Effectiveness of Telepresence Learning Environment for Opera Singing: First Results from a Case Study Teresa Rojas-Rajs Multimedia Applications Laboratory, Polytechnic University of Catalunya, Spain trojas2010@gmail.com Francesc Alpiste Multimedia Applications Laboratory, Polytechnic University of Catalunya, Spain alpiste@ege.upc.edu Pedro Lorente Audiovisual Cluster, i2cat Foundation, Spain pedro.lorente@i2cat.net Francisco Iglesias Audiovisual Cluster, i2cat Foundation, Spain francisco.iglesias@i2cat.net Joaquín Fernández Multimedia Applications Laboratory, Polytechnic University of Catalunya, Spain jfernandez@ege.upc.edu Josep Monguet Multimedia Applications Laboratory, Polytechnic University of Catalunya, Spain jm.monguet@gmail.com Abstract: this brief paper shows the first results of a case study on a remote learning telepresence environment, specialized on lyric singing at higher education level. Opera eLearning study uses high bandwidth for delivering quality audio and video experience, and has been reviewed and evaluated by singing teachers, chorus and orchestra directors, and other professional musicians. Besides the exhaustive material recorded in video, which allows contrasting each of the singing lesson against a set of categories, usability testing and a user acceptance questionnaire has been applied. The analysis of results shows that the environment can be effectively used for masterclass lessons. Introduction Nowadays remote learning systems are wide spread across the World Wide Web. Several solutions offer virtual classroom through videoconference with students, also sharing documents, forums, and other asynchronous means of delivering learning contents. Regarding music education there are interesting examples in blended learning solutions, mixing presence tutoring with the availability of contents, community support and automated review of exercises, like VEMUS (Virtual European Music School, 2009). This and other projects had advanced on supporting the execution and evaluation of the student in automated manners, like controlling gestures, measuring pitch and tone, involving sensors, visual aids and computer-assisted feedback. Such researches are crucial for the future of music education, and are maturing to the stage of even deliver standards for further applications, like the Symbolic Music Representation (i-maestro, 2009). From the musical point of view, the instruments selected are winds, strings, percussions or piano. These kind of multimedia systems can fulfill the help required in the daily practice by computer-assisted games and motivating feedback (Percival, Wang & Tzanetakis, 2007). All users can benefit from them, but user with special needs, children, starters and lifelong learners are key users, since in this stage of the musical learning the main objective is to obtain the proficiency in execution due a consistent repetition and check up. In the other hand, one-to-one tutoring is at the core of learning process at the conservatory level; in a recent study (Gaunt, 2008), teachers have characterized such relationship like parental, friendly, collaboratively curious or like a doctor and patient. Teachers seek transmitting the skills acquired in their own professional life and experience, and are strongly committed to achieve students’ own development, literally to “develop a personal artistic voice” (Gaunt, 2008). Even more, when it comes to music at advanced higher education or graduated level, the council and support given by an experienced teacher is not oriented to teach the instrument: this knowledge is supposed to be mastered already to a high degree of proficiency. Therefore the teacher is devoted to review beyond the technique the interpretation of musical repertoire, conducting the expression (Karlsson & Juslin, 2008), the emotions and artistic quality of the performance, often using images and metaphors that make easier for the student to understand the feeling required (Woody, 2006). For the present study is important to denote that a very common practice, the masterclass, that involves one teacher reviewing the work of one student with other students attending as observers, it is also a one-to-one tutoring (Gaunt, 2008), since it does not involve a proper peer review, that is, the other students are usually not allowed to participate in the evaluation process. Several sceneries justify the need for a remote class. Some opera singing teachers are active musicians, so they have seasons of concerts and travel a lot; the need to continue the tutoring with their students is a good chance for remote learning sessions. Offering a masterclass remotely could mean that students that may not be able to travel can reach an important master. Such virtual classroom involves requirements on audio and video quality: fast speed synchronous communication, excellent view, and a high quality sound approach. According to “experiential telepresence” definition (Draper, Kaber & Usher, 1998), the present study implements a telepresence environment, because promotes that the users feels to share space with distant interlocutors loosing awareness of the distance, and emulates the remote space, specially in the sound experience for the teacher. Music higher education institutions are already delivering remote classes and educative content through high quality videoconference and broadcasting (Manhattan School of Music, 2009). High-bandwidth availability allows improving videoconference’s quality. Such is the case of the use of Internet2 (Internet2, 2009) in Miami for orchestral higher education level (NWS America's Orchestral Achademy, 2009) . Our case study would be the first