Fuzzy Logic Meets Pen Computing

Fuzzy logic is a big winner in Japan where it is used to control everything from cameras to cars, from washing machines to subway systems. Likewise, pen-based computing has grabbed the Japanese imagination more so than in the U.S. Consequently, it was only a matter of time until the two technologies found each other.

The Fuzzy Spline Curve Identifier (FSCI), developed by Tecnova Corporation (an image-processing company), is a programmer’s tool for pen-based developers. More specifically, FSCI is a C library for creating freehand drawing programs that efficiently translates “rough” input into “perfect” curves, circles, lines, and the like.

The demonstration system we tried at the recent IEEE International Conference on Fuzzy Systems was DOS-based, using the Wacom tablet and a 386 PC. The demo program, which was indeed very natural and fast, ran in less than 1 Mbyte of RAM (making use of DOS extenders), taking up less than 500 Kbytes of hard disk space. (The company claims that the final product release will require much less space, along the same lines as a device driver.) Although the demo was DOS-based, the original version was written under UNIX; Windows and Macintosh versions are also planned.

The FSCI is based on a technique called “fuzzy spline interpolation” which, according to its developers Sato Saga and Hiromi Makino (both of Tecnova), was first introduced to handle curves which involve vagueness in their positional information. FSCI segments a freehand curve into a sequence which is then categorized in terms of a line, circle, circular arc, ellipse, elliptic arc, closed free curve, or open free curve. In doing so, FSCI distinguishes definite partition points from mere candidates of partition points according to the roughness of drawing at each point.

Saga and Makino have prepared a detailed technical paper entitled “Fuzzy Spline Interpolation and its Application to On-line Freehand Curve Identification.” Contact them at Tecnova Corp., Techno Park 2-3-3, Shimonopporo, Atsubetsu-ku, Sapporo, 004 Japan. Tel: +81-11-807-6881.

Similarly, two researchers at the Hong Kong Polytechnic Institute have applied fuzzy logic to the problem of handwritten character recognition. The approach take by Man Gary and Poon Joe is to use a thinning algorithm to produce skeletons from raster scanned images which contain characters to be recognized. Based on this provision of skeletons, each character is decomposed into a number of connected strokes. Each stroke is then classified as a primitive according to a fuzzy similarity criterion. When the structural information is extracted, a fuzzy-attributed graph can be constructed to represent the character. Gary and Poon’s technique includes a new means of measuring the similarity between two fuzzy-attributed graphs of a known character class and an unknown character. When this similarity measure is applied in the recognition stage, an input character is correctly classified. Initial experiments shows an accuracy rate of about 93 percent.

Gary and Poon presented a technical description of the algorithm in a paper entitled “A Fuzzy-attributed Graph Approach to Handwritten Character Recognition” at the IEEE conference.

By the way, if you’re not familiar with fuzzy logic, a good discussion of what it is, its history and how the Japanese are using it is presented in the recently published Fuzzy Logic: The Discovery of a Revolutionary Computer Technology — and How it is Changing Our World by Paul Freiberger and Daniel McNeill, Simon & Schuster, 1993.