| Age | Commit message (Collapse) | Author |
|---|
|
think the solution is to wrap the globals in a struct, declare it in a header, and wrap with a definition guard, then define _once_ at library init time. (Sounds like a recipe for something...)
|
|
|
|
only filled in freq_bands+1 values, when it should fill in freq_bands+2
|
|
|
|
xtract_init_fft() for creating plans, which have global scope. Updated examples to reflect the change. New configure option: --with-fft_optimisation (0 = FFTW_ESTIMATE, 1 = FFTW_MEASURE, 2 = FFTW_PATIENT).
|
|
|
|
zeroing a load of elements despite the fact that they're ignored by the DCT process called next, and never used for anything. This was writing to an assumed large result array (same size as number of FFT bins) despite the fact that only a small number of MFCCs (typically less than 50) are required, therefore either wasting memory or writing to memory it shouldn't do!
|
|
|
|
|
|
|
|
bindings tests.
|
|
|
|
Linux build (sigh)).
|
|
|
|
compile. Some kind of path error, but I can't work out how to fix it.
|
|
|
|
anyone can't make it work with their compiler
|
|
functions: powf, roundf etc). Added a few PD examples/tests.
|
|
|
|
|
|
|
|
specific architectures
|
|
|
|
|
|
they don't get a samplerate 44100 is assumed (I'm not sure if this is a good idea!).
|
|
the basic code is there
|
|
|
|
|
|
|
|
|
|
|
|
|
|
nyquist/DC inclusion. Added new boolean argument 'withDC' to select whether the DC component is required in the output
|
|
the same in theory, but saves on one assignment operation. But the real reason I did it was because my compiler (gcc 4.0.1, Apple intel edition) was getting it wrong (presumably during some optimisation). It does save on the assignment though.
|
|
|
|
ChangeLog and AUTHORS files.
|
|
return a value.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|