experience only devoted to opera singing, and is using the Géant2 European network platform (Géant2, 2009). The Study Opera eLearning project included a case study focused on remote learning of lyric singing, which has delivered 14 masterclasses, 4 of them between cities in Spain and 10 between Sabadell (nearby Barcelona), and Amsterdam. As part of the project, prior to this masterclasses several pilot tests has been developed inside Spain in order to define the proper audio and video settings. A few lessons learned from the pilot stage were the definition of a minimum standard for the microphones and speakers, to dismiss the possibility of using headphones, to confirm the absolute requirement of use a piano accompanist at student´s site due the communications delay, and how the immersive sound emulating the space on the rooms could be optionally used to improve the listening of the teacher, although stereo configuration allows to deliver the class as well. The system is intended to be used soon in a opera singing graduate course, and also includes a Web Platform with the common tools (forums, documents, messages) that won´t be discussed beyond in this paper, but is an important part of the whole blended learning proposal of Opera eLearning project. The recorded sessions compressed in MPEG-4 format are between the published documents for further student’s revision. In the recording and publishing process we had followed most of the instructional strategies for video-assisted music education mentioned by Anderson & Ellis (Anderson & Ellis, 2005). At the masterclasses two teachers, ten professional opera singers as students and six pianists participated. Two kind of repertoire has been reviewed, opera and lied singing. In each session there were observers at free demand, which included students, and until five evaluators. Evaluators were opera singing school directors, orchestra or chorus directors, or audio engineers. All evaluators have wide experience (over 25 years) and works nowadays at the more recognized institutions related to opera from Catalonia, in Spain. Goals and Methods Opera eLearning had the goal of design a remote learning environment for opera singing at the graduate level. It was an exploratory case study (Yin, 2003), where previous hypothesis are well defined but we were committed to find additional information trough data recovered. It was a common work in engineering, because we were testing an artifact. In this sense, we could also define the study as a validation or verification study, from the usability point of view (Rubin, 1994). The hypothesis were user oriented: the model of remote singing learning room (1) was enough for the teacher to fully evaluate the student; (2) allowed the student to understand observations and emulate the teacher’s sound; (3) observers could use the information received as well and (4) a high quality, immersive and space distributed sound could improve teacher’s listening. The main question to be answered, at last, is prove if it was possible to deliver proper masterclass remotely with this artifact. During all the classes we had video and audio recording; for the analysis, categories of research were defined, which are listed in the following table. Such categories can be reviewed against each of the hypothesis. Category Learning objectives Subcategory Technique Expression Sound quality objectives Classroom effectiveness Tones Dynamics Harmonics Class support Communications fluency Specific information about subcategories Collocation Air Support (“Appoggio”) Articulation and diction Conducting (“piano”, “forte”, entrances, coordination) Emotion and feeling High, low and medium frequencies Changes between piano and forte Harmonics, timbre and color Seamless remote teaching and learning Speaking, singing and visually communicate Table1. Research categories defined for Opera eLearning Case Study A questionnaire has been applied to teachers, pianists, evaluators and students as part of their participation. Only teachers and some pianist charged for their services, but students got the classes for free. Observers and tech staff present in the room were requested to answer the questionnaire as volunteers. We recovered 195 entries, including qualitative and quantitative data. Five point Likert-Scales were used, with 1 indicating the lowest (‘‘bad’’) and 5 indicating the greatest sound quality (‘‘excellent’’), or with 1 indicating the lowest (“nothing”) and 5 indicating the greatest classroom effectiveness (‘‘very much”). Subcategories have been mapped into constructs for the questionnaire. Learning objectives achievement has not been asked with Likert-Scales to everyone, but commented by teachers and evaluators with free text fields. All questionnaire´s participants had a free text field for the classroom effectiveness category. The questionnaire was filled after each masterclass when possible, or after a set of three of them. In this brief paper we will discuss how effective the classroom seemed from the teacher´s and student´s point of view, measured by the class support construct. Further results will be developed using other variables of the questionnaire and exploiting in a deeper review the qualitative data from recordings. Tools and Material “GÉANT2 is the high-bandwidth, academic Internet serving Europe’s research and education community” (Géant2 2009), equivalent to the Internet2 in the U.S. Communication software selected was Ultravideoconferencing (McGill Ultra-videoconferencing, 2009) for audio and DVTS (Digital Video Transport System, 2009) for video, sending separately PAL video signal and audio channels (DV 720×576 at 25 FPS, which