Introduction

To put this article into context, also see the review of OpenBook with Pearl, which is a similar product. Also see the Eye-Pal and OpenBook with Pearl comparison.

Who can benefit?

It is relevant to people who, for any reason, have difficulty reading text from paper documents. This includes people who have little or no vision, those who have visual perception problems and those whose physical limitations make holding paper documents difficult or impossible.

What is it?

The Eye-Pal converts paper-based printed material into electronic text almost instantly. It is used in conjunction with a Windows-based or Apple Macintosh computer and connects to the computer via a USB port (it must be a USB2 port). The Windows version was tested during this evaluation. All operations are performed on the computer through software supplied with the Eye-Pal. Power is supplied by the computer.

Rather than employing a conventional scanner, the Eye-Pal captures contents of a page through a camera. A light in the camera allows use in a wide range of lighting, including complete darkness. Using a camera means that the capture process is very rapid. While that is not surprising, the process of converting the image to text to be displayed on screen and spoken by the speech synthesizer is usually almost as fast.

Installation

The software is installed prior to connecting the Eye-Pal. Before connecting the Eye-Pal, it is “unfurled”. The horizontal rails that form an L shape on the tabletop are locked into position by small swivelling latches. The arm holding the camera is raised to the horizontal and clicks into place. This process is very straight forward for anyone with reasonable dexterity.

When the USB cable is connected for the first time, a “found new hardware” dialog appears. Select, “No, not this time” and click Next. Select “install the software automatically”. On subsequent connections to that USB port, the Eye-Pal will be recognised automatically.

Using the Eye-Pal

A document (either single page or an open book) is placed on the table that the Eye-Pal is standing on, using the horizontal rails of the Eye-Pal as a guide. The rails define an area of an A4 page. Smaller items, down to business cards, can be processed. Options for processing a book or magazine where two adjacent pages occupy a larger area are discussed below.

There are two modes for processing material. The most basic processes one page at a time. Each time a new page is captured, text from the previous one is lost. Contents of each page can, however, be saved prior to processing the next. The more sophisticated mode involves creating a book. This process requires entering a title and, optionally, an author. Each captured page is saved as a separate jpeg file (discussed further below). When working in this mode, several options are available.

1. The book can be processed one page at a time. This will usually involve rotating the book for odd and even pages. That some of the text on the page adjacent to the one being processed is within the camera’s view usually does not matter.
2. All odd pages can be processed. The page count is then reset to 2 (assuming commencement at page 1) and the process is repeated for even pages. The software then collates the pages.
3. If the book is small enough and there is sufficient gap between text on the adjacent pages, two pages can be scanned at once.

When processing material as a book, the camera can be activated manually or automatically. To capture text manually, the camera is activated by pressing the spacebar on the computer once the document has been put in position. When utilising the automatic function, the camera is triggered when no movement has been detected for several seconds. This proved to be a very effective method of processing multiple pages, the camera never being activated prematurely. It also meant that effort was not taken away from holding bulky books in position.

Accuracy

The Eye-Pal was tested in lighting conditions ranging from a large office with a good coverage of fluorescent lighting, a small room with natural light through a window and total darkness. Having resolved an issue resulting from an undocumented feature discussed below, changing light made little difference to accuracy. As photographers know, varying lighting conditions can significantly affect results if the camera and/or photographer do not adjust for different conditions. Unlike a conventional scanner, products such as the Eye-Pal must deal with this issue.

A variety of printed material was used during testing. This ranged from relatively uncomplicated text on A4 paper from a laser printer to textbooks and magazines. The text from the laser printer was almost completely error free. While there were more errors from textbooks, results were still very impressive. Information at tops of pages was at the greatest risk of being misinterpreted. Importantly, books were quite thick, resulting in considerable curving of the paper towards the spine. Complex magazine layouts were handled very effectively, material being consistently presented in an appropriate sequence. There was greater likelihood of errors when pictures were intermingled with text. While this occasionally led to severe deterioration of accuracy, results were generally very acceptable. In summary, an informal assessment of accuracy put it at a comparable level to that achieved with a conventional scanner and commercially available OCR software.

