Design and Implementation of an object-oriented Media Composition FrameworkSkip other details (including permanent urls, DOI, citation information)
This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 3.0 License. Please contact email@example.com to use this work in a way not covered by the license. :
For more information, read Michigan Publishing's access and usage policy.
Page 220 ï~~Design and Implementation of an Object-oriented Media Composition Framework Philipp Ackermann University of Zurich, Department of Computer Science, MultiMedia Laboratory Winterthurerstrasse 190, CH-8057 Zurich, Switzerland e-mail: ackerman@ i f i. uni zh. ch Abstract The goal of the presented project is a general compositonal environment for multimedia with interactive editing facilities. Multimedia presentations are regarded as hierarchical compositions of time objects that define serial or parallel synchronisation of the inserted media objects. Such media composition hierarchies support automatic temporal layout mechanisms and are integrated into an object-oriented application framework that provides direct manipulative interaction on temporal structures through user interface components. In this papaer reusable and extendable components for audio and music processing are presented. Keywords: multimedia composition, audio, 3D graphics, music, time synchronisation, animation, user interface interaction, object-oriented application framework, C++, MET++. 1. Introduction Today computer workstations are used to integrate different media such as text, graphics, audio, and video into one single system in which these media types are handled and manipulated fully digital. In such multimedia systems, editing of media data can be done with high accuracy, no loss of quality, and extensive control possibilities. These editing features enable users to add more expressiveness into the creative process of media composition. Incorporating visual and audible media comes to interest in many fields, e.g. mixed-media art, information presentation, learning environments, audio-video postproduction, data audification/sonification. In order to support the creation of multimedia data, authoring systems must permit direct manipulative interaction on graphical representation of the large amount of parameters defining content, composition, synchronisation, and execution of media data. The hardware technology that is needed for multimedia applications is available today as separate components. However, programming software for multimedia applications is a difficult task because the following requirements must be fulfilled: " integration of several media types with different realtime constraints at low-level device interfaces " support for media compositions to define presentations with high-level intermedia synchronisation specification and real-time control " intuitive user interface interactions with direct manipulation of graphical representations and high semantic feedback " flexibility and protability in view of different hardware configurations and platforms, having in mind the rapid hardware evolution and high hardware dependencies because of analog-to-digital & digital-to-analog converters, digital signal processors, and other special purpose processors. The complexity of interactive multimedia software systems compels attention to the software engineering aspect. In this project the powerful concepts of object-oriented application frameworks are used to integrate audio-visual multimedia data types such as audio and 3D graphics into a reusable software system. An application framework is a set of interconnected objects that provides the basic functionality of a working application, but which can be easily specialized (through subclassing and inheritance) into individual applications. An application framework not only allows the reuse of code as a class library, but also the reuse of design structures, because the dependencies between objects are preimplemented through predefined object composition, event dispatching mechanisms, and message flow control. This means that the glue between components such as media representations, real-time controlers, user interface components, media views, user actions, etc. is inherited as functionality from the framework. 