AudioBuffer Class

Class for storing and passing audio signals around. Can be instantiated with an array of audio samples or location to load audio from disk. Stores sample rate, number of channels, and the file name if loaded from disk.

Also has utility functions for loading and saving audio from disk, normalizing, resizing buffers, and plotting spectrograms.

Examples

Create a new empty AudioBuffer:

>>> audio = spiegelib.AudioBuffer()

Create an AudioBuffer from existing audio samples:

>>> audio = spiegelib.AudioBuffer(audio_samples, 44100)

Load audio from disk into a new AudioBuffer:

>>> audio = spiegelib.AudioBuffer('./some_audio_file.wav')

Load an entire folder of audio samples into a list of AudioBuffer objects:

>>> audio_files = spiegelib.AudioBuffer.load_folder('./audio_folder')

Plot a spectrogram

1
2
3
4
5
6
import spiegelib as spgl
import matplotlib.pyplot as plt

audio = spgl.AudioBuffer('./audio_file.wav')
audio.plot_spectrogram()
plt.show()
../../_images/audio_buffer_spect.png
class spiegelib.AudioBuffer(input=None, sample_rate=None, **kwargs)
Parameters

  • input (np.ndarray, list, str, file-like object) – Can be an array of audio samples (np.ndarray or list), or a path to a location of an audio file on disk. Defaults to None.

  • sample_rate (int) – Rate of sampled audio if audio data was passed in, or rate to resample audio data loaded from disk at, defaults to None. If loading audio from a file, will automatically resample to 44100 if no sample rate provided.

  • kwargs – keyword args to pass into librosa load function.

get_audio()

Getter for audio data

Returns

Array of audio samples

Return type

np.ndarray

get_sample_rate()

Getter for sample rate

Returns

Sample rate of audio buffer

Return type

int

load(path, sample_rate=44100, **kwargs)

Read audio from a file into a numpy array at specific audio rate. Stores in this AudioBuffer object.

Parameters

  • path (str, int, pathlib.Path, or file-like object) – location of audio file on disk

  • sample_rate (int) – resample audio to this rate, defaults to 44100. None uses native sampling rate.

  • kwargs –

    keyword args to pass into librosa load function.

plot_spectrogram(**kwargs)

Plot spectrogram of this audio buffer. Uses librosas specshow function. Uses matplotlib.pyplot to plot spectrogram.

Defaults to logarithmic y-axis in Hertz and seconds along the x-axis.

Parameters

kwargs –

see specshow.

Returns

The axis handle for figures

Return type

axes

replace_audio_data(audio, sample_rate)

Replace audio data in this object

Parameters

  • audio (np.ndarray) – Array of audio samples

  • sample_rate (int) – Rate that audio data was sampled at

resize(num_samples, start=0)

Resize audio to a set number of samples. If new length is less than the current audio buffer size, then the buffer will be trimmed. If the new length is greater than current audio buffer, then the resulting buffer will be zero-padded at the end.

Parameters

  • num_samples (int) – New length of audio buffer in samples

  • start_sample (int) – Start reading from certain number of samples into current buffer

save(path, normalize=False)

Save an audio to disk as WAV file at given sample rate. Can normalize before saving as well.

Parameters

  • path (str or open file handle) – location to save audio to

  • normalize (boolean) – normalize audio before writing, defaults to false

static peak_normalize(audio)

Peak normalize audio data

Parameters

audio (np.ndarray) – array of audio samples

Returns

normalized array of audio samples

Return type

np.ndarray

static load_folder(path, sort=True, sample_rate=44100, **kwargs)

Try to load a folder of audio samples

Parameters

  • path (str) – Path to directory of audio files

  • sort (bool) – Apply natural sort to file names. Default True

  • sample_rate (int) – Sample rate to load audio files at. Defualts to 44100

  • kwargs –

    keyword args to pass into librosa load function.

Returns

List of AudioBuffer objects

Return type

list