Function FILTER_BANK
Package
filters
Short description
Create a cell array of filter banks
Usage
Input
N (int): The size of the input data.
options (struct): Filter parameters, see below.
Output
filters (struct): A cell array of filter banks corresponding to the
data size N and the filter parameters in options.
Description
The behavior of the function depends on the value of options.filter_type,
which can have the following values:
'morlet_1d', 'gabor_1d': Calls MORLET_FILTER_BANK_1D.
'spline_1d': Calls SPLINE_FILTER_BANK_1D.
'selesnick_1d': Calls SELESNICK_FILTER_BANK_1D.
The filter parameters in options are then passed on to these functions.
If multiple filter banks are desired, multiple parameters can be supplied
by providing a vector of parameter values instead of a scalar (in the
case of filter_type and filter_format, these have to be cell arrays).
The function will then split the options structure up into several
parameter sets and call the appropriate filter bank function for each one
returning the result as a cell array. If a given field does not have
enough values to specify parameters for all filter banks, the last ele-
ment is used to extend the field as needed.
The specific parameters vary between filter bank types, but the follow-
ing are common to all types:
options.filter_type: Can be 'morlet_1d', 'gabor_1d', 'spline_1d',
or 'selesnick_1d' (default 'morlet_1d').
options.boundary: Sets the boundary conditions of the wavelet trans-
form. If 'symm', symmetric boundaries will be used, if 'per', per-
iodic boundaries will be used (default 'symm').
options.precision: Either 'double', or 'single'. Determines the preci-
sion of the filters stored and consequently that of the resulting
wavelet and scattering transform (default 'double').
options.filter_format: Specifies the format in which the filters are
stored. Three formats are available:
'fourier': Filters are stored as Fourier transforms defined
over the whole frequency domain of the signal.
'fourier_multires': Filters are stored as Fourier transforms
defined over the frequency domain of the signal at all
resolutions. This requires much more memory, but speeds up
calculations significantly.
'fourier_truncated': Stores the Fourier coefficients of each
filter on the support of the filter, reducing memory
consumption and increasing speed of calculations (default).
For parameters specific to the filter bank type, see the documentation
for MORLET_FILTER_BANK_1D, SPLINE_FILTER_BANK_1D, and
SELESNICK_FILTER_BANK_1D.
See also
List of all packages