ERCIM News No.33 - April 1998

Information Technology for Education of Visually Impaired Students

by Ivan Kopecek and Karel Pala

Information technology is nowadays the most important instrument used to support visually impaired people. In addition, some of the blind impaired take it as a chance to break the information barrier and they try to apply this technology for solving some of their problems. Exploitation of information technologies for education of visually impaired students is thus important and provoking problem. For this purpose the special speech oriented hypertext system Audis is being developed at the Faculty of Informatics, Masaryk University Brno with the aim to support the education of visually impaired students. The project is a part of the Grant VS97028 funded by the Czech Ministry of Education.

AUDIS is developed to provide a comfortable access for visually impaired students to their textbooks. It maximally pays attention to the special needs of users and allows for flexible customization. The system is supported by the utilities for text conversion and rendering. Next versions of AUDIS should allow the system to be used as an WWW browser. The system is controlled by speech commands combined with keyboard key commands. This access optimizes the effectiveness of the control by joining advantages of both methods. Every command exists in both versions, ie as a speech command version and keyboard key version. This is due to possible application of the system for print impaired people. Speech commands consist of the key words and are recognized by built-in speech recognizer.

The system enables customization of the commands by renaming key words or by defined alternatives. To use the basic speech commands the recognizer need not be trained, it can be automatically customized to the user's voice. However, some simple training of the recognizer increases reliability of the speech recognizer when larger number of speech commands is used. Sound is the main output of the system. It can be used in the form of synthesized voice produced by the syllable based speech synthesizer DEMOSTHENES which is able to read hypertext data, to produce sampled voice necessary for all standard messages to the user, and non-speech sound that is used in first place for environmental sounds applied to provide feedback to user's actions, and secondly for earcons (non-speech glances used to give the blind user an overview by listening).

The used speech synthesizer applies basic prosodic features based on corpora analysis and in this way makes it possible to use various voice types which can be configured by the user. Various voice types are used to distinguish various types of information (eg normal text, underlined text, links etc).

AUDIS data should be carefully prepared. Although a simple structure based on the defined subset of HTML is acceptable for the system, other text should be added to the basic text and some additional pieces of information should be included to make the data more comfortable for the visually impaired users. Pieces of information that are not convenient for the sight impaired (graphics, mathematical expressions, tables etc) should be reduced and/or explained, if possible, in textual form. Shape of the text should be modified with respect to the user's needs. Graphics, tables, mathematical expressions and other non-textual information have to be rendered and/or interpreted and the sentences containing important information must be phonetically accentuated. In accordance with AUDIS hypertext data structure conventions also summaries have to be added.

Building the tree structure of the system from basic linear structure is supported by the automatic generation of the optimal tree structure and some other pieces of information are generated automatically as well (eg earcons, sound glances, environmental sounds).

Please contact:

Ivan Kopecek, Karel Pala - CRCIM - Masaryk University
Tel: +420 5 4151 2364
E-mail: {kopecek, pala}

return to the contents page