Descriptors

Descriptors are a special kind of ProcessingData that describe numerical attributes computed from the data in a related Processing Data object. This computation is accomplished using 'basic' statistical computations such as mean or nth order moments. For computing these statistics efficiently, CLAM offers a statistical computation module that uses functor objects with memory and template metaprogramming techniques.

Descriptor classes are storage classes where the values of these descriptors are kept. They make extensive use of the XML storage facilities in CLAM as a basic functionality of descriptors is storing them in an appropriate way for later using information retrieval techniques. Furthermore every descriptor class has an associated extractor where the actual value computations are implemented as combination of basic statistics.

CLAM offers low-level Descriptors and associated extractors for Audio, Spectrum, Spectral Peaks, Frame and Segment. This latter processing data has special descriptors as it includes temporal statistics such as mean and deviation for other descriptors computed on a frame base.

This is a list of the low-level descriptors currently implemented in the framework:

Apart from these low-level descriptors the same infrastructure is used for higher-level descriptors such as melodic or rhythmic description.

2004-10-18