HRTFTransform¶

class hrtfpykit.datasets.HRTFTransform¶

Factory namespace for reusable HRTF preprocessing callables.

HRTFTransform creates small callables that receive a subject HRTF, run one HRTF operation, and return the transformed object. These callables can be used as dataset-level transforms through dataset_hrtf_transform or as HRTF-level spec transforms, for example through HRTFSpec.

The factory methods are thin adapters over transform and select(). They do not load files, mutate dataset state, or change the original SOFA resources. They only package transform arguments so the same preprocessing can be applied consistently to every loaded subject.

Notes

When used as dataset_hrtf_transform, the returned callables are executed after a subject HRTF is read and before it is cached. When used on an HRTF-level spec, they are executed before that spec extracts its value. If the callable returns an object that does not behave like an HRTF, dataset loading or value extraction raises an error before sample extraction continues.

static build(method_name, *args, **kwargs)¶

Create an HRTF-level callable from a transform method name.

This is the generic factory used by the named convenience methods below. It stores the requested transform method name and arguments, then delays method lookup until a real HRTF object is loaded. When executed, the returned callable looks up method_name on transform, calls it with the stored arguments, and returns the transformed HRTF.

Parameters:

method_name (str) – Name of the method available through transform.
*args (object) – Positional arguments forwarded to the transform method.
**kwargs (object) – Keyword arguments forwarded to the transform method.

Returns:

Callable that accepts an HRTF object and returns the transformed HRTF object.

Return type:

callable

Raises:

TypeError – Raised by the returned callable if the supplied object does not expose a transform attribute.
AttributeError – Raised by the returned callable if method_name is not available or not callable on the object’s transform namespace.

Notes

The factory marks the returned callable with __hrtf_transform__ so callers can distinguish HRTF transform wrappers from arbitrary user callables when introspection is needed.

Examples

>>> from hrtfpykit.datasets import HUTUBS
>>> from hrtfpykit.datasets import HRTFSpec
>>> from hrtfpykit.datasets import HRTFTransform
>>> transform = HRTFTransform.build("apply_padding", 16, location="end")
>>> dataset = HUTUBS(
...     root="datasets/hutubs",
...     inputs=HRTFSpec(),
...     dataset_hrtf_transform=transform,
... )

static select(*args, **kwargs)¶

Create an HRTF-level source selection callable.

Selection is special because it calls select() directly rather than a method under transform. The returned callable lets a dataset reduce or reorder source positions before specs extract acoustic values. The returned callable can be used through dataset_hrtf_transform or through HRTF-level spec transform hooks.

Parameters:

*args (object) – Positional arguments forwarded to select().
**kwargs (object) – Keyword arguments forwarded to select().

Returns:

Callable that accepts an HRTF object and returns the selected HRTF.

Return type:

callable

Raises:

AttributeError – Raised by the returned callable if the supplied object does not expose a callable select()-compatible method.

Notes

Use this factory when the training or evaluation dataset should expose a source subset, for example a fixed plane or a small set of named positions, without rewriting the source SOFA files.

Examples

>>> from hrtfpykit.datasets import HUTUBS
>>> from hrtfpykit.datasets import HRTFTransform
>>> dataset = HUTUBS(
...     root="datasets/hutubs",
...     dataset_hrtf_transform=HRTFTransform.select(positions=[0, 1]),
... )

static apply_window(window_name, start_sample=None, end_sample=None, ear='both')¶

Create a callable that applies a named window to each subject HRIR.

The returned callable forwards window_name, sample bounds, and ear to apply_window(). The transform operates on time-domain IR values, applies the window along the selected interval of the final sample axis, and refreshes the TF representation before dataset specs read values.

Parameters:

window_name (str) – Window identifier forwarded to the HRTF transform layer, for example hann, hamming, blackman, or rectangular.
start_sample (int or None, default=None) – First sample included in the windowed interval. None starts at sample 0.
end_sample (int or None, default=None) – First sample after the windowed interval. None uses the full IR length.
ear ({both, left, right}, default=``both``) – Ear channel that receives the window.

