emonet.utils module
emonet.utils module#
More miscellaneous functions that probably fit better somewhere else.
- emonet.utils.play_audio(waveform: torch.Tensor, sample_rate: int)[source]#
Play an audio signal within an IPython notebook.
- Parameters
waveform (torch.Tensor) – Audio signal.
sample_rate (int) – Audio signal sample rate.
- Returns
None
- emonet.utils.async_file_operation(files: List, func: Callable, max_workers: int = 50)[source]#
Asynchronously apply a function on a list of files.
- Parameters
files (List) – Files to apply function to.
func (Callable) – Function to apply.
max_workers (int) – Maximum number of concurrent threads.
- Returns
None
- emonet.utils.get_rating_encoder(ratings: List[str]) speechbrain.dataio.encoder.CategoricalEncoder[source]#
Get a categorical encoder for emotion severity labels.
- Parameters
ratings (List[str]) – List containing all possible categorical ratings.
- Returns
CategoricalEncoder – A fitted Speechbrain CategoricalEncoder object.
- emonet.utils.binarize_ratings(ratings: Union[int, List, torch.Tensor]) Union[int, torch.Tensor][source]#
Convert emotion severity levels to binary.
Assigns None and Low (0 and 1, respectively) to 0; assigns Med and High (2 and 3, respectively) to 1.
- Parameters
ratings (Union[int, List, torch.Tensor]) – Rating(s) to convert; expects {0, 1, 2, 3}.
- Returns
Union[int, torch.Tensor] – Binarized label(s).
- emonet.utils.binarize_labels(labels: Dict) Dict[source]#
Convert all labels within a dictionary to binary.
- Parameters
labels (Dict) – A dictionary mapping a specific file or sample to its respective emotion severity label.
- Returns
Dict – The original labels dictionary with binarized labels.
- emonet.utils.get_sample(file: pathlib.Path, sample_rate: int) Tuple[torch.Tensor, int][source]#
Read an audio signal from a file and resample, if necessary.
- Parameters
file (pathlib.Path) – Path to audio file.
sample_rate (int) – Desired sample rate.
- Returns
Tuple[torch.Tensor, int] – Audio signal (potentially resampled) and output sample rate.
- emonet.utils.get_random_segment(wav: torch.Tensor, seconds: int = 7, sample_rate: int = 16000) torch.Tensor[source]#
Get a random n-second sample from an audio signal. :Parameters: * wav (torch.Tensor) – Original audio signal.
seconds (int) – Desired duration of random sample.
sample_rate (int) – Original audio sampling rate.
- Returns
torch.Tensor – A random segment of the original audio signal.