A Comprehensive Review
by Richard Ring
From the Editor: Richard Ring is the Director of the International Braille and Technology Center for the Blind at the National Center for the Blind. The Omni 1000 is Dr. Raymond Kurzweil's latest entry in the reading-machine market. Here is what Mr. Ring has to say about it:
In November, 1996, Kurzweil Educational Systems, Inc. of Waltham, Massachusetts, released Omni 1000, a software-reading system for blind persons. The company simultaneously released the Omni 1000 stand-alone reading machine. Recently the International Braille and Technology Center for the Blind acquired the latest release of this software, which is version 2.0. This article discusses this software at some length. The reason that we are devoting an entire article to this software is that we believe that the Omni 1000 is currently the most feature-rich and powerful OCR package for blind people on the market today. Keep in mind, however, that this market is always changing, and it is quite possible that the Omni 1000 will be surpassed by one of its competitors in the months or years ahead.
What is a PC-based reading system? A reading system consists of four components: a scanner to acquire an electronic image of the page; software to analyze that image and convert it into text, a microprocessor (computer) to run that software; and finally an output device, either a speech synthesizer or a refreshable Braille display, that will allow you to listen to or read the text the system produces.
Why choose a PC-based reading system? One question we are often asked at the International Braille and Technology Center is, "Why not purchase a stand-alone reading machine instead of a PC-based reading system if all you intend to do is read and save documents?" This is a fair question. We believe that a PC-based reading system offers better value and increased versatility. If you have questions concerning this assessment, please feel free to contact the International Braille and Technology Center for the Blind. All this now said, let us turn to the Omni 1000 from Kurzweil Educational Systems.
Kurzweil Educational Systems recommends that you have the following minimum system requirements: a Pentium processor with a clock speed of at least 100 megahertz, a hard disk with storage capacity of at least 500 megabytes (five hundred million bytes), and at least twenty-four megabytes of Random Access Memory (RAM). If you do not choose to use the voice-recognition feature, the program will run on a system with only sixteen megabytes of RAM. We believe, however, that the software will run more efficiently with thirty-two megabytes of RAM. You will need a sound card and speakers or headphones as well as a CD-ROM drive. The Omni 1000 software comes on a CD-ROM disk and requires the Windows 95® operating system. It will also run on a Windows NT® machine. And, since we are talking about a reading system, you'll also need a scanner. Kurzweil has tested many of the Hewlett Packard (HP) scanners, e.g., the 2P, 2C, 3P, 4P, 4C, and 5P. Kurzweil says the Omni 1000 should be capable of supporting most scanners that use the TWAIN® scanner interface, which allows Windows 95® applications to communicate with your scanner. If you are uncertain whether your scanner is supported, you should contact Kurzweil Educational Systems.
What's in the Package?
When you receive the Omni 1000 software, you will get the CD-ROM containing the program, a print manual, a seventeen-key keypad, and a registration card. A Braille manual is also available and can be requested when ordering the software.
Installing the Omni 1000 software is almost too easy. Many Windows 95® applications install with little or no user intervention, which can be frustrating to a computer user who wishes to understand what is happening during the process. Omni 1000 is even more automatic than most Windows 95® applications. Simply place the CD in your drive, put the key disk in drive A, close both drive doors, and the process begins. Speech will occur automatically because along with the OCR program itself, Omni 1000 will install AT&T Flextalk®, a software speech synthesizer that will generate speech through your sound card. And of course the first speech you will hear is Omni 1000 reading the license agreement, which you must accept. The only other time that user-intervention is requested is when the software asks you if you would like Omni 1000 to start each time the computer is turned on. To say no, simply press enter after hearing the question. If you do nothing, Omni 1000 will run immediately after Windows 95® is loaded.
Though Omni 1000 comes with the AT&T Flextalk®, you may wish to continue to use the hardware synthesizer you already own. Omni 1000 supports several SSIL-compliant synthesizers, e.g., DECtalk PC or Express, Keynote Gold, Doubletalk, Accent PC or SA, and the Braille 'n Speak product line. One problem with the nearly automatic installation procedure is that you cannot choose to use one of these supported synthesizers. In order to install such support, you will need to have some knowledge of Windows 95® since the process involves finding a folder on the Omni 1000 CD-ROM disk that bears the name of the synthesizer you want to use.
