(This article is sponsored by The Boston Group)

Pen computing as a field broadly includes computers and applications in which a pen is the main input device. This field continues to draw a lot of attention from researchers because there are a number of applications where pen is the most convenient form of input. These include:

1. Preparing a first draft of a document and concentrating on content creation.
2. A socially acceptable form of capturing information in meetings, that is quieter than typing and creates minimal visual barrier.
3. Applications that need privacy.
4. Entering letters in ideographic languages like Chinese and Japanese; and non-letter entries like graphics, music and gestures.
5. Interaction with multi-modal systems.

The advent of electronic tablets in the late 1950s precipitated considerable activity in the area of pen computing. This interest ebbed in the 1970’s, and was renewed in the 1980’s, primarily due to advances in pen hardware technology and improvement in user-interfaces and handwriting recognition algorithms. There are still however, a number of challenges that need to be addressed before pen computing can address the needs listed above to an acceptable level of user satisfaction.

The function of the pen input device is to convert pen tip position into X,Y coordinates at a sufficient temporal and spatial resolution for handwriting recognition and visual presentation. Pen tip contact with paper also in some cases needs to be sensed when ink is deposited on paper. A pen input system consists of a combination of pen, paper, and pad. To digitize handwriting, at least one of these elements must contain electronics.

There are 4 main tracking technologies used in Pen devices: Magnetic, Electronic, Ultrasonic and Optical. Magnetic tracking is the widest deployed system due to high spatial resolution, acceptable temporal resolution, reliability and modest cost. Magnetic and electronic tracking require pad electronics, making them thicker and heavier then a conventional clipboard. Ultrasonic tracking does not require pad electronics, making it lower cost and weight. Relative tracking can give high resolutions, but absolute spatial resolution with ultrasonic tracking is limited due to air currents that cause Doppler shifts. Optical tracking offers the highest spatial and temporal resolution. Optical tracking will play a significant role in future pen systems.

Handwriting recognition is the problem of recognizing electronic ink (trace of the pen tip trajectory) representing a sequence of words, drawing or non-alphanumeric symbol indicating an action to be performed. Most recognition systems use models for handwriting, a lexicon of possible words and linguistic information to guide the recognizer.

Handwriting recognition is the problem of recognizing electronic ink (trace of the pen tip trajectory) representing a sequence of words, drawing or non-alphanumeric symbol indicating an action to be performed. Most recognition systems use models for handwriting, a lexicon of possible words and linguistic information to guide the recognizer.

There are three different factors that determine the complexity of a recognition task:

1. Lexicon: The size of a lexicon can vary from very small (for state names) to open (for proper names). In open vocabulary recognition, any sequence of letters is a plausible recognition result and this is the most difficult scenario for a recognizer.
2. Letter shapes: A tightly constrained system forces users to write each letter in a pre-defined way. This reduces variability considerably and makes the recognition task simpler. A completely unconstrained system allows users to write in their natural handwriting style, which improves usability, but makes the recognition task extremely difficult.
3. Handwriting models: Out-of-the box recognition uses writer-independent models trained from the handwriting of multiple writers. This gives a good average performance across different writing styles. However, there is considerable improvement in recognition accuracy that can be obtained by customizing the handwriting models to a specific writing style. Recognition in this case is called writer-dependent recognition

Typically a writer-dependent, tightly constrained, small lexicon system gives very good recognition results.

|