medkit.core.audio

APIs

For accessing these APIs, you may use import like this:

from medkit.core.audio import <api_to_import>

Classes:

AudioAnnotationContainer(doc_id, raw_segment)	Manage a list of audio annotations belonging to an audio document.
AudioBuffer(sample_rate, nb_samples, nb_channels)	Audio buffer base class.
AudioDocument(audio[, anns, metadata, uid])	Document holding audio annotations.
FileAudioBuffer(path[, trim_start, ...])	Audio buffer giving access to audio files stored on the filesystem (to use when manipulating unmodified raw audio).
MemoryAudioBuffer(signal, sample_rate)	Audio buffer giving acces to signals stored in memory (to use when reading/writing a modified audio signal).
PreprocessingOperation([uid, name])	Abstract operation for pre-processing segments.
Segment(label, audio, span[, attrs, ...])	Audio segment referencing part of an AudioDocument.
SegmentationOperation([uid, name])	Abstract operation for segmenting audio.
Span(start, end)	Boundaries of a slice of audio.

class Segment(label, audio, span, attrs=None, metadata=None, uid=None)

Audio segment referencing part of an AudioDocument.

Variables

• uid (str) – Unique identifier of the segment.

• label (str) – Label of the segment.

• audio (medkit.core.audio.audio_buffer.AudioBuffer) – The audio signal of the segment. It must be consistent with the span, in the sense that it must correspond to the audio signal of the document at the span boundaries. But it can be a modified, processed version of this audio signal.

• span (medkit.core.audio.span.Span) – Span (in seconds) indicating the part of the document’s full signal that this segment references.

• attrs (medkit.core.attribute_container.AttributeContainer) – Attributes of the segment. Stored in a :class:{~medkit.core.AttributeContainer} but can be passed as a list at init.

• metadata (Dict[str, Any]) – Metadata of the segment.

• keys (Set[str]) – Pipeline output keys to which the annotation belongs to.

class AudioAnnotationContainer(doc_id, raw_segment)

Manage a list of audio annotations belonging to an audio document.

This behaves more or less like a list: calling len() and iterating are supported. Additional filtering is available through the get() method.

Also provides handling of raw segment.

class AudioBuffer(sample_rate, nb_samples, nb_channels)

Audio buffer base class. Gives access to raw audio samples.

Parameters

• sample_rate (int) – Sample rate of the signal, in samples per second.

• nb_samples (int) – Duration of the signal in samples.

• nb_channels (int) – Number of channels in the signal.

Attributes:

duration

Duration of the signal in seconds.

Methods:

read([copy])	Return the signal in the audio buffer.
trim(start, end)	Return a new audio buffer pointing to portion of the signal in the original buffer, using boundaries in samples.
trim_duration([start_time, end_time])	Return a new audio buffer pointing to a portion of the signal in the original buffer, using boundaries in seconds.

property duration: float

Duration of the signal in seconds.

Return type

float

abstract read(copy=False)

Return the signal in the audio buffer.

Parameters

copy (bool) – If True, the returned array will be a copy that can be safely mutated.

Return type

ndarray

Returns

np.ndarray – Raw audio samples

abstract trim(start, end)

Return a new audio buffer pointing to portion of the signal in the original buffer, using boundaries in samples.

Parameters

• start (Optional[int]) – Start sample of the new buffer (defaults to 0).

• end (Optional[int]) – End sample of the new buffer, excluded (default to full duration).

Return type

AudioBuffer

Returns

AudioBuffer – Trimmed audio buffer with new start and end samples, of same type as original audio buffer.

trim_duration(start_time=None, end_time=None)

Return a new audio buffer pointing to a portion of the signal in the original buffer, using boundaries in seconds. Since start_time and end_time are in seconds, the exact trim boundaries will be rounded to the nearest sample and will therefore depend on the sampling rate.

Parameters

• start_time (Optional[float]) – Start time of the new buffer (defaults to 0.0).

• end_time (Optional[float]) – End time of thew new buffer, excluded (default to full duration).

Return type

AudioBuffer

Returns

AudioBuffer – Trimmed audio buffer with new start and end samples, of same type as original audio buffer.

class FileAudioBuffer(path, trim_start=None, trim_end=None, sf_info=None)

Audio buffer giving access to audio files stored on the filesystem (to use when manipulating unmodified raw audio).

Parameters

• path (Union[str, Path]) – Path to the audio file.

• trim_start (Optional[int]) – First sample of audio file to consider.

• trim_end (Optional[int]) – First sample of audio file to exclude.

• sf_info (Optional[Any]) – Optional metadata dict returned by soundfile.

class MemoryAudioBuffer(signal, sample_rate)

Audio buffer giving acces to signals stored in memory (to use when reading/writing a modified audio signal).

Parameters

• signal (ndarray) – Samples constituting the audio signal, with shape (nb_channel, nb_samples).

• sample_rate (int) – Sample rate of the signal, in samples per second.

class AudioDocument(audio, anns=None, metadata=None, uid=None)

Document holding audio annotations.

Variables

• uid (str) – Unique identifier of the document.

• audio – Audio buffer containing the entire signal of the document.

• anns (AudioAnnotationContainer) – Annotations of the document. Stored in an AudioAnnotationContainer but can be passed as a list at init.

• metadata (Dict[str, Any]) – Document metadata.

• raw_segment (Segment) – Auto-generated segment containing the full unprocessed document audio.

Attributes:

RAW_LABEL

Label to be used for raw segment

RAW_LABEL: ClassVar[str] = 'RAW_AUDIO'

Label to be used for raw segment

class PreprocessingOperation(uid=None, name=None, **kwargs)

Abstract operation for pre-processing segments.

It uses a list of segments as input and produces a list of pre-processed segments. Each input segment will have a corresponding output segment.

Common initialization for all annotators:

• assigning identifier to operation

• storing class name, name and config in description

Parameters

• uid (str) – Operation identifier

• name – Operation name (defaults to class name)

• kwargs – All other arguments of the child init useful to describe the operation

Examples

In the __init__ function of your annotator, use:

>>> init_args = locals()
>>> init_args.pop('self')
>>> super().__init__(**init_args)

class SegmentationOperation(uid=None, name=None, **kwargs)

Abstract operation for segmenting audio.

It uses a list of segments as input and produces a list of new segments. Each input segment will have zero, one or more corresponding output segments.

Common initialization for all annotators:

• assigning identifier to operation

• storing class name, name and config in description

Parameters

• uid (str) – Operation identifier

• name – Operation name (defaults to class name)

• kwargs – All other arguments of the child init useful to describe the operation

Examples

In the __init__ function of your annotator, use:

>>> init_args = locals()
>>> init_args.pop('self')
>>> super().__init__(**init_args)

class Span(start, end)

Boundaries of a slice of audio.

Variables

• start (float) – Starting point in the original audio, in seconds.

• end (float) – Ending point in the original audio, in seconds.

Create new instance of Span(start, end)

Attributes:

end	Alias for field number 1
length	Length of the span, in seconds
start	Alias for field number 0

property start

Alias for field number 0

property end

Alias for field number 1

property length

Length of the span, in seconds

Subpackages / Submodules

medkit.core.audio.annotation
medkit.core.audio.annotation_container
medkit.core.audio.audio_buffer
medkit.core.audio.document
medkit.core.audio.operation
medkit.core.audio.span