consumes 30MBps on each direction, and 96Khz/24bits per channel, requiring 2MBps each). From students’ classroom to teacher’s site five audio channels were sent (emulating Dolby 5.0), and two channels from teacher’s classroom to the students (stereo). Although around 80 Mbps were used, availability of 100 Mbps is recommended. In the pilot phase prior to the case study several communications technologies were tested, including HD video which consumes 1 GB bandwidth. All telepresence classrooms have been installed on single rooms, whether large classrooms in a university, auditoriums and even a TV studio. Nevertheless we had no control on the acoustics of the room, minimum requirements of ambient noise < 40 dB and a pink noise test have been done previously at each site. The masterclass zone included the piano, lighting, one or two plasma monitors, speakers and a video camera. Teachers delivered the masterclass sited at the piano, while singers stood next to it, so each camera should be accommodated to take a mid shot of the teacher at the piano and a wide shot of the singer including the piano accompanist. Video cameras reached a computer (double processor at 2.40 GHz, 2 GB RAM, Windows Operating System) through the IEEE 1394 port for the DVTS video communication. Only professional, condenser, pre-amplified microphones were used. Teachers used a tie clip microphone for the voice and an Omni polar pattern microphone for the piano. At student´s classroom two of such microphones where assigned to the piano, while a stereo microphone (composed by two cardioids polar pattern microphones) was used for the singer. Two microphones were situated far away, at the end of the room for receiving additional acoustics and harmonics. All microphones inputs entered into a professional audio mixer at each site. The mixer sent audio channels to a professional computer audio card with AD/DA (24 bit/96kHz) converter and recorder, installed on a computer (double processor at 3.16 GHz, 2 GB RAM, Linux Operating System) for the Ultravideoconferencing audio transmission and reception. Since audio signal arrives prior to the video signal, we used the mixer´s delay function to coordinate sound and image, by delaying the sound. Each an audio and a network engineer supported the classroom; they were located in another area in the room. A third area was designated to evaluators and observers, who had their own speakers and plasma screen, or have enough visibility to the plasma screen used by the singers. Teacher´s speakers were distributed around the classroom for the Dolby 5.0 simulation using the five channels received from the student´s side. Findings Questionnaire data has been analyzed with SPSS analytical software. For statistical purposes, pianist answers were grouped in the “students”, since they were receiving instructions from teachers and shared closely the space with singers. Figure at the next page (Fig. 1) shows the frequency of class support construct answers grouped by profile. There is a clear tendency to higher values, confirmed by the descriptive statistics for the whole population that results on a mode of 4 with variance of 0,775, with a rank and a median of 4 as well. Only 22.1% of the answers were in the lowest three values (1 to 3). Students were 13.3% of the sample, while teachers contributed with 5.1%. Students keep the median at 4, with a variance of 0.695, while teachers have a median of 5, with 0.143 of variance. Teachers used to make it easier for the students, some recorded sessions includes phrases like: “listen, you may stop, this is a regular class so you may stop, if something happens or you are tired, you may stop”; “we had finished now, is a good experience for us to work with this huge distance, from North Europe to nearby Barcelona…a little nice adventure, because we note that is possible to work this way…some little irritations we have now must be improved, but for the next time we will know how it works, so we may be more relaxed than today…”. In their writing evaluations all teachers agree in participate on this kind of class for next experiences, valuing the sessions as positive and remarking the need to improve any technical detail that may interrupt the telepresence experience. Students have expressed feeling compromised to effort, they wrote things like: “I felt that because of the distance I had to do everything bigger, exaggerated, which it’s very useful for a concert as well”. In the other hand “I was very insecure about our sound on Netherlands, and in a delicate matter like lied singing, I need a closer contact to the teacher that I cannot reach from a screen monitor”. Or even stressed, saying in the class: “I am aware of too many things and I cannot concentrate on my own”. Students that attended to more than one masterclass tended to declare other kind of things in the second session, pointing to technical details like “piano volume was too low in the first class, now it´s better”. In this sense, teachers were more experienced because served many sessions, while students only took one or two. Most of students qualified the experience as positive and also pointed that they would like to obtain feedback from their regular teachers during opera season, while teachers are not available. They also agreed in be open to take a formal, curricular masterclass in this manner. Figure1. Frequency of answers to class support construct by testing profile. Besides the questionnaire data, first demonstration of effectiveness is that the classes have been executed. An interesting fact that may be observed in each recorded session is that after a while teachers and students forgot that