Skip to content

Rename and improve freq parameter in the edc module #152

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
h-chmeruk wants to merge 11 commits into
base: develop
Choose a base branch
from
Open

Rename and improve freq parameter in the edc module #152

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
11 commits
Select commit Hold shift + click to select a range
199a54a
Renamed `freq` to `smoothing_parameter`, extended the functions to ac…
h-chmeruk Mar 13, 2026
ce827cd
Update the docstring for the `smooth_block_length` parameter in _smoo…
h-chmeruk Mar 13, 2026
d3792fc
Fixed the function logic when the RIR data has more than two dimensions
h-chmeruk Mar 20, 2026
dfee484
Improved the `intersection_time_lundeby` function according to the su…
h-chmeruk Mar 20, 2026
d035410
Fixed a ruff error
h-chmeruk Mar 20, 2026
c4769ff
Adjusted the code to the bugfix in the PR #170
h-chmeruk May 5, 2026
3d7a115
Fixed ruff errors
h-chmeruk May 5, 2026
0f81897
Fixed the implementation for cases where the signal has a multidimens…
h-chmeruk May 15, 2026
51b0c60
Updated the docstring according to the suggestions in the review by @…
h-chmeruk May 15, 2026
c1dd764
Added deprecation warnings and deprecation tests, updated the HISTORY…
h-chmeruk May 19, 2026
e379746
Fixed the ruff error, applied the suggestion by @hoyer-a
h-chmeruk May 19, 2026
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
8 changes: 8 additions & 0 deletions HISTORY.rst
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -2,6 +2,14 @@
History
=======

1.1.0 (2026-05-19)
------------------

Changed:
^^^^^^^^
- Replaced the `freq` parameter with `smoothing_parameter` in the edc functions, added deprecation warnings and deprecation tests (PR #152)


1.0.1 (2026-05-05)
------------------

Expand Down
17 changes: 10 additions & 7 deletions examples/energy_decay_curves_and_reverberation_time.ipynb
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -190,10 +190,13 @@
" The room impulse response with dimension [..., n_samples]\n",
"sampling_rate: integer\n",
" The sampling rate of the room impulse response.\n",
"freq: integer OR string\n",
" The frequency band. If set to 'broadband',\n",
" the time window of the Lundeby-algorithm will not be set in dependence\n",
" of frequency.\n",
"smoothing_parameter : int or array_like of int or {'broadband'}\n",
" Used to determine the smoothing time window in the Lundeby\n",
" algorithm. It should represent the center frequency (in Hz) of the\n",
" frequency band(s) in which the RIR data was computed.\n",
" If set to 'broadband', the smoothing time window will not be set\n",
" in dependence of frequency and a fixed time window of 30 ms\n",
" is used.\n",
"noise_level: ndarray, double OR string\n",
" If not specified, the noise level is calculated based on the last 10\n",
" percent of the RIR. Otherwise specify manually for each channel\n",
Expand Down Expand Up @@ -276,7 +279,7 @@
"source": [
"intersection_time, late_reveberation_time, noise_level_lundeby = \\\n",
" ra.edc.intersection_time_lundeby(\n",
" rir_1_noise, freq='broadband', is_energy=False,\n",
" rir_1_noise, smoothing_parameter='broadband', is_energy=False,\n",
" time_shift=True, channel_independent=False, plot=True)\n",
"\n",
"output_string = \"The estimated intersection time is {}s, the late reverberation time is {}s and the estimated noise is {}dB.\".format(intersection_time, late_reveberation_time, 10*np.log10(np.abs(noise_level_lundeby)))\n",
Expand All @@ -297,7 +300,7 @@
"outputs": [],
"source": [
"edc_truncation = ra.edc.energy_decay_curve_truncation(\n",
" rir_1_noise, freq='broadband', is_energy=False, time_shift=True,\n",
" rir_1_noise, smoothing_parameter='broadband', is_energy=False, time_shift=True,\n",
" channel_independent=False, normalize=True, plot=False)\n",
"\n",
"plt.figure(figsize=(10, 4))\n",
Expand Down Expand Up @@ -422,7 +425,7 @@
"outputs": [],
"source": [
"edc_chu_lundeby = ra.edc.energy_decay_curve_chu_lundeby(\n",
" rir_1_noise, freq='broadband', is_energy=False, time_shift=True,\n",
" rir_1_noise, smoothing_parameter='broadband', is_energy=False, time_shift=True,\n",
" channel_independent=False, normalize=True, plot=False)\n",
"\n",
"plt.figure(figsize=(10, 4))\n",
Expand Down
40 changes: 28 additions & 12 deletions pyrato/dsp.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -142,7 +142,7 @@ def _smooth_rir(
The room impulse response with dimension ``(..., n_samples)``.
sampling_rate: integer
Defines the sampling rate of the room impulse response.
smooth_block_length : double
smooth_block_length : double or array-like of double
Defines the block-length of the smoothing algorithm in seconds.

