The development of computers capable of handling complex objects requires nonverbal interfaces that can bidirectionally mediate nonverbal communication including the gestures of both people and computers. Nonverbal expressions are poweful media for enriching and facilitating humancomputer interaction when used as interface languages. Four gestural modes are appropriate for human-computer interaction: the sign, indication, illustration and manipulation modes. All these modes can be conveyed by a generalized gesture interface that has specific processors for each mode. The basic component of the generalized gesture interface, a gesture dictionary, is proposed. The dictionary can accept sign and indicating gestures in which postures or body shapes are significant, pass their meaning to a computer and display gestures from the computer. For this purpose it converts body shapes into gestural codes by means of two code systems and, moreover, it performs bidirectional conversions of several gesture representations. This dictionary is applied to the translation of Japanese into sign language; it displays an actor who speaks the given Japanese sentences by gesture of sign words and finger alphabets. The performance of this application confirms the adequacy and usefulness of the gesture dictionary.

This paper concerns a subclass of regular languages, called strictly regular languages, and studies the problem of identifying the class of strictly regular languages in the limit from positive data. We show that the class of strictly regular languages (SRLs) is polynomial time identifiable in the limit from positive data. That is, there is an algorithm that, for any strictly regular language L, identifies a finite automaton accepting L, called a strictly deterministic finite automaton (SDFA) in the limit from positive data, satisfying the property that the time for updating a conjecture is bounded by O(mN2), where m is the cardinality of the alphabet for L and N is the sum of lengths of all positive data provided. This is in contrast with the fact that the class of regular languages is not identifiable in the limit from positive data.

A rigorous theoretical method for predicting "ratio of desired signal power to interference power [c/i]" and "ratio of signal power to noise plus interference power [c/(n+i)]" where both desired and interference signals vary with time under the Nakagami-Rice fading conditions is presented. An alternative simple prediction method which is more desirable from the viewpoint of engineering application is then proposed. Prediction errors given by the simple method are evaluated by comparing to the errors given by the rigorous method, and it is confirmed that the simple method gives reasonable accuracy. This method is expected to serve in the development of frequency re-use technologies and the coordination of various systems for mobile satellite communications in the near future.