Online Braille Translation System from Digital Music s on the Internet BrailleMUSE Toshiyuki Gotoh, Reiko Minamikawa-Tachino, and Naoyoshi Tamura Graduate School of Environment and Information Science, Yokohama National Univ., The Tokyo Metropolitan Institute of Medical Science, E-mai: gotoh@sci.ynu.ac.jp, tachino-rk@igakuken.or.jp, tam@ynu.ac.jp Abstract This paper describes an Internet environment called BrailleMUSE to trnalate digital music scores into Braille scores. In experiments with expert translators, a translation error zccurred only at extremely slight frequency. Also, BrailleMUSE has a mirror page to a music site showing about 4,000 digital scores. It was confirmed that the system provided Braille scores by working together with the music site. It is suggested that the system is suitable for practical use. 1. Introduction Visually impaired people need Braille scores based on Braille music notation using Braille cells invented by Louis Braille in 19th century. The Braille scores are usually produced by expert translators who have learned Braille music notation. As it is extremely time-consuming task, there are not enough Braille scores to meet the needs of visually impaired people. The Internet enables sighted people to remotely access music as recorded sound and even music scores. However, visually impaired people cannot enjoy the benefits of the Internet on music scores. Our work is motivated by reducing the information gap between the sighted and the visually impaired on music scores. Here we propose an Internet environment called BrailleMUSE (Braille Music Support Environment) to design to translate digital scores shown on the Internet into Braille scores and to enable users to freely download them. Programs to automatically transcribe into Braille scores have been proposed using computer-based musical formats of MIDI, NIFF, and MusicXML. However, all their programs are used by stand-alone. Users are not allowed to use them via the Internet. The programs have been developed for MIDI [1] and NIFF [2, 3]. Also, GOODFEEL has been developed as a Braille music translator using MusicXML[4]. In addition,
DaCapo is an ongoing project to develop a software packages that will allow translating the MusicXML documents into Braille notes and vice versa [5]. 2. Web-based Braille translation The preliminary goal of the system is to associate digital scores shown on the Internet with the corresponding Braille scores (Figure l). The system assumes digital scores are mostly based on the MusicXML. The system provides free downloads of Braille scores so as to use them for personal purpose. Expert Braille translators judge whether revision is necessary for the downloaded Braille scores. If necessary, they may revise the Braille scores by themselves. The system intends to support expert Braille translators by cooperation with a web-based Braille translation system, an upload manager for data management, a converter into the MusicXML, and a web-site showing digital scores. The system provides a digital score database of about 100 scores and a mirror page linked to the music site showing 4,000 scores in Japanese [6] which enables users to legally download digital scores. The users select a music title in the database or in the mirror page, and then the system accesses automatic any the sites to download the digital score with the title. The system translates it into a Braille score. The users are able to download the translated Braille score. Recognition & Notation soft Visually Impaired User Braille Music BrailleMUSE Figure 1. System concept of the BrailleMUSE. Recently, digital scores are getting popular and wildly used among sighted people. Only we have a PC, we can access scores without going a shop or a library. However, visual impaired people need Braille scores; there still exists a gap for score information. In order to bridge this information gap between sighted and visually impaired people, we have started to develop BrailleMUSE since 10 years before. The final goal of our study is to create an environment to be able to get scores evenly.
Braille scores are produced based on Braille music notation that has signs and conventions with the same Braille cells as used in reading and writing sentences. Some symbols are often omitted in staff notation. The symbols, such as 5 th -point rests and 5 th - points accidental symbols, should be covered on translation. Braille music notation provides, as it were, a language system. However, Braille music notation is different from literary Braille and has its own syntax and abbreviations. 3. System architecture The system is implemented by web-based configuration with JAVA technology (Figure 2). A MusicXML document specified on a user s web browser is passed to the Braille translator through a servlet and an obtained Braille score is displayed on the user s web browser. The system enables the user to download it for personal use. For the mirror page linked to the music site, an upload manager is prepared to legally download digital scores. If necessary, the digital scores are converted into MusicXML documents. Web Browser Select File input MusicXML result download Server Upload List Maneger Servlet Braille Translation JavaServer Pages File Downloader URL Digital Braille Digital Digital Base file Figure 2. Web-based configuration of BrailleMUSE.
