mindspore.dataset.audio.MFCC

class mindspore.dataset.audio.MFCC(sample_rate=16000, n_mfcc=40, dct_type=2, norm=NormMode.ORTHO, log_mels=False, melkwargs=None)

Create MFCC for a raw audio signal.

Parameters

• sample_rate (int, optional) – Sampling rate of audio signal (in Hz), can’t be less than 0. Default: 16000.

• n_mfcc (int, optional) – Number of mfc coefficients to retain, can’t be less than 0. Default: 40.

• dct_type (int, optional) – Type of DCT (discrete cosine transform) to use, can only be 2. Default: 2.

• norm (NormMode, optional) – Norm to use. Default: NormMode.ORTHO.

• log_mels (bool, optional) – Whether to use log-mel spectrograms instead of db-scaled. Default: False.

• melkwargs (dict, optional) –

  Arguments for mindspore.dataset.audio.MelSpectrogram. Default: None, the default setting is a dict including

  • ’n_fft’: 400

  • ’win_length’: n_fft

  • ’hop_length’: win_length // 2

  • ’f_min’: 0.0

  • ’f_max’: sample_rate // 2

  • ’pad’: 0

  • ’window’: WindowType.HANN

  • ’power’: 2.0

  • ’normalized’: False

  • ’center’: True

  • ’pad_mode’: BorderType.REFLECT

  • ’onesided’: True

  • ’norm’: NormType.NONE

  • ’mel_scale’: MelType.HTK

Raises

• TypeError – If sample_rate is not of type int.

• TypeError – If log_mels is not of type bool.

• TypeError – If norm is not of type mindspore.dataset.audio.NormMode .

• TypeError – If n_mfcc is not of type int.

• TypeError – If melkwargs is not of type dict.

• ValueError – If sample_rate is a negative number.

• ValueError – If n_mfcc is a negative number.

• ValueError – If dct_type is not 2.

Supported Platforms:

CPU

Examples

>>> import numpy as np
>>> import mindspore.dataset as ds
>>> import mindspore.dataset.audio as audio
>>>
>>> # Use the transform in dataset pipeline mode
>>> waveform = np.random.random([5, 500])  # 5 samples
>>> numpy_slices_dataset = ds.NumpySlicesDataset(data=waveform, column_names=["audio"])
>>> transforms = [audio.MFCC(4000, 128, 2)]
>>> numpy_slices_dataset = numpy_slices_dataset.map(operations=transforms, input_columns=["audio"])
>>> for item in numpy_slices_dataset.create_dict_iterator(num_epochs=1, output_numpy=True):
...     print(item["audio"].shape, item["audio"].dtype)
...     break
(128, 3) float32
>>>
>>> # Use the transform in eager mode
>>> waveform = np.random.random([500])  # 1 sample
>>> output = audio.MFCC(4000, 128, 2)(waveform)
>>> print(output.shape, output.dtype)
(128, 3) float32

Tutorial Examples:

• Illustration of audio transforms