callcut.extractors.BaseExtractor

class callcut.extractors.BaseExtractor(sample_rate)

Abstract base class for feature extractors.

All feature extractors must implement:

• extract(): Extract features from a waveform

• n_features: Number of output feature channels

• hop_ms: Hop length in milliseconds (determines frame rate)

The base class provides utility methods for time-frame conversion.

Parameters:

sample_rateint

Expected sample rate in Hz. Waveforms should be resampled to this rate before extraction.

Attributes

frame_rate	Frame rate in Hz (frames per second).
hop_ms	Hop length in milliseconds.
hop_s	Hop length in seconds.
n_features	Number of output feature channels.
sample_rate	Expected sample rate in Hz.

Methods

__call__(waveform)	Extract features (alias for extract()).
extract(waveform)	Extract features from a waveform.
frames_to_seconds(frames)	Convert number of frames to duration in seconds.
seconds_to_frames(seconds)	Convert duration in seconds to number of frames.

Examples

Create a custom extractor by subclassing BaseExtractor:

>>> class MyExtractor(BaseExtractor):
...     def __init__(self, sample_rate: int, hop_ms: float = 10.0):
...         super().__init__(sample_rate)
...         self._hop_ms = hop_ms
...
...     @property
...     def n_features(self) -> int:
...         return 64
...
...     @property
...     def hop_ms(self) -> float:
...         return self._hop_ms
...
...     def extract(self, waveform: Tensor) -> tuple[Tensor, Tensor]:
...         # Implementation here
...         ...

__call__(waveform)

Extract features (alias for extract()).

Parameters:

waveformTensor

Audio waveform of shape (1, samples) or (samples,).

Returns:

featuresTensor

Extracted features of shape (n_features, n_frames).

timesTensor

Time axis of shape (n_frames,) in seconds.

abstractmethod extract(waveform)

Extract features from a waveform.

Parameters:

waveformTensor

Audio waveform of shape (1, samples) or (samples,). Should be mono and sampled at sample_rate.

Returns:

featuresTensor

Extracted features of shape (n_features, n_frames).

timesTensor

Time axis of shape (n_frames,) in seconds.

frames_to_seconds(frames)

Convert number of frames to duration in seconds.

Parameters:

framesint

Number of frames.

Returns:

secondsfloat

Duration in seconds.

Examples

>>> extractor = SNRExtractor(sample_rate=32000, hop_ms=8.0)
>>> extractor.frames_to_seconds(250)
2.0

seconds_to_frames(seconds)

Convert duration in seconds to number of frames.

Parameters:

secondsfloat

Duration in seconds.

Returns:

framesint

Number of frames (rounded to nearest integer).

Examples

>>> extractor = SNRExtractor(sample_rate=32000, hop_ms=8.0)
>>> extractor.seconds_to_frames(2.0)
250

property frame_rate

Frame rate in Hz (frames per second).

Type:

float

abstract property hop_ms

Hop length in milliseconds.

Determines the time resolution of extracted features.

Type:

float

property hop_s

Hop length in seconds.

Type:

float

abstract property n_features

Number of output feature channels.

For SNR features, this is the number of frequency bands.

Type:

int

property sample_rate

Expected sample rate in Hz.

Type:

int