cl-wavelets is a library with a set of algorithms related to
various kinds of wavelet transform. Currently they all work with
arrays of type (simple-array (signed-byte 32) (*)) (i.e. they are
1D). This makes this library suitable for audio processing and
compression. For more info visit the project page
here.
Currently supported algorithms:
- DWT
- Frequency analysis using PWT.
- Best-basis PWT.
Currently supported wavelets:
- Haar wavelet
- CDF (2,2) wavelet
- CDF (3,1) wavelet
- CDF (4,2) wavelet
In examples:
(wavelets:dwt (make-array 8
:element-type '(signed-byte 32)
:initial-contents '(0 1 2 3 4 5 7 8))
:wavelet :cdf-2-2
:boundary-style :mirror)
#(2 4 0 3 0 0 0 1)
(wavelets:dwt-inverse *
:wavelet :cdf-2-2
:boundary-style :mirror)
#(0 1 2 3 4 5 7 8)
Generally, there are two kinds of functions: with ! at the end and without
!. Whose with ! are in-place functions, in other words they modify their
first argument. Whose without ! do not modify their first argument.
For more info, generate a documentation with codex like so:
codex:document :cl-wavelets :skip-unsocumented t.
You can load cl-wavelets/examples system which contains packages to
demonstrate some components of this library. For example, you can
build a spectrogram of an uncompressed WAV file making use of
wavelets:frequency-domain function. To build a spectrogram, try this:
(wavelets-spectrogram:spectrogram "/path/to/audio.wav"
"/path/to/spectrogram.jpg")
This will produce spectrogram.jpg image with the spectrogram. Note,
that the time axis is the vertical one, with the time going
up-down. This is a spectrogram rotated by 90 degrees clockwise,
actually. Also, the quality of the spectrogram will be much worse
compared to the qualily of a spectrogram obtained via FFT. This is
because the filters used in the process are far from ideal.