4. Performance of the BrailleMUSE The BrailleMUSE can translate wide scores from single-part to symphonic type. It is confirmed that scores with more than 40 parts, more than 3 staves/part, and more than 10 voices/part have been processed. Due to the hardware performance of our server, it can accept the MusicXML source file whose size is upto 5-7 MByte/score. It has server linkage mechanism via the Internet, which enables users to directly download the Braille scores translated using remote digital score sites without accessing original staff scores. If there is even a small scale personal computer conneted to the Internet, users can use the full functions of the BrailleMUSE. For the translation function, the BrailleMUSE supports a variety of Braille representations, such as partial- & full measure-accords, Braille repeats, and abbr. Users can specify representation according to their indivicual requirements by selecting a menu on the entry screen shown in Figrue 3. The supported representation and symbols are summarized in Table 1. Figure 3. Entry screen with option menus.
Table 1 Braille representation supported by the BrailleMUSE 1) Supported Braille representation; -- Flow: Full-measure in-accord, Partial in-accord, Braille repeats, Rhythmic group, Supplement of rest, accidental signs, and Pitch transcription. -- Symbols: Notes, Chords, Ornaments, articulations, dynamic marks, pedal, expression words and the other major symbols, except the following. 2) Unsupported major Braille signs; fingering & bowing signs, rehearsal marks, Lyrics (except Japanese), some repeat-marks (such as coda, Fine), double-note tremolo, and Inter-part arpeggio. 5. Trial use and discussion System interfaces to the mirror page to the music site in Japanese are shown in Figure 4, respectively. http://mysmaker.com/ http://gotoh-lab.jks.ynu.ac.jp/mirrorofmakerpage/ Over 4,000 s Figure 4. System interfaces of the mirror page to the music site.
The system enables to translate MusicXML documents which users scanned music sheets using music recognition & notation software. In our trial experimental, it was confirmed that converted MusicxML documents included conversion errors. The Braille scores were tested with the cooperation of expert translators to check whether there there were translation errors. Figure 5 shows an example of music sheet and the obtained Braille score by the system. Sinfonia4 (BMV.790), J.S.Bach Figure 5. A sample music sheet and the Braille score translated by the BrailleMUSE. As MusicXML documents were revised on convert errors in the digital score data base and the music site, translation error occurred at slight frequency. The translation error rate is nearly equal to the conversion error rate. Trial use suggests that the Braille translator may be suitable for practical use. The trial web-based Braille translation English version has been available at the page [7][8]. There are about 400 use of Braille translation a month on the page. 6. Conclusion This paper describes an Internet environment called BrailleMUSE to translate digital scores into Braille scores. BrailleMUSE supports a variety of Braille representation and symbols and the translated Braille scores are produced to meet individual requirements by selecting menu options. In experiments with expert translators, a translation error occurred only at extremely slight frequency. BrailleMUSE provides data & server linkage mechanism through the Internet. Using this function, we have also developed a
mirror page of a music site showing about 4,000 digital scores. It was confirmed that the system provided Braille scores by working together with the music site. Acknowledgments The authors would like to thank TONICA (The group of expert Braille music translators) and the group of expert Braille translators of the Japanese Red Cross Society Mie branch for helpful discussions and valuation. The work was supported in part by grants of Grant-in-Aid for Scientific Research from Ministry of Education, Culture, Sports, Science and Technology of Japan and by grants of The Technology Promotion Foundation. And lastly the authors are very thankful to the music site of Maker Page that directly helped us in extension of BrailleMUSE into Internet environment. References [1] K.Sato, M.Komatsu, Y.0yama, S. Matsuzaki, and K. Kanda, Construction of the Braille score system, IPSJ SIG, Japan, 2000-MUS-36-7, 2000. [2] D. Langolff, N.Baptiste-Jessel, and D. Levy, NIFF transcription and generation of Braille musical score. Computing in Musicology 12, 34-44, 1999-2000. [3] Toccata Braille Music Translation Software Toccata 2008 http://members.optusnet.com.au/~terryk/toccata.htm. [4] W.R. McCann, New tools for transcribers and musicians from Duxbury and Dancing Dots. Technology and Persons with Disabilities Conference 2005. [5] T. Kahlisch, M. Leopold, and C. Waldvogel, DaCapo, a project on transforming ink printed to Braille notes semi-automatically: the project s current state, Proceedings of ICCHP,224-227,2004. [6] A page of Maker. 2008. http://www.smaker.sakura.netp/index.html [7] D. Goto, T. Gotoh, R. Minamikawa-Taehino, and N. Tamura, A transcription system from MusicXML format to Braille music notation. EURASIP, Article ID 42498, 9 pages, 2007. [8] T. Gotoh, BrailleMUSE --Braille music transcription server--. 2008. http://gotoh-lab.jks.ynu.ac.jp/{braille_music_score_english/etop-new.html/braille_music_score /index.html}.