Digitization Of The Book: A Report On Present Trends

Back to table of contents

• A New Generation of Audio Books
  • 6.5 Archiving/Storage and Retrieval
  • 6.6 Daisy (Digital Talking Book System)

A New Generation of Audio Books

6.5 Archiving/Storage and Retrieval

Current digital technology facilitates the upgrading of the sound quality of analog recordings prior to the archiving process. Some deteriorating analog audio books 20 years old or more can be restored digitally and archived for future generations. Due to the rapid evolution of digital audio technology, it is possible that audio files currently produced and/or archived digitally will in future have to be transferred to an entirely different storage medium.

Storage media can be characterized respectively as slow (archival), medium (near line storage) and fast (primary storage) in terms of retrieval response time. A system utilizing a mix of these media, that is, magnetic tape, hard drive and optical technology to achieve the above requirements should be transparent to the user. Different levels of storage may be required. An hierarchical storage management system that facilitates high speed retrieval of frequently requested titles as well as cost effective storage of less frequently used materials is desirable.

The degree to which audio is compressed during mastering or archiving is usually considered in relation to its effect on the digital audio. With lossy compression, as the level of compression increases, the quality of the audio decreases and the 'bits' of data that have been removed for the purposes of compression are not restored when the file is decoded. However, the level of lossy compression which is possible far exceeds that of lossless compression. For archival purposes, lossless rather than lossy compression appears to be appropriate. Also, as the future cost of digital archival storage decreases it may be preferable to archive digital audio files in an uncompressed state. To this end, the RNIB (England) is currently testing 'archiving to CD'.

6.6 Daisy (Digital Talking Book System)

DAISY is a digital talking book system which is being developed specifically for the purpose of providing human voice recorded books for print handicapped readers. The system consists of four modules: the recording module, the playback module, the tape transfer module and "ToC" (table of contents) maker. All four modules are in their alpha version (0.65) and are presently running under Microsoft Windows. A beta 1 version has just been released and beta 2 is under development with release planned for the very near future. The compression currently used is ADPCM-type coding and decoding. Implementation of MPEG 3 compression is planned for 1996. A DOS version is in development.

The design objectives of the DAISY Digital Talking Books System are:

• "To give the blind reader the same -- or better -- flexibility in accessing the book's information as a fully able user has when reading a printed book. This means fast, random access to any part of the recording."
• "The system should be independent of distribution media to adapt to future distribution technologies. Furthermore, the system should not be fixed to any particular method for digital audio data compression, since these technologies are evolving"

This system requires a minimum hard drive size of one gigabyte. Master production is done directly onto the hard drive. DAISY produces digitized human voice recordings, stored in a compressed data format on CD-ROM. CD-ROM technology is utilized to produce copies, but mass production from a digital master is possible.

DAISY differs from other digital audio recording systems in that there is no waveform display. Standard hard disk recording of audio is digitization, that is, there is no recognition of the meaning of what is being recorded. DAISY is designed exclusively for recording human voice. It structures the digitized voice into "phrases" (voice data blocks) which are groups of spoken words representing the descriptors for a narrated sentence or short paragraph. The system utilizes a voice analysis module which does not record or store periods of silence in the incoming speech as normal data, but rather registers them as silence and measures the length of each period of silence. As the periods of silence are not stored, the storage of the recorded voice is compacted significantly to approximately 15 to 50 hours of storage per CD.

Files relating to each segment of recorded material are created in the hard drive, thereby reducing the amount of actual recorded voice. Although the storage capacity in relation to the format of the digitized audio may be 15 hours, there are not 15 hours of digitized speech on the CD. The sampling rate which provides 15 hours of storage results in an acceptable playback sound quality. This is lossy compression in that the actual breaks in narration or periods of silence are not recorded and when the audio is played back, artificial 'timed' silence is inserted in its place.

The DAISY format specification potentially allows for other data types, such as text or graphics, to be stored together with the voice data. All types of data would be linked together at the "phrase" level.

During post production editing, an electronic text 'talking book table of contents' is created and the headings/topics within this table of contents are cross referenced to the relevant point within the digitized audio. Users have almost instantaneous access to any point within the book when selecting table of contents entries. Within the text, it is possible to 'skip' from one 'phrase' to the next. In addition, the system facilitates note making and 'bookmarking' within the text.

The DAISY system provides recorded human speech, compression to increase storage and expanded search/indexing and locate capabilities. However, it requires post production indexing to create search points, thus increasing production time. Its retrieval capabilities compared to the flexible search capability of electronic texts are limited to the selected index terms.

The DAISY talking book reading system consists of a standard IBM compatible PC (486 SX or better), equipped with a standard CD-ROM XA and photo CD-compatible drive, a 16 bit audio card and the playback software. The user reads the digitized book with DAISY specialized software, currently running under Microsoft Windows. The reading software will have a 'talking interface' which employs synthesized speech. The user, when reading a DAISY book, can play the digital audio back with the length of pause between the phrases reduced, thereby 'speeding' up the playback without pitch change. Testing of the DAISY playback system by a sample group of blind Swedish university students will be done in 1996.

To date, RNIB, the Danish Students' Library and the Swedish Talking Book Library have been able to transfer prerecorded analog audio into the system, but have not recorded a live narration. The transfer rate of analog recorded material into the DAISY System is twice normal playback speed. Specifications for hardware required to record (original or transfer) are: IBM compatible PC, Pentium, 16 Mb RAM, 1 gigabyte hard drive capacity,SoundBlaster 16 soundboard or compatible, PC speaker and quality microphone with microphone amplifier.

Although DAISY is still in development, an early version of the system has been installed in the following countries: Japan, Russia, Finland, Norway, Denmark (the Danish National Library for the Blind and the Danish Students' Library), Holland, United Kingdom (RNIB), and the United States (Library of Congress NLS/BPH, and, RFB & D). A consortium is currently being established to help fund its further development and to test it. Sweden, England, Spain, Japan, Holland and Switzerland are expected to join this consortium. In addition, organizations in the following countries have expressed an interest in serving as test sites: Norway, Denmark (2 organizations), Finland, the United States (RFB & D) and the CNIB Library for the Blind in Canada. DAISY 1.0 is planned for release by autumn 1996.