Formatting

While the Eye-Pal presents material in a linear manner, suitable for synthetic speech or electronic Braille output, it can also report on page layout. Available information includes the number of sections, columns and headings. Another important detail is whether the page is upsidedown, which allows easy orientation of a book by a blind user. There are commands for navigating by character, word, sentence and section.

A single page can be saved as either a jpeg (image) or text file. A jpeg can be reopened in the Eye-Pal software.

A book can be saved as a text, rtf or xml file or as a series of jpegs, one for each page. A book saved as jpeg files can be navigated page by page or a specific page number can be selected. The disadvantage of this option is that errors cannot be corrected. When saved as an RTF file, each page is formatted as a single paragraph. Text files are saved with a hard return at the end of each line and multiple hard returns following a paragraph. Page numbers are included at the top of each page. The chosen format may vary depending on such factors as personal preference, nature of material and the type of editing (if any) to be undertaken.

Each page can also be saved as an audio file, using the currently set speed of the speech synthesizer. Because the wave format is used, audio files are quite large.

Documentation

Installation instructions, a quick guide and a manual are supplied in both MS Word 2003 and PDF formats. The PDF files are not structured for assistive technology access and lack bookmarks, a valuable navigation feature. A product primarily aimed at blind people could be expected to include a fully accessible PDF manual. People who use a screen reader with relevant facilities will appreciate inclusion of formal heading styles in the Word files.

Information is provided in an informal, friendly tone. Apart from one omission, information on the various aspects of the product is easy to locate and to digest. The omission is that there is no reference to turning the light on the camera on and off. This is done by pressing ctrl-l, but see under Issues below for more on this.

Issues

The only problem encountered during testing involved the camera light. If, after pressing the spacebar to start the camera, ctr-l is pressed, the light is toggled on or off and the change is announced by the speech synthesizer. This command should not toggle the light if the camera has not been started, but it does, with no audible feedback. In early testing, the command was pressed inadvertently while issuing a screen reader command. No reference to turning the light on and off could be found in the manual and it took some discussion with the distributor and manufacturer to resolve the issue.

Specifications

Introduction

To put this article into context, also see the review of Eye-Pal, which is a similar product. Also see the Eye-Pal and OpenBook with Pearl comparison.

OpenBook is optical character recognition (OCR) and reading software specifically intended for use by people who have limited vision or visual reading difficulties for other reasons. The product, which runs on the Windows platform, was acquired by Freedom Scientific some years ago and continues to evolve. Until recently, hardcopy material was processed by OpenBook via a conventional scanner connected to the computer. This option is still available, but Freedom has recently released the Pearl, a camera that communicates with the software via the computer’s USB port. This allows considerably faster processing of material and, when coupled with a notebook computer, makes a very portable scanning and reading package.

Installation

Under best circumstances, installation of the software is straight forward. It is important to do this before connecting the Pearl to the computer and this is clearly indicated in the accompanying documentation. The only problem encountered during testing involved activating OpenBook. While activation can be done over the internet, the Department’s firewall prevents this. An alternative involves sending specific information to Freedom Scientific via the distributor and receiving an activation code. After three weeks of swapping emails and entering codes and receiving error messages, this process was abandoned. A physical dongle was then used and this was successful on the second attempt. Without activation, OpenBook runs in demonstration mode, requiring a computer restart every 40 minutes.

Highly versatile software

The OpenBook software is much more than an easy-to-use OCR package (that process will be discussed below). Facilities include:

• Synthetic speech that can be turned on or off to suit individual needs
• Material can be displayed in a wide variety of sizes and colours
• A scanned document can be saved in OpenBook’s proprietory format and also exported to text, Word (.doc), RTF and HTML formats
• Material can be sent to a Braille printer and also saved as an audio file xx
• An OpenBook file can be read in either “text” (a simplified view) or “exact” view, depending on needs and circumstances
• Once scanned, a document can be easily edited

It is easy to use basic features of the software without reading the accompanying documentation. This, however, is likely to result in the user missing out on many valuable features. Documentation, while perhaps being a little verbose, is well written and organised. It can be accessed from within OpenBook and is also provided in text and RTF formats.