Returns:

Dataset-level transform callable accepting and returning an HRTF object.

Return type:

callable

Examples

>>> from hrtfpykit.datasets import HUTUBS
>>> from hrtfpykit.datasets import HRTFSpec
>>> from hrtfpykit.datasets import HRTFTransform
>>> transform = HRTFTransform.apply_window(
...     "hann",
...     start_sample=0,
...     end_sample=128,
...     ear="left",
... )
>>> dataset = HUTUBS(
...     root="datasets/hutubs",
...     inputs=HRTFSpec(),
...     dataset_hrtf_transform=transform,
... )

static apply_crop(start_sample=None, end_sample=None, window=None, ear='both')¶

Create a callable that crops each subject HRIR.

The returned callable forwards start_sample, end_sample, window, and ear to apply_crop(). The transform removes the selected IR sample interval, shifts later samples left, appends zeros to preserve the original IR length, and rebuilds TF values before dataset specs read acoustic data.

Parameters:

start_sample (int or None, default=None) – First sample removed from each impulse response. None starts at sample 0.
end_sample (int or None, default=None) – First sample after the removed interval. This value is required. The interval is Python-style and excludes end_sample.
window (str or None, default=None) – Optional window name used to taper the final retained samples before the trailing zero padding. None skips tapering.
ear ({both, left, right}, default=``both``) – Ear channel that receives the crop.

Returns:

Dataset-level transform callable accepting and returning an HRTF object.

Return type:

callable

Examples

>>> from hrtfpykit.datasets import HUTUBS
>>> from hrtfpykit.datasets import HRTFSpec
>>> from hrtfpykit.datasets import HRTFTransform
>>> transform = HRTFTransform.apply_crop(64, 128, window="hann", ear="left")
>>> dataset = HUTUBS(
...     root="datasets/hutubs",
...     inputs=HRTFSpec(),
...     dataset_hrtf_transform=transform,
... )

static apply_padding(padding_length, location='end', value=0, preserve_length=False, ear='both')¶

Create a callable that pads each subject HRIR.

The returned callable forwards padding_length, location, value, preserve_length, and ear to apply_padding(). The transform pads the final IR sample axis and rebuilds TF values with the active FFT length before sample extraction.

Parameters:

padding_length (int) – Number of samples added to each impulse response.
location ({start, end}, default=``end``) – Side of the IR sample axis where padding is applied.
value (float, default=0) – Constant value used in the selected padded region.
preserve_length (bool, default=False) – If True, only start padding is accepted and the transformed HRIR keeps its original sample length.
ear ({both, left, right}, default=``both``) – Ear channel that receives the requested padding.

Returns:

Dataset-level transform callable accepting and returning an HRTF object.

Return type:

callable

Examples

>>> from hrtfpykit.datasets import HUTUBS
>>> from hrtfpykit.datasets import HRTFSpec
>>> from hrtfpykit.datasets import HRTFTransform
>>> transform = HRTFTransform.apply_padding(
...     16,
...     location="start",
...     preserve_length=True,
...     ear="left",
... )
>>> dataset = HUTUBS(
...     root="datasets/hutubs",
...     inputs=HRTFSpec(),
...     dataset_hrtf_transform=transform,
... )

static upsampling(new_sample_rate)¶

Create a callable that upsamples each subject HRIR.

The returned callable forwards new_sample_rate to upsampling(). The transform resamples IR values to the target rate, updates IR sample-rate metadata, and recomputes TF values and frequency bins from the resampled IR.

Parameters:: new_sample_rate (float) – Target sample rate in hertz. It must be greater than the current subject IR sample rate.
Returns:: Dataset-level transform callable accepting and returning an HRTF object.
Return type:: callable

Notes

Dataset-level upsampling keeps all loaded subjects in the same transformed sampling context before HRTF, ITD, ILD, SH, or sample-indexed specs resolve their values.

