PyGaborSTM¶
core
¶
High-level user-facing class for PyGaborSTM.
Wraps :class:AuditorySpectrogram and :class:GaborFilterbank behind a
single object so users don't have to manage the two stages by hand.
PyGaborSTM
¶
PyGaborSTM(config: Config | None = None)
Main interface for spectro-temporal modulation analysis.
Holds an :class:AuditorySpectrogram and a :class:GaborFilterbank
configured from the same :class:Config, and exposes both per-stage
methods (:meth:spectrogram, :meth:rsf) and full-pipeline
convenience methods (:meth:compute, :meth:compute_device).
| PARAMETER | DESCRIPTION |
|---|---|
config
|
Configuration object. If
TYPE:
|
| ATTRIBUTE | DESCRIPTION |
|---|---|
config |
The configuration used to build both internal stages.
TYPE:
|
Examples:
>>> import pygaborstm as stm
>>> model = stm.PyGaborSTM(config=stm.Config(use_gpu=True))
>>> spec = model.spectrogram(audio)
>>> rsf = model.rsf(spec)
Or chained without an intermediate host transfer:
spectrogram
¶
spectrogram(audio: ndarray) -> Spectrogram
Compute the auditory spectrogram and return it on host.
| PARAMETER | DESCRIPTION |
|---|---|
audio
|
1D audio signal.
TYPE:
|
| RETURNS | DESCRIPTION |
|---|---|
Spectrogram
|
Spectrogram with data and axis metadata on the host. |
rsf
¶
rsf(spec: Spectrogram) -> RSF
Compute the RSF representation from a spectrogram, on host.
| PARAMETER | DESCRIPTION |
|---|---|
spec
|
Auditory spectrogram, typically produced by
:meth:
TYPE:
|
| RETURNS | DESCRIPTION |
|---|---|
RSF
|
RSF representation with data and axis metadata on the host. |
compute_device
¶
Run the full pipeline on device with no intermediate host transfer.
Spectrogram output stays on the device and feeds directly into
the Gabor stage. Use this when you do not need the intermediate
:class:Spectrogram dataclass.
| PARAMETER | DESCRIPTION |
|---|---|
audio
|
1D audio signal.
TYPE:
|
| RETURNS | DESCRIPTION |
|---|---|
ndarray or ndarray
|
RSF tensor of shape
|
compute
¶
compute(audio: ndarray) -> RSF
Run the full pipeline and return an RSF dataclass on host.
Chains both stages on device with no intermediate host transfer,
then copies the final result to host and wraps it in an
:class:RSF.
| PARAMETER | DESCRIPTION |
|---|---|
audio
|
1D audio signal.
TYPE:
|
| RETURNS | DESCRIPTION |
|---|---|
RSF
|
Host-side RSF with data and axis metadata. |