Note: This bibliographic page is archived and will no longer be updated. For an up-to-date list of publications from the Music Technology Group see the Publications list .

An Object-Oriented Metamodel for Digital Signal Processing with a focus on Audio and Music

Title An Object-Oriented Metamodel for Digital Signal Processing with a focus on Audio and Music
Publication Type PhD Thesis
Year of Publication 2005
University Universitat Pompeu Fabra
Authors Amatriain, X.
Advisor Serra, X.
Academic Department Department of Information and Communication Technologies
Abstract Classical models for information transmission such as Shannon and Weaver's still tend to be looked at as the only possible scenarios where signal processing applications can be formally modeled. Meanwhile, other disciplines like Computer Science have developed different paradigms that offer the possibility of looking at the same problem from a different perspective. One of the most favored approaches for software analysis and design is the Object Oriented paradigm, which proposes to model a system in terms of objects and relations between objects. An object is an instance of a real world or abstract concept and it is made up of an identity, a state, and a behavior. An object oriented system is thus described in terms of its internal objects, messages that are passed in between them and the way these objects respond to incoming messages by executing a particular method. Although object oriented technologies have been applied to signal processing systems, no previous comprehensive approach has been made to translate all the advantages and consequences, both practical and formal, of this paradigm to the signal processing domain. This dissertation defends the thesis that a generic signal processing system can be thoroughly and effectively described using the object oriented paradigm. For doing so, the Digital Signal Processing Object Oriented Metamodel offers a classification of signal processing objects in terms of their role in a DSP system. Objects are classified into two main categories: objects that process and objects that act as data containers. This 00 metamodel turns out to be closely related to Dataflow Process Networks, a graphical model of computation that has already proven useful for modeling signal processing systems. In our study we highlight the similarities of both models to conclude that object-orientation is in fact a superset of process-oriented models and therefore the object-oriented paradigm can be proposed as a general approach for system modeling. Furthermore, it turns out that nowadays the natural target for many signal processing applications is the computer and its software environment and the objectoriented paradigm becomes a natural conceptual framework where the different development phases fit. CLAM (C++ Library for Audio and Music) is a framework for developing music and audio applications that has been designed bearing this conceptual model in mind. CLAM is both the origin and the proof of concept of the Metamodel. On one hand its design process and rationale has led to the definition of the metamodel. On the other hand, it demonstrates that the metamodel proposed is more than an abstract wish-list and can be used to model working and efficient applications in the music and audio domain. The basic Object Oriented metamodel for signal processing systems can be extended to include the idea of Content Based Processing. 00 concepts like Inheritance Hierarchies, Polymorphism or Late Binding can be used to model run-time classification of media objects and to deal with the semantic information present in the signal rather than just treating the signal itself. This leads us to the definition of a new metamodel of information transmission that, unlike the traditional ones, does care about meaning. Finally, the 00 paradigm can also be used to model higher-level symbolic domains related to signal processing. For example, music (as a whole) can be effectively modeled using the 00 paradigm. An 00 model for music is proposed as an instance of the basic signal processing metamodel and the MetriX language is presented as its proof of concept.
Final publication