Examples

>>> from hrtfpykit.datasets import HUTUBS
>>> from hrtfpykit.datasets import HRTFSpec
>>> from hrtfpykit.datasets import HRTFTransform
>>> transform = HRTFTransform.upsampling(96000.0)
>>> dataset = HUTUBS(
...     root="datasets/hutubs",
...     inputs=HRTFSpec(),
...     dataset_hrtf_transform=transform,
... )

static downsampling(new_sample_rate)¶

Create a callable that downsamples each subject HRIR.

The returned callable forwards new_sample_rate to downsampling(). The transform resamples IR values to the target rate, updates IR sample-rate metadata, and recomputes TF values and frequency bins from the resampled IR.

Parameters:: new_sample_rate (float) – Target sample rate in hertz. It must be lower than the current subject IR sample rate.
Returns:: Dataset-level transform callable accepting and returning an HRTF object.
Return type:: callable

Notes

Downsampling is intended for workflows that need a lower acoustic sampling rate than the original SOFA files while keeping the dataset interface unchanged.

Examples

>>> from hrtfpykit.datasets import HUTUBS
>>> from hrtfpykit.datasets import HRTFSpec
>>> from hrtfpykit.datasets import HRTFTransform
>>> transform = HRTFTransform.downsampling(44100.0)
>>> dataset = HUTUBS(
...     root="datasets/hutubs",
...     inputs=HRTFSpec(),
...     dataset_hrtf_transform=transform,
... )

static apply_fir_filter(filter, cutoff=None, num_taps=101, window=None, ear='both')¶

Create a callable that applies FIR filtering to each subject HRIR.

The returned callable forwards filter, cutoff, num_taps, window, and ear to apply_fir_filter(). Filtering is performed in the time domain, then TF values are rebuilt before specs read the loaded subject.

Parameters:

filter (str) – FIR filter type accepted by the HRTF transform layer, such as lowpass, highpass, or bandpass.
cutoff (float | tuple[float, float] | None, default=None) – Cutoff frequency or frequency pair in hertz.
num_taps (int, default=101) – FIR tap count used to design the filter.
window (str | None, default=None) – Optional FIR design window.
ear ({both, left, right}, default=``both``) – Ear channel that receives the FIR filter.

Returns:

Dataset-level transform callable accepting and returning an HRTF object.

Return type:

callable

Examples

>>> from hrtfpykit.datasets import HUTUBS
>>> from hrtfpykit.datasets import HRTFSpec
>>> from hrtfpykit.datasets import HRTFTransform
>>> transform = HRTFTransform.apply_fir_filter(
...     "lowpass",
...     cutoff=8000.0,
...     num_taps=101,
...     ear="right",
... )
>>> dataset = HUTUBS(
...     root="datasets/hutubs",
...     inputs=HRTFSpec(),
...     dataset_hrtf_transform=transform,
... )

static apply_iir_filter(filter, cutoff=None, order=10, ear='both')¶

Create a callable that applies IIR filtering to each subject HRIR.

The returned callable forwards filter, cutoff, order, and ear to apply_iir_filter(). Filtering is performed in the time domain, then TF values are rebuilt before specs read the loaded subject.

Parameters:

filter (str) – IIR filter type accepted by the HRTF transform layer, such as lowpass, highpass, or bandpass.
cutoff (float | tuple[float, float] | None, default=None) – Cutoff frequency or frequency pair in hertz.
order (int, default=10) – Butterworth filter order.
ear ({both, left, right}, default=``both``) – Ear channel that receives the IIR filter.

Returns:

Dataset-level transform callable accepting and returning an HRTF object.

