Journal of Electronic Library, volume 21, number 6, p. 575-586, 2003

ABSTRACT

The right of blind people to access the Internet is simply ignored in many countries because Web pages have been designed for normal people. As a result, many blind people are not enjoying the benefits of the Internet and the improvement in the quality of life that Internet use can bring. In order for visually impaired persons to surf the Internet, it is necessary to develop a special human-computer interface system. This paper presents the design of a Web project for the blind. The aim of this research is to develop a new human-computer interface model and an associated computer system for visually impaired people so that they can browse the World Wide Web via Internet. An assessment of the potential of a wide range of applications and their impact are also presented.

Keywords: Human Computer Interface, Internet Surfing, Blind

INTRODUCTION

It is estimated that there are 54 million people in the United States with a disability. The Congress of United States enacted the Americans with Disabilities Act (ADA) in 1990 and passed amendments in subsequent years that “prohibit discrimination on the basis of disability in employment, programs and services provided by state and local governments, goods and services provided by private companies, and in commercial facilities.”(http://www.usdoj.gov/crt/ada/publicat.htm). Web sites and pages are also covered under the ADA. The US Access Board also issues standards (Access Board, 2000) for electronic and information technology covered by section 508 of the Rehabilitation Act Amendments of 1998. However, many visually impaired people today still have access problems with most Web sites (Cynthia et al, 1999). The reason for this phenomenon is simple - many Web page designers do not test the accessibility of their designs with disabled persons in mind. The accessibility problem has grown significantly (Marquand, 2000) because more business and government agencies are relying on the Internet to disperse information and services.

As it is difficult to define accessibility, the World Wide Web Consortium (W3C) has outlined an accessibility guideline document in its website (W3C, 1999) to help web designers. Although this document is quite bulky (34 pages), the idea is quite straight forward. If information is conveyed through color, sound, or image, an alternative description should be placed in the html file. The alternative description can then be read by a “Screen Reader” for people with disabilities . Row and column headings should be used to give direction to users if tables are used in the web pages.

This document recommends 14 guidelines and 105 checkpoints. These checkpoints are classified into three priority levels. Conformance Level “A” will be awarded to web pages which satisfy all priority 1 checkpoints. Conformance Level “Double-A” will be given to web pages which satisfy all priority 1 and 2 checkpoints. Conformance Level “Triple-A” is the highest level. A web page must satisfy all priority checkpoints in order to be awarded “Triple-A” conformance.

It is quite time consuming to validate all 105 checkpoints for each web page. Automatic validation tools do exist and are generally fast and convenient, but they cannot validate all accessibility issues. Human review is still required to ensure a web pages’ conformance. The World Wide Web Consortium thus recommends both automatic validation and human review.. Among the available automatic validation tools, Bobby (http://www.cast.org/bobby) is one of most well known software packages.

Bobby was developed by a non-profit organization called the Center for Applied Special Technology (CAST). Users can submit a web page to Bobby by typing the URL of the page at CAST’s web site. Bobby can then examine the page and report accessibility problems. This method will only check one page at a time in order to keep the server available to all. A downloadable version of Bobby, which can check web pages in a whole web site in batch mode, is also available. A web designer earns the right to display a Bobby icon on his/her web page if it passes the Bobby test.

Even with the protection of ADA and availability of automatic tools, recent accessibility studies (Jackon-Sanborn et al, 2001) using Bobby show that the majority of U.S.-basedweb sites do not meet the Web Content Accessibility Guideline in http://www.w3.org/WAI/GL (Waddell 1999). In many developing countries, the lack of access to the Internet for disabled persons is even worse. People with disabilities in these countries are not protected by laws similar to ADA in the U.S. The access problem is simply ignored by web designers as they do not believe that they should make the Web sites accessible to people with disabilities. It is even more urgent to find a solution in these countries.

Many governments realize the importance of Internet and the benefits it can bring to their populations. These governments have invested heavily to promote the use of the Internet. However, blind persons cannot receive benefits from these investments as they cannot see the information presented via Internet on a computer screen. It is practically impossible for them to use the Internet as they cannot position the cursor to a particular location on the screen using the mouse.

This paper presents a new human computer interface designed¹ to solve Internet accessibility problems encountered by blind people. English “screen reader” programs for blind people are already on the market, but an efficient Chinese-language screen reader for Web browsing is not available yet. Hong Kong-based Web sites routinely contain both Chinese and English characters on the same Web page, making a screen reader for this market more difficult to develop. This new human computer interface can deal with mixed language content and can be used by very young and very old segments of the population as well as by visually impaired people.