Another problem with Omni 1000 is that right now there is no direct support for refreshable Braille displays. This can be a drawback for those who rely on this method of output. At the very least it makes it more difficult for deaf-blind people to use Omni 1000.
There is a way to get around this limitation, however. Although Omni 1000 does not support the use of refreshable Braille displays itself, it can be configured to operate like any other Windows 95® application. If you use a screen-access program which supports a refreshable Braille display, you will be able to use the display with Omni 1000; and as is the case with any traditional Windows 95® application, you will need to be familiar with conventional Windows 95® menu procedures. It would be better if Omni 1000 included built-in support for refreshable Braille output.
A Word about Copy Protection
The Omni 1000 program is copy-protected. This practice has never been popular and has nearly ended in the world of mainstream software. But it is still fairly prevalent with programs written for blind people. Most of the screen readers (programs that allow blind people to use computers with speech synthesis or refreshable Braille) are copy-protected in one form or another.
When installing Omni 1000, you must place the supplied key disk in your floppy drive while running the installation from your CD-ROM. Kurzweil Educational Systems allows three installs. This means that, if you are a licensed owner of Omni 1000, you may install the software on three computers, provided you are not using these three computers simultaneously. For most people in most situations, this is sufficient. In circumstances involving hard disk failures which result in the accidental loss of an install, Kurzweil says that it will furnish you with an additional installation if required. However, it is possible that because of copy protection you could be without the use of the software for several days while waiting for an additional key disk to arrive.
Memory and Processor Speed
We know that the Omni 1000 software will run on a machine with only sixteen megabytes of RAM. However, to get the most out of this software, thirty-two megabytes is preferable. Several factors contribute to this conclusion.
First, analyzing and converting a page uses a large amount of your computer's resources. Accordingly, the faster your system and the more memory you have available, the more rapidly the recognition process will occur. Second, Omni 1000 employs a software speech-engine, which requires additional system resources. Added to the optical-character-recognition operation, this creates a challenge for your machine. We have tested the Omni 1000 software on machines ranging from a Pentium 133 with sixteen megabytes of RAM to a Pentium 200 with sixty-four megabytes. Naturally the faster machine significantly outperformed the others.
Since a reading system turns scanned images into text which must be spoken, it must provide the user with the means to control the reading and management of scanned documents. Not only does the Omni 1000 provide the tools needed to perform these tasks, but its many features expand upon these concepts. Those who design reading systems for blind people do everything they can to create programs that are useable regardless of the level of computer expertise. Omni 1000 is no exception. What distinguishes it from its competitors is that it provides so many interfaces or ways to talk to the computer program you're running. Omni 1000 provides four distinct methods for the user to communicate with it. We will describe all of them as we discuss system features.
When you purchase the Omni 1000, you get a seventeen-key keypad. This keypad does not differ substantially from the numeric keypad customarily found on the right side of a standard 101- or 104-key computer keyboard. If the user of the Omni 1000 knows little or nothing about computers and wishes to preserve that blissful state, this keypad could be attached to the computer in place of the full-size keyboard. The keypad has three distinct modes or layers: the reading, file management, and settings keypad. When Omni 1000 is loaded, you are placed in the reading keypad. From this keypad you can scan and read, move to different parts of a document, control speaking speed and volume, and access the dictionary and spell-word features. In the file management keypad, you can create folders (once known as directories); delete folders; move between folders; and open, close, and delete documents. You can also save documents in the file format you choose. Omni 1000 offers 127 possible file formats. From the settings keypad you can change some system characteristics; for example, Omni 1000 can use one voice for reading documents and one for system information, e.g., prompts and error messages. From the settings keypad layer you can assign the voices you like to these functions. You can also change the way in which Omni 1000 scans documents. For example, from the settings keypad you can regulate the brightness of the image and determine whether to scan and recognize a file immediately or scan files to image for later recognition. Moving from keypad layer to keypad layer is a fairly easy task.
In the early 1970's, when Dr. Raymond Kurzweil was developing the first reading machine, he worked closely with the National Federation of the Blind. One of our suggestions incorporated in that early reading machine was to include a nominator key. By pressing the nominator key followed by any other key on the keyboard, the user could hear a brief help message about the second key. Omni 1000 preserves this function with the help-and-status key. If you need help with the function of any key in any layer of the keypad, first press the help-and-status key followed by the key for which you need information. A brief help message will follow. If you press the help-and-status key and hold down the key about which you want help, you will receive a more detailed message. This feature provides a quick way to familiarize yourself with the keypad and its functions. There is also a keypad command that allows you to load the online manual when more in-depth help is required.