2. System Architecture The media framework is based on the object-oriented application framework ET++ [Weinand, 1989], which was developed at the University of Zurich in the programming language C++. ET++ is an objectoriented class library integrating user interface building blocks, basic data structures, support for object input/output, printing, and high level application framework components. It eases the development of interactive textual and graphical applications with direct manipulation and high semantic feedback. In our project the 2D graphics model of ET++ was extended to 3D graphics, audio classes were added, and time structures were designed [Ackermann, 1993a/b]. Hardware dependencies are hidden in a portability layer, that provides abstract interfaces to operating system, window system, as well as audio and MIDI I/O (Figure 1). The main development platform of the MET++ (= Media-ET++) project is currently a Silicon Graphics Indigo workstation. 3. Time Synchronisation For multimedia applications time synchronisation is an important aspect. Looking at the the structure in Figure 2, the music piece consists of many parts on several layers. This hierarchy contains implicit information about grouping and temporal constrains. The root node "music piece" defines part-of relations for intro, theme, and solo, meaning that they Composition, Composition Systems 220 ICMC Proceedings 1994
Page 221 ï~~Pronramminn Environment JI Application I L.vy.awrru.ruay rr>rrrvrwr.v..aJ l Application Framework Classes Graphical Building Blocks Media Classes Application, Manager, EventHandler, VObject, Media, MediaView, Timer, Document, Window, View, CompositeVObject, Menu, TEvent, Synchro, Loop, Command, Dialog, Data, Clipper, Scroller, Box, Sequence, Conductor, Clipboard, Converter,.. Slider, Button, Image, TimeFunction, AudioUnit, TextView, View3D, Samples, Note, Scale, Camera, Light, ThreeD, Chord, MusicInstrument, Material, - MusicPlayer, TimeView,.. Applications Media Application Framework Class Library -1 Basic Building Blocks Object, Class, Collection (List, Set, Array), Point, Rectangle, Ink, Font, Text, Bitmap, Time, Intensity, Pitch, Beat, File, Stream, Filter,. (Graphics) Port (Operating) Systen MIDI Audio _nPortability Printer Port Window Port / Window System SystemSystem _ _1Layer,,,,, W in PEXPHGS+ i]y.....Audio Adaptors 3D Graphics DB IDILaJ for Sytm 1 r-- - Liray Concrete SyteILCT~ i Lir SUN Audio System GL / nventor"m 3DLir a[ y j SW & W pGoraphis DB DEC VMS'' [~I IOpenGLr L ---*P udo -d Vid graphics::: CPU- Memory Audio.DSP Workstation ft g:--: -:,:.: -......:........... r-...":T.......... r_...,-Z.......1-.., -1_;_ _i_._..so --,-, - ___. ]; i ll: id:o.. mebufer....:...: Serial Ports..Network'; SCSI A..:..!D.. C-I. Figure 1. The System Architecture of the MET++ Media Application Framework. TEvent CompositeTEvent TSynchro TSequence TimeLine TFunction TBoolFunction TlntFunction TFloatFunction TStringFunction TTimeFunction NoteList TShift TLoop TBool Tint TFloat TString TTime Note TVObject TTextltem TCamera TThreeD TAudio TAudioSamples MusicPlayer MusicalContext Conductor Table 1. The Class Inheritance Hierarchy of TEvent. are played in sequence. The duration of the music piece is specified by the sum of the duration of its child objects. The chord node "Cm7" specifies a parallel temporal structure of its child objects, e.g. the notes should be played in parallel and the duration of the chord is the maximum of its child duration. Abstracting from this music example, we introduce temporal specification by hierarchical composition as a generic mechanism that includes information about part-of relations, grouping, and temporal constrains and that allows automatic time layout calculation based on its part hierarchy. Furthermore, the composition structure can visually presented so that its implicit and often hidden temporal information will be explicit and available for direct manipulation. music piece intro theme solo piano sax piano drums / \ /1\ /"\ /1\ Cm7 G7 Figure 2. Temporal Composition. 3.1 Time Events In the media framework time layout objects and media objects are the basic building blocks of multimedia presentations, which are interleaved together in a hierarchical composition structure. The components of this structure are modelled as time events and have their own starting point, duration, and virtual time line. The time events are combined through a flexible composition mechanism with automatic object positioning along the time axis. In the time line approach, objects are attached independently from each other to a time axis, and removing an object does not affect the time position of other objects. In contrary, removing a time event from a composition hierarchy will initiate a new time position calculation so that the duration change will be propagated along the tree. Time dependencies are realized as an object-oriented inheritance hierarchy whose base class is called TEvent (time event). All time dependent classes are subclasses of TEvent (Table 1) and inherit its synchronisation features. The high-level synchronisation accuracy is based on millisecond time resolution. 