The Pearl camera

When folded, the Pearl can be housed in the supplied neoprene case. The case offers very good protection and includes a shoulder strap. Both the camera and a notebook computer will comfortably fit into a small backpack or similarly sized luggage.

Setting up the camera is quick and easy. The arm holding the camera hinges to a roughly horizontal position and locks into place. A small button at the pivot point is pressed to release it for folding. The Pearl is powered from the computer’s USB port. A light on the camera can be toggled on and off by pressing ctrl-l, but testing did not include an environment with dim light.

Scanning

OpenBook automatically recognises scanners that are connected to the computer, including the Pearl camera. If more than one scanner is connected, the user chooses between them.

By default, scanning is commenced by pressing the spacebar. If the Pearl is being used, it takes only a few seconds to capture and process a page. In “acquire and read” mode, a fresh scan is performed with each press of the spacebar and new material is placed on a separate page. In “multipage automatic” mode, the Pearl recognises when movement of the document has ceased and captures the next page. If a book is small enough, it can be turned sideways to allow both pages to be scanned simultaneously. On bulky books however, presumably due to pages curving near the binding, this mode was highly inaccurate during testing. Another option is to scan all odd pages, then even pages and then have them collated.

Three scan qualities are available. The user can choose from fastest, mixed and most accurate. Over a reasonably wide variety of material during testing, however, the “most accurate” mode, while taking a little longer, usually yielded less accuracy than the mixed mode. As with other facets of the software, a large number of scanning settings can be adjusted. While providing considerable flexibility, an inexperienced or less technically oriented user is likely to find the array of options daunting. As already mentioned, reading the documentation will help. Fortunately, default settings are well chosen for many situations.

OpenBook as a reading tool

With its in-built synthetic speech and highly flexible visual display options, it is very effective as a reading tool for people who have little or no vision and will also suit some who have visual reading difficulties. A scanned document retains information from each page discretely, which can be important in study situations. Bookmarks and comments can be placed in the document and the dictionary and thesaurus can be used to check the meaning of the current word.

Specifications

It is suggested that, before reading this comparison, you read the separate reviews of the Eye-Pal and OpenBook with Pearl . The products do similar things, but there are some distinct and some subtle differences.

Copy protection

The Eye-Pal software is not copy protected. That is, it can be installed on as many computers as required without restriction. The only scanning option, however, is the Eye-Pal camera. While the camera must be connected to capture material, material already processed can be read on any machine hosting the software.

The OpenBook software is copy protected. A single licence allows installation on several computers, but only one copy is to be used at one time. Activation can be done over the internet. An alternative is to use a dongle. The computer to which the dongle is connected will then have a fully functioning copy of the software.

File format

The Eye-Pal stores scanned material as a series of images. Each time a page is accessed, text is recognised again. While this is usually done very quickly, a difficult page can take somewhat longer. Another important issue is that errors cannot be corrected without exporting the material to another format.

OpenBook uses its own proprietory format, which includes text and images. The default read-only mode does not allow editing, which prevents accidental alteration. It is, however, easy to switch to edit mode to correct errors.

Accuracy

Without having both systems available at the same time, precise comparison is difficult. Another factor that makes accuracy comparisons difficult is that the same software can yield different results when scanning the same page more than once. That said, the Eye-Pal seemed to be slightly more accurate. It also usually processed material a little faster. On the other hand, as mentioned above, OpenBook allows fast and easy correction of errors.

The Eye-Pal’s motion sensing was superior. It automatic mode, the camera did not trigger prematurely. The Pearl did so occasionally, which could be particularly inconvenient when processing in duplex mode, where pages are to be subsequently collated. In manual mode, the Eye-Pal would not capture the page until all movement had ceased. The Pearl, on the other hand, did not utilise movement detection except when in automatic mode.