The Menu Bar
Omni 1000 uses a menu bar similar to those found in most Windows 95® applications. Therefore, if you are familiar with such menu bars, you will already be able to take advantage of this interface. If you don't already know how to navigate in a menu, learning to use the Omni 1000 menu bar will provide an excellent way to familiarize yourself with the use of this standard Windows 95® feature. What is unique about Omni 1000 in this regard is that, if you are using the program's built-in speech, all of the menus talk.
From within the menus you can open, close, save, and delete files and open and create folders. You can begin scanning, change scanner settings, and change other Omni 1000 settings such as reading speed, reading voice, and system voice. Some menu choices have sub-menus; others perform a specific action, such as start-reading or start-scanning. When a sub-menu exists, Omni 1000 informs the user.
Other Omni 1000 features, settings, and operations are accessed using other standard Windows 95® controls (e.g., dialog boxes, edit boxes, check boxes, and the like). All of these controls will speak if you are using the built-in Omni 1000 speech.
Finally, many Omni 1000 menu items can be accessed directly with short cut keys, a common feature in most Windows 95® applications.
The fourth and final interface is the voice-recognition command system. With voice commands you can start scanning, start reading, spell a word, get a dictionary definition for a word, raise and lower the speaking volume, and raise or lower the reading speed as well as invoking many other program functions. To use voice commands, you must have a microphone properly connected to your sound card. You must also configure your microphone correctly for use with voice-recognition commands. If you don't have a Windows 95® screen reader, you will require some assistance because it cannot be done from within Omni 1000 itself.
Voice-recognition commands require a lot of system resources. Unless you actually need the feature, we recommend that you leave this function out.
We did not find the voice interface reliable. We had to repeat our requests many times to get the system to understand what we wanted. We tested this feature using three different computers and four different microphones and achieved the same poor results every time.
Omni 1000 can be used simply to scan and read a page of text. However, there are a number of features that can add more flexibility to the process. As we have discussed, you can initiate a scan from either the keypad, the menu bar (using shortcut keys), or voice recognition commands. When you begin a scan, Omni 1000 responds with the message, "Scanning, please wait." We tested Omni 1000 using two different scanners=FEthe HP4P and HP5P, both manufactured by Hewlett Packard. When using the HP4P, we noticed that scanning speed was somewhat more sluggish than when using the HP5P. This was the only significant difference between these scanners.
During the scanning process Omni 1000 provides progress messages. These quickly become tedious. One of the first things we did was to disable them. You can cancel a scan in progress, provided that the scan was not completed. Simply pressing the start-and-stop-scanning key on the keypad or F9 on the keyboard will immediately terminate a scanning operation. If you are using the keypad to perform Omni 1000 operations, you can scan only when in the reading keypad layer.
Omni 1000 appends pages as they are scanned to the open file. As soon as a page has been scanned, you can begin scanning the next page. You need not wait for recognition to be complete.
Scanning and Reading Simultaneously
The Omni 1000 allows you to scan and read at the same time. You can scan a document while the system is reading the previously scanned pages. When using Flextalk®, the software speech synthesizer installed with Omni 1000, we noticed a lot of choppiness in the speech during scanning. Words are skipped while scanning is going on. However, when testing Omni 1000 using a traditional hardware speech synthesizer such as a DECtalk PC or an Accent PC, this choppiness did not occur.
Re-scanning a Page
Omni 1000 offers the ability to re-scan a page. You may scan a page and then realize that the results may be flawed, e.g., the brightness is set incorrectly, or the page was not properly aligned on the scanner. If this occurs, you can choose the re-scan page option from the scan menu, and the page you have just scanned will be automatically erased and scanned again.
Inserting a Page
When scanning lengthy documents, it is easy to skip a page inadvertently. Inserting a page means that you can correct this problem. When choosing this option, you can scan the missing page and have it inserted before the current page in your document. The pages will be renumbered to reflect the addition of the new page. This option is not available in other OCR reading systems designed for blind users. Until now, the only way to insert missing text has been to save the insertion as a file and use a word processor to add the page to the document.
When the print quality of a document is poor, one way to improve the scanning results is to adjust the brightness of the image. Many scanners support this feature. Omni 1000 can accept brightness values ranging from zero to one hundred, where zero is the darkest setting and one hundred is the lightest. For most documents we tested, a brightness setting of fifty resulted in reasonably good recognition.