3.2 Time-dynamic Media Objects In a multimedia presentation, the temporal relationships between media objects are configured by composing universal time objects such as TSequence, TimeLine, TSynchro, TShift, and TLoop with special media objects. Any time dependent object inherited from TEvent can be inserted into such an object composition structure. There exist predefined media classes derived from TEvent for animated visual 2D & 3D objects, camera, audio, and music components (Table 1). They are realized as time wrappers to existing objects in the ET++ class hierarchy. It is easy to extend the time framework by realizing a new time wrapper for a specific object class. The time dynamic behaviour of a time wrapper is supported through time functions that manipulate values of the controlled object. Actions such as fading, scaling, or positioning of media objects are expressed with time functions and are handled within each class independently. A time function can manipulate the time itself, so that each TEvent can specify a local (object) time, which maps the global presentation time to a specific temporal behavior. TEvent objects are synchronized to the global presentation time at the beginning and ending, whereas in this interval, it is possible to deform the linear time progress. 3.4 Real-time Presentation For real-time multimedia presentation, the root object, usually a Conductor object, periodically sends Perform messages to the time dependent media objects along the composition hierarchy. The ICMC Proceedings 1994 221 Composition, Composition Systems
Page 222 ï~~media objects compute only the data that is necessary to perform the next interval in time. Continous media data is computed incrementally ahead into a buffer. Each media class itself is responsible for solving low-level hard real-time constraints and hides media stream dependencies (e.g. writing to device drivers, using DMA to transfer streams to DSP chips, starting a subprocess or thread, etc.). Interactions for fast-forwarding, rewinding, or slow-motion are realized by changing the parameters of the message Perform. The parameters of the Perform message are the start time and duration of the interval that should be performed, plus the realtime interval duration. This means that each media object can adopts its behavior according to the relation between presentation duration and real-time duration. A graphical controller of the Conductor provides interaction for start, stop, pause, fastforward, rewind, speed change, looping, loop points, locator points, and time position setting. 4. Visualization of Temporal Structures In contrast to synchronisation specification by language (scripts), the benefit of hierarchical object composition having time functions is that its structure can be presented and specified graphically and allows interactive multimedia authoring with direct manipulation techniques. The time event graph (Figure 3) is an obvious visualization of the temporal composition structure and reflects the partof hierarchy. Figure 4. Time Composition View. 5. Audio and Music Classes The following classes are basic building blocks for audio and music processing. The class Intensity is used for representing rational values for amplitude, volume, gain level, etc. It can operate on and convert itself to byte, integer, floating-point, dezibel and MIDI values. The Pitch class is able to handle information about pitch key and pitch frequency. It uses the TonalSystem class to map symbolic keys to physical cycles per seconds (Hz). The class Beat models beat and bar properties and delegates quantisation and the mapping from symbolic beat measure to pysical time in seconds to the MusicalContext class. The PitchScale class can represent symbolic scales (e.g. chromatic, dorian) or physical intervals for tuning purposes (e.g. equal-tempered in cents). Figure 3. Event Graph View. The boxes in the event graph represents the hierarchical structuring where the leaves of the tree are time-dynamic media objects and the inner nodes are grouping elements with a specific temporal layout. Unfortunately, the correlation between objects and the time axis is not visible. Without a time axis it makes no sense to visualize time functions. Therefore, a new time view was created, called time composition view, which depends on a time axis and still visualizes the hierarchical grouping structure (Figure 4). It displays the temporal structure in a linearized form along the time axis in the x direction and the hierarchical composition structure in the y direction. The spatial and temporal placement of the components is automatically determined in each time object according to its child objects. Flexible direct manipulative editing features support the media creation process. AudioUnit AudiolO SGIAudiolO SUNAudiolO AudioSamples AudioFile SNDPort WAVEPort AIFFPort Sound CDPort SGICDPort DATPort SGIDATPort Amplifier Mixer MonoStereo StereoMono Delay Oscillator WaveTable SoundPlayer Table 2. Inheritance Hierarchy of AudioUnit. Audio resources are modelled with the abstract Audio Unit class (Table 2) as modules of a source-filter-sink architecture. An AudioUnit object is either a source that produces audio samples, a sink that consumes audio samples, or a filter through which audio samples flow. The output of an Audio Unit can be sent to different other Audio Units and thus define an audio signal flow graph. This flow and its parameters can be configured in the audio system editor shown in Figure 5. ICMC Proceedings 1994 Composition, Composition Systems 222
