A UGen for extracting mel frequency cepstral coefficients. It generates a set of MFCCs; these are obtained from a band-based frequency representation (using the Mel scale by default), and then a discrete cosine transform (DCT). The DCT is an efficient approximation for principal components analysis, so that it allows a compression, or reduction of dimensionality, of the data, in this case reducing 42 band readings to a smaller set of MFCCs. A small number of features (the coefficients) end up describing the spectrum. The MFCCs are commonly used as timbral descriptors.
The output values are somewhat normalised for the range 0.0 to 1.0, but there are no guarantees on exact conformance to this. Commonly, the first coefficient will be the highest value. The number of output channels corresponds to the number of coefficients specified. Technical note: The 0th coefficient is not generated as it consists of multiplying all bands by 1 and summing
Note that despite running at control-rate, the output remains constant for each FFT frame. E.g. with an FFT size of 1024 and 50% overlap, a new measure is generated every 512 audio frames, or (at control block size 64) every 8 control blocks.
- Value Params
- chain
the output (buffer) of an FFT UGen which transforms the audio input to track. For the FFT chain, with a standard hop of half FFT size, the FFT size should be 1024 at 44.1 or 48 kHz and 2048 at 88.2 or 96 kHz sampling rate.
- numCoeffs
the number of coefficients, defaults to 13, maximum of 42; more efficient to use less of course! Since this number determines the number of output channels of the UGen, it has to be an
Int.
- Companion
- object