electron beam melted titanium (EBM-Ti), zirconia, and acrylic TMJ parts when subjected to biomechanical simulation testing. Eighteen prosthetic TMJ parts (condyle, glenoid fossa) were custom-fabricated using computer aided design and manufacturing (CAD/CAM) techniques based on patient’s radiographic images. Biomechanical simulation testing of TMJ parts (in different combinations) were done in a modified chewing simulator (108,000 cycles, 1 Hz frequency, 45–60 N compression, strokes-downward 0.15–0.25 s/horizontal, 0.4–0.5 s/upward, 0.25–0.45 s/displacement, 1.5–2.0 mm). Qualitative analysis using scanning electron microscopy revealed wear facets on leading edges of vertical and horizontal simulation strokes. Measurement of pre-test and post-test weights of TMJ parts revealed non-significant reduction in weights due to wear. EBM-Ti and acrylic TMJ glenoid fossae articulating against zirconia condyles during simulation testing had significantly higher wear, evidenced by greater mean reduction in weights. Based on results of this preliminary study, custom-fabricated alloplastic prosthetic TMJ are a viable alternative to stock alloplastic joints. While EBM-Ti and acrylic are suitable biomaterials for custom-fabrication, use of zirconia results in greater wear and requires further studies to optimize their role in customized alloplastic TMJ.

During the first and second centuries A. D., Roman cursive developed a system of ligatures between letters that was to remain virtually unchallenged until a new cursive structure came about in the twelfth and thirteenth centuries. This remarkably coherent system was based on top-down connections, in the form of angular joins between the last, horizontal stroke of one letter and the first, vertical and downward stroke of the next; but from the outset different combinations, although sporadic, even exceptional, are evidence that other solutions were technically possible, namely bottom-up connections, particularly between vertical strokes. Why, then, was a broad range of artifices developed to enable the use of top-down, angular connections even with letters whose «natural» forms were incompatible, instead of resorting more widely to upstrokes? Mainly because this type of ligature was quite similar to the manner in which scribes tended to arrange the strokes constituting letters; the ligatures were consistent with a certain organisation of handwriting, which they in turn conditioned. This is all the more apparent in the fourth century when the structure of cursive letter forms and ligatures both developed in the same direction: although bottom-up connections remained relatively scarce and specific to certain variant letter forms (quite different in nature from medieval cursive ligatures), they henceforth became part of the system.

Deterministic Finite-Automata Abstractions of Time-Variant Sequential Behaviours

Expert system knowledge represents expertise obtained through formal education, training, and/or experience. Formal education provides deep knowledge of a particular domain; experience and training result in heuristic knowledge. A knowledge base defines the range of information and understanding with which the system is capable of dealing; therefore, its information must be structured and filed for ready access. The objective of this symposium is to address the challenges associated with establishment of valid expert system knowledge, specifically, knowledge to be used by expert system shells. As expert system knowledge is obtained, structured, and stored, it is formulated. In this symposium, knowledge formulation is addressed as a three-phase process: knowledge acquisition, the mechanics associated with structuring knowledge, and knowledge porting. Knowledge acquisition is the process of extracting expertise from a domain expert. Expertise may be collected through a series of interviews between the expert and a knowledge engineer or through sessions the expert holds with an automated knowledge acquisition tool. Thus, the ultimate outcome of knowledge acquisition is a collection of raw knowledge data. The following human factors issues become apparent: documenting mental models (where mental models are the expert's conceptualization of a problem), recording cognitive problem-solving strategies, and specifying an appropriate interface between the domain expert and the acquisition methodology. The knowledge structuring process involves the refinement of raw knowledge data, where knowledge is organized and assigned a semantic structure. One issue that must be considered is how to interpret knowledge data such that formal definitions, logical relationships, and facts can be established. Finally, formulation involves knowledge porting, that is, the movement of an expert system shell's knowledge base to various other shells. The outcome of this process is a portable knowledge base, where the challenges lie in maintaining consistent knowledge, understanding the constraints inherent to a shell (the shell's ability to incorporate all relevant knowledge), and designing an acceptable user-expert system interface. The fundamental component of any expert system is its knowledge base. The issues to be presented in this symposium are important because they address three processes that are critical to the development of a knowledge base. In addition to presenting computer science challenges, knowledge base formulation also presents human factors challenges, for example, understanding cognitive problem-solving processes, representing uncertain information, and defining human-expert system interface problems. This symposium will provide a forum for discussion of both types of challenges.

In this paper we analyse the explanation facilities of several expert systems using a framework based on factors relevant to the structure and content of a generated explanation. We compare this with the lack of such facilities in expert system shells, and lay out criteria for an explanation facility within an expert system shell for education. We then describe their realisation in the explores (explanation oriented expert system) shell, where domain-independent templates allow questions to be asked about the domain of any knowledge base being consulted. In response, the shell generates domain-based explanations, with further information available through follow-up questions related to the original question, the generated explanation, and the available knowledge. This facility allows a user to explore and learn about a knowledge domain, and has implications for the use of expert system shells in education.