BACKGROUND INFORMATION

The Internet is the most well-known component (Kalakota and Whinston, 1996) of the Information Superhighway network infrastructure which spans several continents and is the backbone of electronic commerce. Indeed, Internet use is expanding faster than any other communication technology in history and has the potential to significantly impact the major portion of the population in any society. The Internet’s ability to transmit multimedia content overcoming time and space constraints has created exciting and unforeseen opportunities in commerce, communication, education, science, politics, international relations, and many other fields. The Internet has played a major role in stimulating the global economy and has a profound impact on the quality of life for its users. However, a digital divide exists. People with disabilities are often left out of this Internet revolution.

In the early stages of Internet, only text information was available on the Internet. The text LineMode Browser (Walsh 1996) in 1991 was quite different from the Web navigation tools we know today. It did not support the mouse or graphics and was difficult to use. The first multimedia browser (MOSAIC) to boast a user-friendly graphical user interface (GUI) was released in 1993. MOSAIC was considered to be a breakthrough software product as it advanced the World Wide Web into a multimedia system. Today, the Internet can deliver text, video, sound, human speech and graphics. The mouse and hypermedia are employed to make it easy to navigate the Internet and search for information. The latest Web browsers, Netscape Navigator and Microsoft Internet Explorer, have further improved on the functions of MOSAIC and use similar technologies. Although the mouse and hypermedia are great interface tools for "normal" people, ironically they create barrier for visually impaired persons to access the Internet. Today, there are many new applications for the Internet such as Internet banking, Internet shopping, Internet voting, Internet telephone and Internet television. Internet is also being used for education and for seeking employment. However, visually impaired persons are not able to obtain the full benefits of Internet because it is nearly impossible for them to navigate the Internet with the existing browsers. Thus, new human-computer interfaces need to be developed to enable them to enjoy the benefits of the Internet.

Human-Computer Interface

Research in human-computer interfaces (HCI), an important area of software design, has been very active and references abound. However, most research has been based on the assumption that the user possesses normal eyesight. Research work on access tools for blind people is lacking. For example, visual design is often stressed in the design of human-computer interfaces with the objective of providing visual attributes that contribute valuable impressions and communicate important cues to a user. Various approaches have been suggested, and technologies developed, via which visually impaired persons can access the Internet and surf the Web. These approaches and their limitations are presented in the following sections.

Text Browsers

To avoid problems of using the mouse and hypermedia, most visually impaired persons use text-based Web browsers (e.g. Lynx) that will ignore graphics on Web pages and allows the use of the keyboard to activate hyperlinks. However, since many Web designers only test their designs on popular browsers such as Netscape and Microsoft Explorer, they often use features that are not supported by text browsers; blind users often have problems accessing such web sites. Text browsers cannot completely solve the problems of Internet surfing for the blind.

Screen Readers

Speech synthesis technology (Allen et al, 1981; Suen, 1981) has been available since the late 1970s. “Screen readers” (Blenkhorn and Caulderwood, 1992) were developed in 1980s and blind people can now access most text-based computer displays using speech generated by screen readers (Meyers and Schreier, 1991). However, simply reading the text and converting it to human speech will not solve Internet navigation problems for blind people. First of all, screen reading is usually done in a batch mode. A real time mode is required for Internet navigation. In addition, "reading aloud" every item on a Web page and asking the user to make subsequent choices constitutes a heavy burden on a humans’ short term memory (Zetie, 1995) making it a poor HCI technique. Also, most “text reading” programs work independently and cannot interact with popular Web browsers such as I.E. and Netscape. In order to activate the next Web page, the user still needs to point to a specific hypertext link and click the mouse, an action which is nearly impossible for a visually impaired person. Innovative methods must be developed if visually impaired people are to have uninhibited access to the Internet.

Braille Printout and Braille Devices

Thirty years ago, the output of computer systems was primarily conveyed to humans via paper printout. As blind computer users cannot read ordinary paper, they had to read computer output by touching paper specially indented with a pattern of raised dots called “Braille” (Lightowler, 1994; Blenkhorn and Evans, 1988). This technology was named after its inventor - Mr. Louis Braille. He was a blind Frenchman and his blindness was caused by an accident in his childhood. Braille is not the only reading and writing system for the blind, but it was considered to be the best according to several independent studies (Keeler, 1986). Through out the years, his system has been adopted by many countries all over world. Over 600,000 books, newspapers and magazines are printed in Braille every year. However, it is much more expensive than ordinary computer printout and a special printer is required.

A Braille device (Kay, 1984; Leventhal et al, 1991; CSUN’95, 1995) is another alternate output device for the blind. A small part of the image of a computer screen can be generated on the device; a visually impaired person can read it quickly by touching the device and does not have to wait for the generation of the Braille paper. However, Braille devices are very expensive. A typical device costs about US$6,000 while the cost of a Pentium-based computer is only US$1,000. People with disabilities generally have far lower incomes than other citizens (National Council on Disabilities, 2001). Most visually impaired computer users cannot afford to buy a Braille device. A cheaper and more reliable output method for the blind is necessary. The speech option meets these criteria and thus it is chosen as the major navigation method for this project.