Returns
Expand All @@ -156,26 +156,42 @@ def _smooth_rir(

"""
cshape = data.shape[:-1]
data = np.atleast_2d(data)
n_samples = data.shape[-1]
data = data.reshape(-1, n_samples)
smooth_block_length = (np.atleast_1d(smooth_block_length)).flatten()
n_channels = len(smooth_block_length)
n_samples_nan = np.count_nonzero(np.isnan(data), axis=-1)

n_samples_per_block = int(np.round(smooth_block_length * sampling_rate, 0))
n_samples_per_block = (np.round(
smooth_block_length * sampling_rate, 0)).astype(int)
n_blocks = np.asarray(
np.floor((n_samples-n_samples_nan)/n_samples_per_block),
dtype=int)

# average data in blocks
n_blocks_min = int(np.min(n_blocks))
n_samples_actual = int(n_blocks_min*n_samples_per_block)
reshaped_array = np.reshape(
data[..., :n_samples_actual],
(*cshape, n_blocks_min, n_samples_per_block))
time_window_data = np.mean(reshaped_array, axis=-1)

n_blocks_min = (np.min(n_blocks)).astype(int)
n_samples_actual = (n_blocks_min*n_samples_per_block).astype(int)
time_window_data = np.empty((n_channels, n_blocks_min))
# Use average time instances corresponding to the average energy level
# instead of time for the first sample of the block
time_vector_window = \
if n_channels > 1:
for ch in range(n_channels):
reshaped_array = np.reshape(
data[ch, :n_samples_actual[ch]],
(n_blocks_min, n_samples_per_block[ch]))
time_window_data[ch, :] = np.mean(reshaped_array, axis=-1)
time_vector_window = (
(0.5+np.arange(0, n_blocks_min)).reshape(1, -1) * (
n_samples_per_block/sampling_rate).reshape(-1, 1))
time_window_data = np.reshape(
time_window_data, (cshape + (n_blocks_min,)))
time_vector_window = np.reshape(
time_vector_window, (cshape + (n_blocks_min,)))
else:
reshaped_array = np.reshape(
Comment on lines +176 to +190

@h-chmeruk h-chmeruk Mar 13, 2026

Copy link
Copy Markdown
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I'm not entirely satisfied with this cumbersome if statement, so I would appreciate a suggestion for improvement

data[..., :n_samples_actual[0]],
(-1, n_blocks_min, n_samples_per_block[0]))
time_window_data = np.mean(reshaped_array, axis=-1)
time_vector_window = \
((0.5+np.arange(0, n_blocks_min)) * n_samples_per_block/sampling_rate)

# Use the time corresponding to the sampling of the original data
Expand Down
111 changes: 80 additions & 31 deletions pyrato/edc.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -187,9 +187,13 @@ def _schroeder_integration(impulse_response, is_energy=False):
return energy_decay_curve


@pf._utils.rename_arg(
{"freq" : "smoothing_parameter"},
"'freq' will be deprecated in "
"pyrato 1.1.0 in favor of 'smoothing_parameter'")
def energy_decay_curve_truncation(
data,
freq='broadband',
smoothing_parameter='broadband',