The cameras

The Pearl camera is more heavily built than the Eye-Pal. Its stand offers only one straight edge for aligning paper, while the rails forming the base of the Eye-Pal form a right angle, which makes it a little more obvious where documents are to be placed.

The Pearl’s carry case is more robust than the somewhat untidy bag housing the Eye-Pal. Not only does it offer much better protection, but it is aesthetically more pleasing.

Conclusion

Both products offer very fast access to hardcopy printed material. Despite the vagaries associated with text recognition, under most circumstances both offer high levels of accuracy. Becoming fully acquainted with the Eye-Pal requires considerably less time and effort than is needed for OpenBook. Especially with its recent substantial price cut, the Eye-Pal can be a very effective tool for those simply wanting rapid access to printed material. While quite some time and effort is needed to get the most out of OpenBook, the reward is a very powerful and versatile package. Even with minimal familiarisation, it can be used effectively to provide access to hardcopy material.

Among OpenBook’s many facilities are settings to assist people who have limited vision or visual reading difficulties. Equivalent facilities are not provided in Eye-Pal, but are available in a separate product, Zoom Ex.

It is also helpful to consider these products in the context of broadly available scanners and OCR (optical character recognition) software. Quite high levels of OCR accuracy can be achieved for as little as several hundred dollars. For that, however, scanning speed would be very slow – at least 30 seconds per page or slower – and formatting options would be limited. A scanner designed to handle bound books and magazines and powerful OCR software would cost in the order of $1,000. Processing speed would still be significantly slower than with the Eye-Pal or Pearl. Nor would such an option be portable and people with visual reading difficulties would not have immediate access to the processed material.

Thanks to Loree

This item was prompted by a question from Loree:

I have a blind friend who wants to create a book. He wants to dictate the book and then verbally edit it since he can’t use a computer. What is the best technology for him to use? Once he gets the book recorded, I assume I will need to get an audio-to-text converter for my computer to make the “book.” Do you have any suggestions for that?

Introduction

While the above question relates to a blind person, the following information is relevant to a much wider section of society. Those who, for any of a variety of reasons, have literacy problems will often benefit from this technology. Similarly, people who have difficulty or suffer pain when using a keyboard may find dictation to be a viable option. The portable equipment discussed below is also appealing to people who do not like being “tied to a computer” when working.

Simple options

While fast disappearing, a standard cassette recorder could be used to dictate the material. The major benefit of this option is low cost. Disadvantages include:

• Poor ergonomics;
• Limited opportunity for editing; and
• Slow (real time) transfer to a computer for subsequent transcription.

A dedicated dictation machine using a micro cassette would overcome the ergonomics issue. Cost would be higher than that of a basic cassette recorder. Transferring material to a computer would also be slow. An alternative to transferring to a computer would be to key the material into a computer while listening to the recording. This could be done from the original dictation machine or, ideally, from a dedicated transcription machine equipped with a foot pedal.

Digital dictation equipment

This equipment offers good ergonomics and fast transfer to a computer. Some examples also allow deletion of unwanted sections of a recording and insertion of material. That is, a recording can be treated very much like text is in a word processor. There is an important issue to be aware of when considering a digital recorder. Models intended specifically for dictation allow the user to stop and start recording while in the one file. “domestic” examples, on the other hand, usually start a new file each time recording is recommenced. A very fluent author may find this latter approach acceptable, but the creation of multiple short files can become a logistical problem for those who like to listen to a passage and then resume recording.

Both Olympus and Philips make sophisticated digital dictation machines. Examples that allow insertion and deletion within a recording are not cheap. They do, however, allow a person to dictate and edit material with a high degree of precision. A file can be transferred to a computer via a USB cable for further processing, discussed below.

Computer-based dictation software

While Loree’s request was specifically for dictation options that did not involve a computer, the following is worth consideration. Subject to issues such as motivation and dexterity, a blind person with limited computer knowledge could use a computer to dictate material and edit it. Inexpensive software such as Dictation Buddy from High Criteria for Windows offers excellent facilities. Ideally, the software would be used in conjunction with a screen reader. The free NVDA screen reader would be more than adequate.

