Calculates time-frequency decomposition; not intended for direct use. Please use 'RAVE' pipelines (see 'Examples').
Usage
run_wavelet(
subject,
electrodes,
freqs,
cycles,
target_sample_rate = 100,
kernels_precision = "float",
pre_downsample = 1,
verbose = TRUE
)Arguments
- subject
'RAVE' subject or subject ID
- electrodes
electrode channels to apply, must be imported and
'LFP'type- freqs
numeric vector of frequencies to apply
- cycles
number of wavelet cycles at each
freqs, integers- target_sample_rate
the resulting 'spectrogram' sampling frequency
- kernels_precision
double or single (default) floating precision
- pre_downsample
down-sample (integer) priory to the decomposition; set to 1 (default) to avoid
- verbose
whether to verbose the progress
Value
The decomposition results are stored in 'RAVE' subject data path; the function only returns the wavelet parameters.
Details
The channel signals are first down-sampled (optional) by a ratio of
pre_downsample via a 'FIR' filter. After the down-sample,
'Morlet' wavelet kernels are applied to the signals to calculate the
wavelet coefficients (complex number) at each frequency in freqs.
The number of cycles at each frequency controls the number of
sine and cosine waves, allowing users to balance the time and power
accuracy. After the decomposition, the 'spectrogram' is further down-sampled
to target_sample_rate, assuming the brain power is a smooth function
over time. This down-sample is done via time-point sampling to preserve the
phase information (so the linear functions such as common-average or
bi-polar reference can be carried over to the complex coefficients).
Examples
# Check https://rave.wiki for additional pipeline installation
if (FALSE) { # \dontrun{
# ---- Recommended usage --------------------------------------------
pipeline <- ravepipeline::pipeline("wavelet_module")
pipeline$set_settings(
project_name = "demo",
subject_code = "DemoSubject",
precision = "float",
pre_downsample = 4,
kernel_table = ravetools::wavelet_cycles_suggest(
freqs = seq(1, 200, by = 1)),
target_sample_rate = 100
)
# Internally, the above pipeline includes this function call below
# ---- For demonstration use, do not call this function directly ----
# Original sample rate: 2000 Hz
# Downsample by 4 to 500 Hz first - 250 Hz Nyquist
# Wavelet at each 1, 2, ..., 200 Hz
# The number of cycles log-linear from 2 to 20
# The wavelet coefficient sample rate is 500 Hz
# Further down-sample to 100 Hz to save storage space
run_wavelet(
subject = "demo/DemoSubject",
electrodes = c(13:16, 2),
pre_downsample = 4,
freqs = seq(1, 200, by = 1),
cycles = c(2, 20),
target_sample_rate = 100
)
} # }