@f-brinkmann f-brinkmann May 8, 2026

Copy link
Copy Markdown
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

We have to deprecate the freq parameter for backwards compatability. This can be done with the pyfar._utils.rename_arg decorator. You can search pyfar for example use cases of this :)

noise_level='auto',
is_energy=False,
time_shift=True,
Expand Down Expand Up @@ -218,10 +222,13 @@ def energy_decay_curve_truncation(
----------
data : pyfar.Signal
The room impulse response.
freq: integer OR string
The frequency band. If set to 'broadband',
the time window of the Lundeby-algorithm will not be set in dependence
of frequency.
smoothing_parameter : int or array_like of int or {'broadband'}
Used to determine the smoothing time window in the Lundeby
algorithm. It should represent the center frequency (in Hz) of the
frequency band(s) in which the RIR data was computed.
If set to 'broadband', the smoothing time window will not be set
in dependence of frequency and a fixed time window of 30 ms
is used.
noise_level: ndarray, double OR string
If not specified, the noise level is calculated based on the last 10
percent of the RIR. Otherwise specify manually for each channel
Expand Down Expand Up @@ -301,7 +308,7 @@ def energy_decay_curve_truncation(

intersection_time = intersection_time_lundeby(
energy_data,
freq=freq,
smoothing_parameter=smoothing_parameter,
initial_noise_power=noise_level,
is_energy=True,
time_shift=False,
Expand Down Expand Up @@ -350,9 +357,13 @@ def energy_decay_curve_truncation(
return edc


@pf._utils.rename_arg(
{"freq" : "smoothing_parameter"},
"'freq' will be deprecated in "
"pyrato 1.1.0 in favor of 'smoothing_parameter'")
def energy_decay_curve_lundeby(
data,
freq='broadband',
smoothing_parameter='broadband',
noise_level='auto',
is_energy=False,
time_shift=True,
Expand Down Expand Up @@ -381,10 +392,13 @@ def energy_decay_curve_lundeby(
----------
data : pyfar.Signal
The room impulse response.
freq: integer OR string
The frequency band. If set to 'broadband',
the time window of the Lundeby-algorithm will not be set in dependence
of frequency.
smoothing_parameter : int or array_like of int or {'broadband'}
Used to determine the smoothing time window in the Lundeby
algorithm. It should represent the center frequency (in Hz) of the
frequency band(s) in which the RIR data was computed.
If set to 'broadband', the smoothing time window will not be set
in dependence of frequency and a fixed time window of 30 ms
is used.
noise_level: ndarray, double OR string
If not specified, the noise level is calculated based on the last 10
percent of the RIR. Otherwise specify manually for each channel
Expand Down Expand Up @@ -462,7 +476,7 @@ def energy_decay_curve_lundeby(
intersection_time, late_reverberation_time, noise_estimation = \
intersection_time_lundeby(
energy_data,
freq=freq,
smoothing_parameter=smoothing_parameter,
initial_noise_power=noise_level,
is_energy=True,
time_shift=False,
Expand Down Expand Up @@ -653,9 +667,13 @@ def energy_decay_curve_chu(
return edc


@pf._utils.rename_arg(
{"freq" : "smoothing_parameter"},
"'freq' will be deprecated in "
"pyrato 1.1.0 in favor of 'smoothing_parameter'")
def energy_decay_curve_chu_lundeby(
data,
freq='broadband',
smoothing_parameter='broadband',
noise_level='auto',
is_energy=False,
time_shift=True,
Expand Down Expand Up @@ -686,10 +704,13 @@ def energy_decay_curve_chu_lundeby(
----------
data : pyfar.Signal
The room impulse response.
freq: integer OR string
The frequency band. If set to 'broadband',
the time window of the Lundeby-algorithm will not be set in dependence
of frequency.
smoothing_parameter : int or array_like of int or {'broadband'}
Used to determine the smoothing time window in the Lundeby
algorithm. It should represent the center frequency (in Hz) of the
frequency band(s) in which the RIR data was computed.
If set to 'broadband', the smoothing time window will not be set
in dependence of frequency and a fixed time window of 30 ms
is used.
noise_level: ndarray, double OR string
If not specified, the noise level is calculated based on the last 10
percent of the RIR. Otherwise specify manually for each channel
Expand Down Expand Up @@ -770,7 +791,7 @@ def energy_decay_curve_chu_lundeby(
intersection_time, late_reverberation_time, noise_level = \
intersection_time_lundeby(
energy_data,
freq=freq,
smoothing_parameter=smoothing_parameter,
initial_noise_power=noise_level,
is_energy=True,
time_shift=False,
Expand Down Expand Up @@ -824,9 +845,13 @@ def energy_decay_curve_chu_lundeby(
return edc


