mindspore.dataset.audio.Resample

class mindspore.dataset.audio.Resample(orig_freq=16000, new_freq=16000, resample_method=ResampleMethod.SINC_INTERPOLATION, lowpass_filter_width=6, rolloff=0.99, beta=None)

Resample a signal from one frequency to another. A resample method can be given.

Parameters

• orig_freq (float, optional) – The original frequency of the signal, must be positive. Default: 16000.

• new_freq (float, optional) – The desired frequency, must be positive. Default: 16000.

• resample_method (ResampleMethod, optional) – The resample method to use, can be ResampleMethod.SINC_INTERPOLATION or ResampleMethod.KAISER_WINDOW. Default: ResampleMethod.SINC_INTERPOLATION.

• lowpass_filter_width (int, optional) – Controls the sharpness of the filter, more means sharper but less efficient, must be positive. Default: 6.

• rolloff (float, optional) – The roll-off frequency of the filter, as a fraction of the Nyquist. Lower values reduce anti-aliasing, but also reduce some of the highest frequencies, in range of (0, 1]. Default: 0.99.

• beta (float, optional) – The shape parameter used for kaiser window. Default: None, will use 14.769656459379492.

Raises

• TypeError – If orig_freq is not of type float.

• ValueError – If orig_freq is not a positive number.

• TypeError – If new_freq is not of type float.

• ValueError – If new_freq is not a positive number.

• TypeError – If resample_method is not of type mindspore.dataset.audio.ResampleMethod .

• TypeError – If lowpass_filter_width is not of type int.

• ValueError – If lowpass_filter_width is not a positive number.

• TypeError – If rolloff is not of type float.

• ValueError – If rolloff is not in range of (0, 1].

• RuntimeError – If input tensor is not in shape of <…, time>.

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, 16, 30])  # 5 samples
>>> numpy_slices_dataset = ds.NumpySlicesDataset(data=waveform, column_names=["audio"])
>>> transforms = [audio.Resample(orig_freq=48000, new_freq=16000,
...                              resample_method=audio.ResampleMethod.SINC_INTERPOLATION,
...                              lowpass_filter_width=6, rolloff=0.99, beta=None)]
>>> 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
(16, 10) float64
>>>
>>> # Use the transform in eager mode
>>> waveform = np.random.random([16, 30])  # 1 sample
>>> output = audio.Resample(orig_freq=48000, new_freq=16000,
...                         resample_method=audio.ResampleMethod.SINC_INTERPOLATION,
...                         lowpass_filter_width=6, rolloff=0.99, beta=None)(waveform)
>>> print(output.shape, output.dtype)
(16, 10) float64

Tutorial Examples:

• Illustration of audio transforms