Omni 1000 supports the automatic contrast feature, which is available only when you use a Hewlett Packard scanner. This feature is referred to by Omni 1000 as dynamic thresholding. This means that the lightness or darkness of the image is adjusted automatically during the scanning process based on the quality of the page. When dynamic thresholding is on, previous brightness settings are disabled until dynamic thresholding is turned off. This feature is useful when you are uncertain about the quality of a document. When testing Omni 1000 using the HP4P scanner, the scanner hesitated (stopped and started repeatedly) when dynamic thresholding was enabled, but this problem did not occur when we used the HP5P scanner.
The term "batch scanning" means scanning a number of pages and deferring their recognition or processing until later. Omni 1000 supports this feature, which it calls scanning to image files. You can scan pages which are saved as image files but start processing them when it is convenient. To do this, you must change the scanning mode from scan and recognize, the default setting, to scan to image file. An image file is a digital picture of the page. It cannot be read since the text has not been recognized.
To extract text from the digital image, you must switch the scanning mode to recognize image files. Then you must invoke the start-scanning function. As each image file is recognized, it is automatically deleted. One useful feature of Omni 1000 batch scanning is that you can safely interrupt the recognition of previously scanned pages. However, you must save the already recognized pages with a file name. When you wish to resume recognition, open the saved file, switch scanning mode to recognize-image-files, and invoke the start-scanning function.
Omni 1000 provides the usual reading features=FEread by line, read by sentence, and read by paragraph. You can use any of the interfaces described earlier to move around by the reading unit you have selected. In addition, you can invoke a continuous reading function that can be halted with a single keystroke.
You can move around within the document using standard arrow keys, and you can also move quickly from page to page or jump to a specific page.
Dictionary and Spell Features
Omni 1000 incorporates an electronic version of the American Heritage Dictionary in its software. Accordingly, you may obtain a definition for a word by simply positioning the cursor on the desired word and using either the keypad or the menu bar to select the dictionary feature.
Omni 1000 has a spell-word feature. When listening to text read by synthetic speech, it is easy to misunderstand a word. Using the spell feature can eliminate these problems. When the spell command is invoked, Omni 1000 first says the word and then spells it.
Perhaps the most powerful feature in Omni 1000 pertains to editing. It supports many of the editing commands found in word processors. You can delete characters, use insert or typeover mode, and select text for cutting and pasting. The text-selection commands are the same as those found in most Windows 95® applications. Omni 1000 also allows you to cut and paste text, which is another standard capability in most Windows 95® editing programs. Although it might be preferable to edit scanned documents using a word processor with which one is familiar, the fact that such editing capabilities are built into Omni 1000 makes it a powerful tool.
Omni 1000 also includes a find feature, which allows the user to search for specific text strings within the document. This feature is not case-sensitive, providing another powerful tool to aid in the examination of a document.
Omni 1000 allows the user to create and save bookmarks. You can create only one bookmark per page. A bookmark is useful when you have scanned a long document and you wish to resume reading at a specific place, or if you wish to select portions of the document to revisit. Please note that a document saved in a text format or as a word-processed file cannot use the bookmark feature. Omni 1000 stores the location of the sentence containing a bookmark and will return upon request. Bookmarks can also be deleted.
File Management Features
Omni 1000 allows you to save a scanned document in any of 127 possible file formats. We prefer to save documents in plain-text format, since file-conversion routines often misinterpret the format of the original document.
When Omni 1000 is installed, it creates a number of folders in which documents can be stored: books, manuals, recipes, etc.
You can delete these folders and create your own. You can also
create a favorites list of folders. You can create a file before
you have scanned any text into it, or you can scan the text you
wish to read and save the document to a file when you have
finished. Omni 1000 allows you to name files from the numeric
keypad. When you use the save-file key on the numeric keypad,
Omni 1000 prompts you to accept the default name for the file or
to enter a name of your own. Specific keys on the keypad
correspond to letters and digits, but the system is cumbersome
and slow. It is much easier to use the keyboard.=20
You can also create a favorites list of the files you are most likely to use. All of the favorites commands may be accessed using either the keypad or the menu bar. A favorites list allows you to choose only those files or folders you wish to access regularly.
Omni 1000 provides the ability to alter display colors, to highlight text as it is spoken, and to magnify text on the screen. These features could be of some use to those with some remaining vision.