@pf._utils.rename_arg(
{"freq" : "smoothing_parameter"},
"'freq' will be deprecated in "
"pyrato 1.1.0 in favor of 'smoothing_parameter'")
def intersection_time_lundeby(
data,
freq='broadband',
smoothing_parameter='broadband',
initial_noise_power='auto',
is_energy=False,
time_shift=False,
Expand All @@ -843,10 +868,15 @@ def intersection_time_lundeby(
----------
data : pyfar.Signal
The room impulse response
freq: integer OR string
The frequency band. If set to 'broadband',
the time window of the Lundeby-algorithm will not be set in dependence
of frequency.
smoothing_parameter : scalar or array_like of scalar or {'broadband'}
Used to determine the smoothing time window in the Lundeby
algorithm. It should represent the center frequency (in Hz) of the
frequency band(s) in which the RIR data was computed.

@f-brinkmann f-brinkmann Mar 13, 2026

Copy link
Copy Markdown
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

It would be good to mention that length must match data.cshape[0] if it is an array like

If set to 'broadband', the smoothing time window will not be set
in dependence of frequency and a fixed time window of 30 ms
is used.
If set to an array_like, the length of the array must match
the number of frequency bands in the RIR data, i.e., data.cshape[0].
initial_noise_power: ndarray, double OR string
If ``'auto'``, the noise level is calculated based on the last 10
percent of the RIR. Otherwise specify manually for each channel
Expand Down Expand Up @@ -942,11 +972,30 @@ def intersection_time_lundeby(
sampling_rate = np.round(1/np.diff(data.times).mean(), decimals=4)
energy_data = energy_data.time

if freq == "broadband":
# number of frequency bands given by first channel axis
n_bands = data.cshape[0]
if smoothing_parameter == "broadband":
# broadband: use 30 ms windows sizes
freq_dependent_window_time = 0.03
freq_dependent_window_time = np.atleast_1d([0.03] * n_bands)
elif isinstance(smoothing_parameter, (int, float, list,
tuple, np.ndarray)):
smoothing_parameter = np.atleast_1d(smoothing_parameter)
if smoothing_parameter.size == 1:
smoothing_parameter = np.tile(smoothing_parameter, n_bands)
elif smoothing_parameter.size != n_bands:
raise ValueError(
"The size of smoothing_parameter must match the number of "
"frequency bands.")
freq_dependent_window_time = (800 / smoothing_parameter + 10) / 1000
else:
freq_dependent_window_time = (800/freq+10) / 1000
raise TypeError(
"smoothing_parameter must be an int or float, an array_like"
" of int or float, or 'broadband'")

freq_dependent_window_time = np.broadcast_to(
freq_dependent_window_time.reshape(
n_bands, *([1] * (len(data.cshape) - 1))),
data.cshape).copy()
Comment on lines +994 to +998

@hoyer-a hoyer-a May 20, 2026

Copy link
Copy Markdown
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Maybe a bit more readable.
data.cdim is equivalent to len(data.cshape) :)

Suggested change
freq_dependent_window_time = np.broadcast_to(
freq_dependent_window_time.reshape(
n_bands, *([1] * (len(data.cshape) - 1))),
data.cshape).copy()
shape = (n_bands,) + (1,) * (data.cdim -1)
freq_dependent_window_time = np.broadcast_to(
freq_dependent_window_time.reshape(shape),
data.cshape).copy()


