Audio Preprocessing

audioclass.preprocess

Module for preprocessing audio data.

This module helps loading audio data and preprocessing it into a standardized format for audio classification models.

Provides functions for loading audio, resampling, and framing into fixed-length buffers.

Functions

load_clip(clip, samplerate, buffer_size, audio_dir=None)

Load an audio clip from a soundevent Clip object.

This function will load the clip from the audio file, preprocess it, and return a numpy array.

Parameters:

Name	Type	Description	Default
clip	Clip	The soundevent Clip object representing the audio segment.	required
samplerate	int	The desired sample rate to resample the audio to.	required
buffer_size	int	The length of each audio frame in samples.	required
audio_dir	Optional[Path]	The directory containing the audio files. If not provided, the clip's default audio directory is used.	None

Returns:

Type	Description
ndarray	A numpy array of shape (num_frames, buffer_size) containing the preprocessed audio data.

load_recording(recording, samplerate, buffer_size, audio_dir=None)

Load an audio recording from a soundevent Recording object.

This function will load the audio file, preprocess it, and return a numpy array.

Parameters:

Name	Type	Description	Default
recording	Recording	The soundevent Recording object representing the audio file.	required
samplerate	int	The desired sample rate to resample the audio to.	required
buffer_size	int	The length of each audio frame in samples.	required
audio_dir	Optional[Path]	The directory containing the audio files. If not provided, the recording's default audio directory is used.	None

Returns:

Type	Description
ndarray	A numpy array of shape (num_frames, buffer_size) containing the preprocessed audio data.

preprocess_audio(wave, samplerate, buffer_size)

Preprocess a loaded audio waveform.

This function performs the following preprocessing steps:

1. Selects the first channel if multiple channels are present.
2. Resamples the audio to the specified sample rate.
3. Frames the audio into fixed-length buffers.

Parameters:

Name	Type	Description	Default
wave	DataArray	The loaded audio waveform.	required
samplerate	int	The desired sample rate to resample the audio to.	required
buffer_size	int	The length of each audio frame in samples.	required

Returns:

Type	Description
ndarray	A numpy array of shape (num_frames, buffer_size) containing the preprocessed audio data.

resample_audio(wave, samplerate)

Resample audio to a specific sample rate.

Parameters:

Name	Type	Description	Default
wave	DataArray	The audio waveform to resample.	required
samplerate	int	The target sample rate.	required

Returns:

Type	Description
DataArray	The resampled audio waveform.

stack_array(arr, buffer_size)

Stack a 1D array into a 2D array of fixed-length buffers.

This function pads the input array with zeros if necessary to ensure that the number of elements is divisible by the buffer size.

Parameters:

Name	Type	Description	Default
arr	ndarray	The 1D array to stack.	required
buffer_size	int	The length of each buffer.	required

Returns:

Type	Description
ndarray	A 2D array of shape (num_buffers, buffer_size) containing the stacked buffers.