ADVANTAGES OF THE KNOWLEDGE BASED APPROACH

The knowledge based screen reader system provides many advantages for the system as it can be extended by adding to/ replacing its knowledge base for a variety of applications. For example, the interface system can be modified so that visually impaired persons will be able to use other popular programs such as Microsoft Word, Excel, etc. by changing the knowledge base of the resident program.

VOCALSURF: A INTERNET SURFING TOOL FOR BLINDS

A HCI system especially designed for Internet surfing by visually impaired people was developed as part of a project funded by the Quality Education Fund (QEF) of Hong Kong. The key objective of the project was to produce a prototype to assist blind people to understand the contents of Web pages through speech and, using simple keyboard instructions, to interact with the various components of a Web page. The design of the prototype and its components are presented below.

Design of the Prototype

As the Overview of the System (Figure 1) shows, the basis of the system is that resident programs read HTML pages downloaded via Web browser, and with the help of dictionary files and knowledge bases. produce human speech. The human speech is used by visually impaired persons to guide their interaction with the browser. They in turn can provide their input through the use of special input device or a regular keyboard that generates emulated mouse.

Figure 1. Overview of the new system

Specifically, the resident programs have the following functions:

• 
  Interaction with the Internet browser
  
• 
  Selectively reading part of the text in Web pages and producing human speech;
  
• 
  Receiving signals from special input unit and emulating a corresponding mouse signal to the browser.
  

Components of the system

Figure 2. Components of the resident programs

As described in Figure2, the resident programs of the HCI system consist of the following modules:

• Input Handler

This module accepts input signals from the Input Device and passes the signals to the user interface unit.

• User Interface

This module takes input messages from the Input Handler module and interprets the signals with the help of an Inference Engine. It then sends the signal to the Mouse Emulator.

• Inference Engine

The Inference Engine gets rules from the Internet Browser and HTML knowledge bases . It matches an input signal against the corresponding “mouse click” if action from the browser is necessary. It also selects sentences/words and pass them to the “voice synthesizer” module.

• Voice Synthesizer

The voice synthesizer generates human speech by matching selected words/sentences on the Web page with those in dictionaries and wave tables.

• 
  Mouse Emulator
  

The Mouse Emulator module emulates a corresponding mouse signal and passes it to the Web server.

Development of the prototype

The prototype of this project was called VocalSurf. It was designed to operate on any Internet-ready personal computer using Microsoft Windows 95/98 as a single application program after installation. The hard disk capacity required is 128 megabytes (MB). Users interacted with VocalSurf using speech and keyboard. Users typed in simple instructions and VocalSurf read back specified Web page content to the users. In other words, VocalSurf was designed as a WWW surfing tool for blind or visually impaired individuals.

Technologies Applied

To make VocalSurf functional, the following technologies were employed, in addition to object oriented programming techniques:

• 
  Microsoft Sound Application Programming Interface (SAPI);
  
• 
  Sound Wave Manipulation;
  
• 
  Component Object Modelling (COM).
  

Microsoft SAPI technology was the core technology applied in VocalSurf’s English speech engine construction. In constructing the Cantonese speech component of VocalSurf, since SAPI for Cantonese is not available from Microsoft, sound wave manipulation using audio compression techniques and COM technologies were adopted to simulate a SAPI for a Cantonese speech engine. Rapid Application Development was adopted in software development to facilitate continuous prototyping.

Mechanisms Implemented

The following diagram illustrated the overall architecture of the sound engine:

Figure 3: Sound Engine

End-users interact with VocalSurf by means of User Interface using the keyboard and control keys are summarized in Table 1. Messages are then carried forward to the VocalSurf Sound Engine, which parses the requested Web page for meaningful content.

Key Combination	Function
CTRL	Focus on URL input
SHIFT B	Begin reading
SHIFT S	Stop reading
SHIFT L	List the current 10 hyperlinks
SHIFT =	Move on to the next 10 hyperlinks
SHIFT -	Return to the previous 10 hyperlinks
SHIFT 0...9	Select a particular hyperlink in the current 10 hyperlink listing. (If the current 10 hyperlink listing is from 11 to 20, 0 will be 20, 1 will be 11, 2 will be 12, etc.
Backspace	Go back
ALT 	Go to a page ahead of the current page

Table 1: Control keys for the System

The engine also determines if the reading content is Chinese or English. English content is directed to an API Wrapper for SAPI to process. If the content is Chinese, every word will be matched against a database for the corresponding wave compressed files. When processing by either Database-WAV or SAPI is complete, the VocalSurf Sound Engine produces the audio output.

Figure 4: Classes in the Sound Engine

Classes and objects in the sound engine are described in Figure 4. The most important class in Figure 4 is the “PlaySound” which produces human voice. Its components are described in Figure 5.