Page 223 ï~~Different audio file formats (AIFF, SUN/NeXT, RIFF/WAVE) are supported. Audio files can be played back in real-time from disc or can be converted to a Sound object, that holds the samples in memory. The samples can be visualized and be edited in a sample view or in an envelope view. The interpretation context of a musical performance is modelled in the time-dependent MusicalContext class. It holds information about the tonal system, tonality, signature, measure, and tempo. The MusicPlayer interprets its notes in this context and delegates the note playing to the abstract MusicInstrument class. The concrete subclasses MIDJInstrument, CSoundlnstrument, and Synthesizer map the internal note representation to device-specific synthesis parameters. The MIDlInstrument sends MIDI messages over a serial interface to external MIDI devices, the CSoundlnstrument controls a subprocess that is running CSound [Vercoe,1990] as a server, and the Synthesizer dynamically allocates Oscillator objects in order to produce the sound with its own audio components of the media framework. A MusicPlayer can switch at run-time the MusicInstrument that it will control. Converters for Standard MIDI Files and CSound Score files are provided. A control object for a Yamaha DMR8 digital mixer/recorder was developed in order to realize mix automation within the framework. Other MIDI devices can easily be integrated. 6. Sample Applications Different applications are under development in order to test the functionality of the time and media extensions and to improve the multimedia framework by redesign. One of these test applications is shown in Figure 5. It is a multimedia authoring tool that handles 2D drawings, images, text, 3D graphics, audio, and musical data. The hierarchical object composition ("Lego"-approach) for visual and temporal structures are displayed in separate windows. Based on the functionality of the media framework, a new audio mixing metapher was realized. Audio so that dependent on their position in the room and the camera view the loudness, panorama and reverb is set accordingly. This allows very intuitive arrangements in setting to music for computer animations and film post-productions and proves the additional possibilities gained by the integration of 2D and 3D graphics with audio and music components. 7. Conclusions A large effort has gone into the design of the MET++ multimedia application framework to provide real reusable classes. This paid off by simplifying the task of writing multimedia applications and reducing development time. Altough the learn effort to understand the framework is comparatively high, the results of projects using it are encouraging. By end of the year 1994 it is planned to release a public domain version of the MET++ multimedia application framework over ftp at f tp. if i. uni zh. ch. 8. Acknowledgements Thanks are due to Andre Weinand and Erich Gamma, which are the principal designers and developers of ET++. The presented work is supported by the Swiss National Science Foundation. References [Ackermann,1993a] Philipp Ackermann: Object-oriented Modelling of Time Synchronisation in a Multimedia Application Framework; in: Audio Engineering Society AES 95th Convention, Preprints, Audio Engineering Society, New York, October 1993. [Ackermann, 1993b) Philipp Ackermann, Dominik Eichelberg: Combining 2D User Interface Components and Interactive 3D Graphics; TOOLS USA '93 Conference Proceedings, Santa Barbara, August 1993. [Vercoe,1990] Barry Vercoe, D. Ellis: Real-time CSound: Software Synthesis with Sensing and Control; in: International Computer Music Conference Glasgow 1990 Proceedings, p..209-211, ICMC Assoc., 1990. [Weinand,1989] A. Weinand, E. Gamma, R. Marty: Design and Implementation of ET++, a Seamless Object-Oriented Application Framework; Structured Programming, Vol. 10, No. 2, June 1989, pp. 63-87. Figure 5. A media authoring tool developed with reusable components of the MET++ Multimedia Application Framework. ICMC Proceedings 1994 223 Composition, Composition Systems