With a basic microphone connected to the computer’s microphone or USB input, the user would need to learn a number of keyboard commands to operate the software. Much of this learning could, however, be avoided by using a dedicated dictation microphone such as the Philips SpeechMike Pro 5274. The microphone includes buttons for controlling the software.

Transcription options

There are two approaches that can be taken. The most appealing is to use speech to text software which, under the best circumstances, produces a near-perfect transcription automatically. On a Windows computer, Dragon Naturally Speaking is the obvious software for this purpose. Subject to a raft of issues, achieving a high level of accuracy may or may not be possible. If many errors are made, the job of editing can be a very tedious one. On the other hand, if all goes well much time and effort is saved.

The more mundane approach is for a competent typist to transcribe the material. The Dictation Buddy software mentioned above is well suited to this task. If this option becomes necessary, a foot pedal to control play, rewind and fast forward would be a valuable asset.

Background

A comprehensive review of the Stream was conducted in 2007. This review covers the optional SoftPack which has recently become available. For $US49, it increases the functionality of the Stream to provide access to:

• Books in the EPUB format
• Enhanced Audio books from Audible
• MP4 audio, including iTunes
• MS-Word docx (Word 2007 and later) files
• Record in MP3 and wav, as well as the unusual 3gp format already available

Purchasing SoftPack

The SoftPack can be purchased online from the SoftPack section of the Humanware website. It is necessary to provide an email address, serial number of the Stream onto which the SoftPack is to be installed and credit card details . The process, however, is not automatic. In the case of Australian customers, information is passed from a North American office to the Australian office. In our case, there was also a problem with the serial number being recognised by the system, which was resolved after several days.

Installation

Before installing the SoftPack, it is important to upgrade the Stream firmware to version 3.2. It is free and is available from the Humanware Support site. A small file is sent via email. This is copied to the root folder of the SD card in the Stream and it is turned on. The Stream announces that the SoftPack has been installed and, from then on, the additional features are part of the system.

Features

As mentioned above, MS Word 2007 and later files can be read on the Stream. If correctly structured, a file can be navigated by the various heading levels. Disappointingly, it cannot be navigated by specific page numbers. Another disappointment is that earlier MS Word files cannot be read. Once installed on a host computer, the free Stream Companion software can convert these to HTML, providing a reasonable level of navigation.

Availability of books in the EPUB format is very welcome. Importantly, protected books cannot be read. However, a large number of free and commercially available unprotected EPUB books are available.

EPUB is a subset of the DAISY format. Unlike DAISY books, an EPUB book is ultimately contained in just one file. These books do not include an audio component and are therefore read via the Stream’s inbuilt synthesizer. There are many thousands of both free and commercially available EPUB books. Access to this format therefore has the potential to expand markedly ones access to literature.

Navigation options are similar to those in DAISY books – various heading levels, paragraph, sentence etc. Navigation to a specific page is not available, at least in the books sampled so far. This is apparently an EPUB formatting issue rather than a limitation in the Stream, but investigation continues. Bookmarks can be inserted at any point. .

Audio recordings made on a Stream that does not include the SoftPack are in the unusual 3gp format. The Softpack adds MP3 and wav files as options. Importantly, the hardware limits recordings to mono. If nothing else, having access to the extra two formats demonstrates just how space-efficient the 3GP format is. one hour of recording in that format produces a 9 mb file, increasing to 41mb for MP3 and a substantial 303mb for a wav file. If recordings are to be used on anything but the Stream, 3GP files must be converted using the Companion software. MP3 and wav files, on the other hand, do not require this process. When listening to recordings made through the Stream’s internal microphone, there was negligible difference in sound quality across the three formats. The benefit of MP3 recordings is immediate compatibility with other equipment. There seems little point in choosing the wav format unless a high quality external microphone is to be used.

Enhanced Audible and MP4 files have not yet been tested. The latter, provided they are unprotected, should not pose a problem. It is necessary to have an Audible account to access that material.