The system just presented should be understood as a proof of concept. It has several limitations that make it unpractical. Just to name a few:
As already mentioned the receiver holds a particular model of a sax/trumpet synthesizer. In its current implementation the model cannot be modified or configured, just controlled. Nevertheless, as we will show in the next chapter, it is easy to foresee a solution to this issue by allowing the model to transmit complete instrument descriptions such as those allowed by the MetriX Instrument Definition Language.
On the other hand, the encoded and transmitted melody has severe limitations. To start with, the instrument that has produced it is not encoded into the stream. Encoding the exact instrument that has produced the melody would mean having a robust and precise instrument classification algorithm. Although many people, including some from our own group (see [Herrera et al., 2000]), are working on it but this is right now far from solved.
Another limitation of the encoded melody is its limited information. As a matter of fact the only data being transmitted is the note boundaries, pitch and energy. This is in fact the same information contained in a MIDI stream. In order to have more realistic performances other information such as the energy envelope, ADSR curve or vibrato characteristics should be encoded. Most of these descriptors though can be easily computed.
Finally, and also on the transmitter side, there is a great limitation on the content analysis component. The algorithms used can only work on monophonic melodic phrases. Other more complex signals should be analyzed using other techniques such as PCA or multipitch, which are still not robust and precise enough.
Nevertheless, and regardless of all these limitations, we believe that the presented sample system already highlights many of the opportunities of the Object-Oriented Content Transmission Metamodel.
To start with, any of the components of the system, though maybe limited, is already useful by itself. We have a content analyzer, a content description XML encoding tool, a decoder that interprets XML and translates it into MIDI or information understandable by the user and a content-based synthesizer.
But the most important novelty is the use of coherent and compatible formats all through the chain. XML, or any other compatible metadata format, can be used in the analysis, transmission and synthesis step, opening up a whole range of possible future applications of the metamodel.