Return type:

callable

Examples

>>> from hrtfpykit.datasets import HUTUBS
>>> from hrtfpykit.datasets import HRTFSpec
>>> from hrtfpykit.datasets import HRTFTransform
>>> transform = HRTFTransform.apply_iir_filter(
...     "highpass",
...     cutoff=200.0,
...     order=4,
...     ear="left",
... )
>>> dataset = HUTUBS(
...     root="datasets/hutubs",
...     inputs=HRTFSpec(),
...     dataset_hrtf_transform=transform,
... )

static minimum_phase(method='homomorphic', fft_length=None, epsilon=1e-12)¶

Create a callable that converts each subject HRIR to minimum phase.

The returned callable forwards method, fft_length, and epsilon to minimum_phase(). The transform replaces each HRIR with a minimum-phase version derived from its magnitude response, then rebuilds TF values for the loaded subject.

Parameters:

method (str, default=``homomorphic``) – Minimum-phase reconstruction method passed to the HRTF transform layer.
fft_length (int | None, default=None) – Optional FFT length used during minimum-phase reconstruction.
epsilon (float, default=1e-12) – Positive numerical floor used by the reconstruction routine.

Returns:

Dataset-level transform callable accepting and returning an HRTF object.

Return type:

callable

Notes

Minimum-phase conversion is a dataset-side preprocessing step only. It lets a dataset expose minimum-phase acoustics without modifying the original SOFA files.

Examples

>>> from hrtfpykit.datasets import HUTUBS
>>> from hrtfpykit.datasets import HRTFSpec
>>> from hrtfpykit.datasets import HRTFTransform
>>> transform = HRTFTransform.minimum_phase(method="homomorphic")
>>> dataset = HUTUBS(
...     root="datasets/hutubs",
...     inputs=HRTFSpec(),
...     dataset_hrtf_transform=transform,
... )

static to_ctf(weights=False, magnitude_average='log', attenuation=None)¶

Create a callable that converts each subject HRTF to CTF form.

The returned callable forwards weights, magnitude_average, and attenuation to to_ctf(). The transform collapses the source axis into a singleton common transfer function per ear before dataset specs consume the subject HRTF.

Parameters:

weights (bool, default=False) – Whether source-position weights are used when estimating the common transfer function.
magnitude_average ({log, linear}, default=``log``) – Magnitude averaging rule used by the CTF transform.
attenuation (float | None, default=None) – Optional attenuation in decibels applied by the CTF transform.

Returns:

Dataset-level transform callable accepting and returning an HRTF object.

Return type:

callable

Notes

The CTF output has a singleton source axis. Dataset specs that index by position should be configured with that changed source layout in mind.

Examples

>>> from hrtfpykit.datasets import HUTUBS
>>> from hrtfpykit.datasets import HRTFSpec
>>> from hrtfpykit.datasets import HRTFTransform
>>> transform = HRTFTransform.to_ctf(weights=False, magnitude_average="log")
>>> dataset = HUTUBS(
...     root="datasets/hutubs",
...     inputs=HRTFSpec(),
...     dataset_hrtf_transform=transform,
... )

static to_dtf(weights=False, magnitude_average='log', attenuation=None)¶

Create a callable that converts each subject HRTF to DTF form.

The returned callable forwards weights, magnitude_average, and attenuation to to_dtf(). The transform estimates and removes a common transfer component while preserving the subject source layout.

Parameters:

weights (bool, default=False) – Whether source-position weights are used when estimating the common transfer function removed from the HRTF.
magnitude_average ({log, linear}, default=``log``) – Magnitude averaging rule used during DTF calculation.
attenuation (float | None, default=None) – Optional attenuation in decibels applied by the DTF transform.

Returns:

Dataset-level transform callable accepting and returning an HRTF object.

Return type:

callable

Notes

DTF conversion is useful when learning or analysis should focus on direction-dependent spectral structure rather than the common ear response.

Examples

>>> from hrtfpykit.datasets import HUTUBS
>>> from hrtfpykit.datasets import HRTFSpec
>>> from hrtfpykit.datasets import HRTFTransform
>>> transform = HRTFTransform.to_dtf(weights=False, magnitude_average="log")
>>> dataset = HUTUBS(
...     root="datasets/hutubs",
...     inputs=HRTFSpec(),
...     dataset_hrtf_transform=transform,
... )

static modify_ir(new_ir)¶

Create a callable that replaces each subject IR data array.