they were in a virtual space, relaxed and start to work fluently and deeply, speaking to the plasma monitor and forgetting about the rest of the people in the room, even trying to reach some functions that are still impossible, like playing together at the same time. Conclusions Regarding the classroom effectiveness from the teacher and student point of view, findings shows the agreement on the viability of the artifact proposed, with improvement required in specific zones well identified. What are those “little irritations” that are mentioned by the teacher above? Masterclasses have been delivered most of the time fluently, although some network issues were experienced in six international sessions, without avoiding the masterclass to be fully delivered, but introducing some interruptions in the sound. Connectivity matters should be addressed and perhaps a workaround must be considered if required, for instance we could turn to a stereo, DVTS only configuration that requires less bandwidth. In some locations, where classrooms where too small, echo could be heard from time to time, due the audio technician’s delay on adjusting the volume for each teacher or student intervention. This is not a problem in proper auditoriums, but should be reviewed anyhow. Another theme in terms of long-term telepresence classrooms is that intelligent agents should take care of functions that in this study were accomplished by technicians. That’s far beyond the scope now, but is clearly an excellent research line. Main weakness of this study is the limited number of cases, and the possible bias of population used. Teachers involved certainly have an open mind about working with technology. It is important to keep researching over a population that uses this classroom regularly, as part of usual curricula. Also we look forward to be able to equip permanently laboratories with the telepresence classroom, which may be another weakness of the study, since some variables could be affected by the different rooms used. The aim of this telepresence application is not substitute the personal tutoring that is so valued in music education, but to allow in some cases the remote tutoring. This study has been developed for voice, but other instruments could be the subject. The idea of a telepresence music learning classroom should be completed with the computer-assisted tools and aids that were introduced at the beginning of this paper for integrate a whole blended learning system (part remote and part presence learning), where support for daily practice is enabled, opportunities for one-to-one tutoring increases and Web platforms enhances the social interaction and information delivery. References New World Symphony America’s Orchestral Academy, NWS. 2009. Internet2. Retrieved January 3, 2009 from NWS official site. Web site: http://www.nws.org/AcademyInternet2.asp. Anderson, A. J. & A. Ellis (2005) Desktop video-assisted music teaching and learning: new opportunities for design and delivery. British Journal of Educational Technology, 36, 915-917. Digital Video Transport System, DVTS. 2009. DV Stream on IEEE1394 Encapsulated into IP. Retrieved January 3, 2009 from WIDE Project. Web site: http://www.sfc.wide.ad.jp/DVTS/. Draper, J. V., D. B. Kaber & J. M. Usher (1998) Telepresence. Human Factors, 40, 354-375. Gaunt, H. (2008) One-to-one tuition in a conservatoire: the perceptions of instrumental and vocal teachers. Psychology of Music, 36, 215-245. Géant2. 2009. Retrieved January 5, 2009 from Géant2 official site. Web site: http://www.geant2.net/. i-maestro. 2009. Retrieved June 1, 2009 from i-maestro official site. Web site: http://www.i-maestro.org/. Internet2. 2009. Retrieved January 5, 2009 from Internet2 official site. Web site:http://www.internet2.edu/. Karlsson, J. & P. N. Juslin (2008) Musical expression: an observational study of instrumental teaching. Psychology of Music, 36, 309-334. Manhattan School of Music. 2009. Special Programs - Distance Learning Overview. Retrieved Fabruary 15, 2009 from Manhattan School of Music official site. Web site: http://www.msmnyc.edu/special/distancelearning/. McGill Ultra-videoconferencing Research Group. Retrieved May 3, 2009 from McGill Ultra-videoconferencing home page. Web site: http://ultravideo.mcgill.edu/. Percival, G., Y. Wang & G. Tzanetakis. 2007. Effective use of multimedia for computer-assisted musical instrument tutoring. In Proceedings of the international workshop on Educational multimedia and multimedia education. Augsburg, Bavaria, Germany: ACM. Rubin, J. 1994. Handbook of usability testing : how to plan, design, and conduct effective tests. New York: Wiley. Virtual European Music School, VEMUS. 2009. Retrieved January 5, 2009 from VEMUS official site. Web site: http://www.vemus.org/. Woody, R. H. (2006) Musicians' cognitive processing of imagery-based instructions for expressive performance. Journal of Research in Music Education, 54, 125-137. Yin, R. K. 2003. Case study research : design and methods. Thousand Oaks, Calif.: Sage Publications. Acknowledgments Opera eLearning has been possible due to collaboration between the City of the Music of Sabadell City Council, directed by the singing professor Àngels Civit, the i2cat Foundation and the Multimedia Applications Laboratory. Another main promoter of the project has been Joan Francesc Marco, nowadays director of the Gran Teatre del Liceu, formerly director of the City of the Music project. Further information may be found at www.opera-elearning.com. First author is studying the Ph.D. at the Polytechnic University supported by grant 208288 from CONACYT, the National Council for Science and Technology of Mexico.