data.dataset.AudioFileDataset

data.dataset.AudioFileDataset(
    metadata,
    processor,
    audio_dir='data',
    sample_rate=16000,
    chunk_size=30,
    alignment_strategy='speech',
)

Loads audio files and corresponding metadata files. Splits the audio into chunks according to metadata, and creates wav2vec2 features for each chunk. Returns an AudioSliceDataset object containing the features for each chunk, along with the metadata.

Parameters

Name	Type	Description	Default
metadata	JSONMetadataDataset or list of AudioMetadata or AudioMetadata	List of AudioMetadata objects or paths to JSON files.	required
processor	Wav2Vec2Processor or WhisperProcessor	The Wav2vec2Processor to use for feature extraction.	required
audio_dir	str	Directory with audio files	"data"
sample_rate	int	Sample rate to resample audio to.	16000
chunk_size	int	When VAD is not used, SpeechSegments are naively split into chunk_size sized chunks for feature extraction.	30
alignment_strategy	str	‘speech’ or ‘chunk’ - determines how chunks are defined.	"speech"

Methods

Name	Description
get_speech_features	Extract features for each speech segment in the metadata.
get_vad_features	Extract features for each VAD chunk in the metadata.