The returned callable forwards new_ir to modify_ir(). Replacement data become the loaded subject’s time-domain values, and TF values are rebuilt before dataset specs extract outputs.

Parameters:: new_ir (np.ndarray | IR | HRTF) – Replacement time-domain data forwarded to modify_ir().
Returns:: Dataset-level transform callable accepting and returning an HRTF object.
Return type:: callable

Notes

The replacement leading shape must be compatible with each loaded subject’s source and ear layout. Use this for controlled experiments where dataset metadata and indexing come from real resources but acoustic arrays are supplied externally.

Examples

>>> import numpy as np
>>> from hrtfpykit.datasets import HUTUBS
>>> from hrtfpykit.datasets import HRTFSpec
>>> from hrtfpykit.datasets import HRTFTransform
>>> transform = HRTFTransform.modify_ir(np.zeros((440, 2, 256)))
>>> dataset = HUTUBS(
...     root="datasets/hutubs",
...     inputs=HRTFSpec(),
...     dataset_hrtf_transform=transform,
... )

static modify_phase(new_phase, unit='degrees')¶

Create a callable that replaces each subject TF phase.

The returned callable forwards new_phase and unit to modify_phase(). The transform preserves current TF magnitude, replaces phase values, and rebuilds IR values before sample extraction.

Parameters:

new_phase (np.ndarray) – Replacement phase array with the same TF layout expected by the HRTF transform layer.
unit ({degrees, radians}, default=``degrees``) – Angular unit used by new_phase.

Returns:

Dataset-level transform callable accepting and returning an HRTF object.

Return type:

callable

Notes

The replacement phase must be compatible with the TF layout of every loaded subject that will receive the callable.

Examples

>>> import numpy as np
>>> from hrtfpykit.datasets import HUTUBS
>>> from hrtfpykit.datasets import HRTFSpec
>>> from hrtfpykit.datasets import HRTFTransform
>>> transform = HRTFTransform.modify_phase(
...     np.zeros((440, 2, 129)),
...     unit="radians",
... )
>>> dataset = HUTUBS(
...     root="datasets/hutubs",
...     inputs=HRTFSpec(),
...     dataset_hrtf_transform=transform,
... )

static modify_tf(new_tf)¶

Create a callable that replaces each subject TF data array.

The returned callable forwards new_tf to modify_tf(). Replacement data become the loaded subject’s complex frequency-domain values, and IR values are rebuilt before dataset specs extract outputs.

Parameters:: new_tf (np.ndarray | TF | HRTF) – Replacement complex frequency-domain data forwarded to the HRTF transform layer.
Returns:: Dataset-level transform callable accepting and returning an HRTF object.
Return type:: callable

Notes

The replacement leading shape must be compatible with each loaded subject’s source and ear layout. Frequency-bin metadata are copied from TF-like inputs when available.

Examples

>>> import numpy as np
>>> from hrtfpykit.datasets import HUTUBS
>>> from hrtfpykit.datasets import HRTFSpec
>>> from hrtfpykit.datasets import HRTFTransform
>>> transform = HRTFTransform.modify_tf(
...     np.zeros((440, 2, 129), dtype=complex),
... )
>>> dataset = HUTUBS(
...     root="datasets/hutubs",
...     inputs=HRTFSpec(),
...     dataset_hrtf_transform=transform,
... )

static modify_magnitude(new_magnitude, scale='linear')¶

Create a callable that replaces each subject TF magnitude.

The returned callable forwards new_magnitude and scale to modify_magnitude(). The transform preserves current TF phase, replaces magnitude values, and rebuilds IR values before dataset specs consume the loaded subject.