# (1) SMOOTH
time_window_data, time_vector_window, time_vector = dsp._smooth_rir(
Expand All @@ -965,12 +1014,12 @@ def intersection_time_lundeby(
noise_peak_level = np.zeros(data.cshape, data.time.dtype)

for ch in np.ndindex(data.cshape):

output = _intersection_time_lundby(
time_window_data[ch], noise_estimation[ch], energy_data[ch],
time_vector_window, dB_above_noise, n_intervals_per_10dB,
use_dyn_range_for_regression, sampling_rate, ch, failure_policy)

time_window_data[ch], noise_estimation[ch], energy_data[ch],
np.squeeze(np.atleast_2d(time_vector_window)[ch]),
dB_above_noise, n_intervals_per_10dB,
use_dyn_range_for_regression, sampling_rate,
ch, failure_policy)
Comment on lines +1018 to +1022

@hoyer-a hoyer-a May 20, 2026

Copy link
Copy Markdown
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

looks like indentation is missing here

@mberz mberz May 21, 2026

Copy link
Copy Markdown
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Please do not comment on code that is not changed in the PR. Instead, please create an issue instead.

if output is None:
reverberation_time[ch] = np.nan
noise_level[ch] = np.nan
Expand Down
16 changes: 8 additions & 8 deletions tests/test_data/generate_test_data.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -100,24 +100,24 @@
substracted_2D = pyrato.edc.subtract_noise_from_squared_rir(rir_array**2)

edc_truncation_1D = pyrato.energy_decay_curve_truncation(
rir_array[0], freq='broadband', is_energy=False, time_shift=True,
rir_array[0], smoothing_parameter='broadband', is_energy=False, time_shift=True,
channel_independent=False, normalize=True)
edc_truncation_2D = pyrato.energy_decay_curve_truncation(
rir_array, freq='broadband', is_energy=False, time_shift=True,
rir_array, smoothing_parameter='broadband', is_energy=False, time_shift=True,
channel_independent=False, normalize=True)

edc_lundeby_1D = pyrato.energy_decay_curve_lundeby(
rir_array[0], freq='broadband', is_energy=False, time_shift=True,
rir_array[0], smoothing_parameter='broadband', is_energy=False, time_shift=True,
channel_independent=False, normalize=True, plot=False)
edc_lundeby_2D = pyrato.energy_decay_curve_lundeby(
rir_array, freq='broadband', is_energy=False, time_shift=True,
rir_array, smoothing_parameter='broadband', is_energy=False, time_shift=True,
channel_independent=False, normalize=True, plot=False)

edc_lundeby_chu_1D = pyrato.energy_decay_curve_chu_lundeby(
rir_array[0], freq='broadband', is_energy=False, time_shift=True,
rir_array[0], smoothing_parameter='broadband', is_energy=False, time_shift=True,
channel_independent=False, normalize=True, plot=False)
edc_lundeby_chu_2D = pyrato.energy_decay_curve_chu_lundeby(
rir_array, freq='broadband', is_energy=False, time_shift=True,
rir_array, smoothing_parameter='broadband', is_energy=False, time_shift=True,
channel_independent=False, normalize=True, plot=False)

edc_chu_1D = pyrato.energy_decay_curve_chu(
Expand All @@ -128,10 +128,10 @@
channel_independent=False, normalize=True, plot=False)

intersection_time_1D = pyrato.intersection_time_lundeby(
rir_array[0], freq='broadband', is_energy=False, time_shift=True,
rir_array[0], smoothing_parameter='broadband', is_energy=False, time_shift=True,
channel_independent=False, plot=False)
intersection_time_2D = pyrato.intersection_time_lundeby(
rir_array, freq='broadband', is_energy=False, time_shift=True,
rir_array, smoothing_parameter='broadband', is_energy=False, time_shift=True,
channel_independent=False, plot=False)

# noise_energy_from_edc_1D = pyrato.edc.estimate_noise_energy_from_edc(
Expand Down
Loading