There are several settings that may be changed in order to allow you to use your screen reader. You can disable active speech so that Omni 1000 no longer communicates directly with your chosen speech synthesizer. You can also disable Omni 1000's use of the numeric keypad. This is essential if you wish to use an existing Windows 95® screen reader since nearly all such programs use the numeric keypad for their commands.
For many users of OCR systems, the most important question is, how accurate is Omni 1000? All reading software is extremely limited in what it will read correctly. Documents with unusual layoutsfont variation, multiple columns, and graphics on the same pagewill cause any of the OCR packages to produce poor results. Print is a visual medium, so documents are usually designed to catch the eye rather than to convey information. When dealers of reading software or reading machines tout their products, one might conclude that all reading tasks can be accomplished with these systems. Nothing could be further from the truth! Even if we assume that an OCR package had an accuracy rate of 99 percent, think of the possible errors contained in that one percent! Therefore, no one should use an OCR package to read bank statements or pay bills. Although OCR is improving, financial matters leave no room for error.
Omni 1000 was tested using a number of documents. When given a typewritten letter, a novel, or a laser-quality computer printout, Omni 1000 did remarkably well. We also found that it identifies and properly presents multiple columns better than any of its competitors.
However, it cannot read inverse text (light text on a dark background), which is found on book jackets, brochures, and catalogs. Stephen Baum, who is primarily responsible for the design of Omni 1000, says that this feature will be present in the next release of the software. Omni 1000 does not read poor-quality faxes as well as does Open Book version 3.5 from Arkenstone. Yet in most cases we found Omni 1000 to be extremely accurate. Keep in mind, however, that no reading system today can produce error-free results.
Omni 1000's recognition speed is fairly fast. On a Pentium 200 with sixty-four megabytes of RAM, it took Omni 1000 under ten seconds to recognize the facing pages of a paperback novel. But as document layouts become more complex, the recognition speed begins to slow down considerably.
Omni 1000 incorporates an automatic page-orientation feature. This is essential for a blind user since he or she has no way of orienting material correctly for scanning if it has not been previously identified. Omni 1000 will notify you if it has scanned a blank page and will not add it to your document.
The overall differences in recognition accuracy between reading systems are not great. Therefore, when choosing reading software, you should look for a program that provides the features you need. Omni 1000 incorporates so many features that it would be difficult to imagine that it could not satisfy the needs of most users.
Room for Improvement
Unfortunately, Omni 1000 is the most costly reading software package for blind people today. Its price of $1,295 is about $300 more than Arkenstone's Open Book Unbound.
The help-and-status key, discussed earlier, provides help for only the keypad=FEnot the menu system. In other words, there is no context-sensitive help of the type that has become a standard for conventional Windows 95® applications. Direct support for refreshable Braille displays is a deficiency that must be addressed if Omni 1000 is truly to be of use to the deaf-blind. We also find the use of copy protection an unnecessary inconvenience that should be eliminated.
One area of concern is technical support. Though we received good technical support from Kurzweil Educational Systems, this support is extremely difficult to obtain because nearly every time we called, we got a busy signal. This is certainly another thing which Kurzweil Educational Systems could improve.
Finally, the installation procedure should include the ability to install support for hardware speech synthesizers such as the DECtalk PC/Express or any other of the currently supported devices. We also hope that the list of supported devices is enhanced in a future release.
Omni 1000 is by far the most feature-rich OCR package for blind people currently available. There are so many different ways to use the program and so many useful tools that it is clear that Kurzweil Educational Systems has established a new standard in this field. The documentation is quite good, and a Braille manual is available.
No brief review can address all the questions that arise when considering the purchase of such complex and expensive software. If you are considering making such a purchase, we strongly suggest that you try the programs yourself. When doing so, you should bring the kinds of documents with you that you would ordinarily scan. One way to accomplish this would be by visiting the International Braille and Technology Center for the Blind. All of the available reading systems are on display here, and we would be glad to assist you in making your decisions. The National Federation of the Blind has been closely involved with reading systems for the blind since the beginning, and we will continue to provide unbiased information regarding this technology.
You may contact Kurzweil Educational Systems at the
following address: Kurzweil Educational Systems, Inc., 411
Waverly Oaks Road, Waltham, Massachusetts 02154, Phone (800) 894-
5374 or (617) 893-8200, Fax (617) 893-4157, E-mail:=20
World Wide Web: http://www.kurzweiledu.com