Parameters:

new_magnitude (np.ndarray) – Replacement magnitude array with the same TF layout expected by the HRTF transform layer.
scale ({linear, db}, default=``linear``) – Magnitude scale used by new_magnitude.

Returns:

Dataset-level transform callable accepting and returning an HRTF object.

Return type:

callable

Notes

The replacement magnitude must be broadcast-compatible with the TF layout of every loaded subject that will receive the callable.

Examples

>>> import numpy as np
>>> from hrtfpykit.datasets import HUTUBS
>>> from hrtfpykit.datasets import HRTFSpec
>>> from hrtfpykit.datasets import HRTFTransform
>>> transform = HRTFTransform.modify_magnitude(
...     np.ones((440, 2, 129)),
...     scale="linear",
... )
>>> dataset = HUTUBS(
...     root="datasets/hutubs",
...     inputs=HRTFSpec(),
...     dataset_hrtf_transform=transform,
... )

static apply_gain(gain, scale='db', ear='both')¶

Create a callable that applies TF-domain gain to each subject.

The returned callable forwards gain, scale, and ear to apply_gain(). Gain modifies TF magnitude while preserving phase, then IR values are rebuilt from the adjusted TF.

Parameters:

gain (float | np.ndarray) – Scalar or broadcast-compatible gain applied to each loaded HRTF.
scale ({db, linear}, default=``db``) – Scale used to interpret gain.
ear ({both, left, right}, default=``both``) – Ear channel that receives the gain.

Returns:

Dataset-level transform callable accepting and returning an HRTF object.

Return type:

callable

Examples

>>> from hrtfpykit.datasets import HUTUBS
>>> from hrtfpykit.datasets import HRTFSpec
>>> from hrtfpykit.datasets import HRTFTransform
>>> transform = HRTFTransform.apply_gain(3.0, scale="db", ear="left")
>>> dataset = HUTUBS(
...     root="datasets/hutubs",
...     inputs=HRTFSpec(),
...     dataset_hrtf_transform=transform,
... )

static add_ild(ild)¶

Create a callable that adds ILD to each subject HRTF.

The returned callable forwards ild to add_ild(). Positive ILD values increase the left ear level relative to the right ear. Negative values increase the right ear level relative to the left ear. The transform applies symmetric TF-domain gains and rebuilds IR values from the modified TF.

Parameters:: ild (float | np.ndarray) – ILD values in dB. A scalar applies the same ILD to every source and frequency bin. An array matching each loaded subject’s TF leading shape applies one ILD per source entry. An array matching the TF leading shape plus the frequency axis applies frequency-dependent ILD values.
Returns:: Dataset-level transform callable accepting and returning an HRTF object.
Return type:: callable

Notes

The underlying transform requires two ear channels and TF data. Array ILD inputs must match each loaded subject’s source and frequency layout.

Examples

>>> from hrtfpykit.datasets import HUTUBS
>>> from hrtfpykit.datasets import HRTFSpec
>>> from hrtfpykit.datasets import HRTFTransform
>>> transform = HRTFTransform.add_ild(6.0)
>>> dataset = HUTUBS(
...     root="datasets/hutubs",
...     inputs=HRTFSpec(),
...     dataset_hrtf_transform=transform,
... )

static delete_ild(epsilon=1e-12)¶

Create a callable that removes frequency-dependent ILD from each subject.

The returned callable forwards epsilon to delete_ild(). The transform measures signed frequency-dependent ILD, applies the inverse symmetric left/right TF-domain gain correction, and rebuilds IR values from the modified TF.

Parameters:: epsilon (float, default=1e-12) – Positive floor used while measuring frequency-dependent ILD.
Returns:: Dataset-level transform callable accepting and returning an HRTF object.
Return type:: callable

Notes

The underlying transform requires two ear channels and TF data. The ILD correction is frequency-dependent for each loaded subject.

Examples

>>> from hrtfpykit.datasets import HUTUBS
>>> from hrtfpykit.datasets import HRTFSpec
>>> from hrtfpykit.datasets import HRTFTransform
>>> transform = HRTFTransform.delete_ild()
>>> dataset = HUTUBS(
...     root="datasets/hutubs",
...     inputs=HRTFSpec(),
...     dataset_hrtf_transform=transform,
... )

static modify_fft_length(new_fft_length)¶

Create a callable that changes each subject FFT length.

The returned callable forwards new_fft_length to modify_fft_length(). The transform keeps current IR values unchanged and recomputes TF values with the requested FFT size.

Parameters:: new_fft_length (int) – FFT length used when recomputing the frequency-domain representation.
Returns:: Dataset-level transform callable accepting and returning an HRTF object.
Return type:: callable

Notes

Changing FFT length affects the frequency-bin count and spacing exposed to frequency-domain specs, but it does not add new time-domain information to the HRIR.

Examples

>>> from hrtfpykit.datasets import HUTUBS
>>> from hrtfpykit.datasets import HRTFSpec
>>> from hrtfpykit.datasets import HRTFTransform
>>> transform = HRTFTransform.modify_fft_length(512)
>>> dataset = HUTUBS(
...     root="datasets/hutubs",
...     inputs=HRTFSpec(),
...     dataset_hrtf_transform=transform,
... )

static add_itd(itd, unit='samples')¶

Create a callable that adds ITD to each subject HRIR.

The returned callable forwards itd and unit to add_itd(). Positive ITD values delay the left ear, negative values delay the right ear, and TF values are rebuilt after the time-domain shift.

Parameters:

itd (float) – Interaural time difference added to each loaded HRTF. Positive values delay the left ear and negative values delay the right ear.
unit ({samples, time}, default=``samples``) – Unit used by itd. time is interpreted in microseconds.

Returns:

Dataset-level transform callable accepting and returning an HRTF object.

Return type:

callable

Notes

This factory is useful for controlled binaural timing perturbations in dataset workflows. The underlying transform requires two ear channels and a delay smaller than the IR length.

Examples

>>> from hrtfpykit.datasets import HUTUBS
>>> from hrtfpykit.datasets import HRTFSpec
>>> from hrtfpykit.datasets import HRTFTransform
>>> transform = HRTFTransform.add_itd(4.0, unit="samples")
>>> dataset = HUTUBS(
...     root="datasets/hutubs",
...     inputs=HRTFSpec(),
...     dataset_hrtf_transform=transform,
... )

static delete_itd(method='threshold', thresh_level=-10.0, upper_cut_freq=3000.0, filter_order=10)¶

Create a callable that estimates and removes ITD from each subject.

The returned callable forwards method, thresh_level, upper_cut_freq, and filter_order to delete_itd(). The transform estimates per-source interaural delay, advances the delayed channel, zero-fills the shifted tail, and rebuilds TF values.

Parameters:

method ({threshold, maxiacce}, default=``threshold``) – ITD estimator used before delay compensation.
thresh_level (float, default=-10.0) – Threshold offset in decibels used by the threshold estimator.
upper_cut_freq (float, default=3000.0) – Low-pass cutoff in hertz used before ITD estimation.
filter_order (int, default=10) – Butterworth low-pass filter order used before ITD estimation.

Returns:

Dataset-level transform callable accepting and returning an HRTF object.

Return type:

callable

Notes

Use this factory when a dataset should expose ITD-normalized acoustics while preserving the rest of the HRTF workflow. The underlying transform requires valid two-ear IR data and sample-rate metadata for the estimator path.

Examples

>>> from hrtfpykit.datasets import HUTUBS
>>> from hrtfpykit.datasets import HRTFSpec
>>> from hrtfpykit.datasets import HRTFTransform
>>> transform = HRTFTransform.delete_itd(method="threshold")
>>> dataset = HUTUBS(
...     root="datasets/hutubs",
...     inputs=HRTFSpec(),
...     dataset_hrtf_transform=transform,
... )