SWT - Stationary Wavelet Transform

class audioflux.SWT(num, fft_length, wavelet_type=WaveletDiscreteType.SYM, t1=4, t2=0)

Stationary Wavelet Transform (SWT)

Parameters

num: int

Number of frequency bins to generate.

fft_length: int

fft length

wavelet_type: WaveletDiscreteType

Wavelet discrete type

See: type.WaveletDiscreteType

Note

t1/t2 settings for wavelet_type:

• DB: t1

  • 2~10/20/30/40

• SYM: t1

  • 2~10/20/30

• COIF: t1

  • 1/2/3/4/5

• FK: t1

  • 4/6/8/14/18/22

• BIOR/DMEY: t1.t2

  • 1.1/1.3/1.5

  • 2.2/2.4/2.6/2.8

  • 3.1/3.3/3.5/3.7/3.9

  • 4.4/5.5/6.8

t1: int

t1 value

t2: int

t2 value

See also

CQT

ST

FST

DWT

WPT

Examples

Read 220Hz audio data

>>> import audioflux as af
>>> audio_path = af.utils.sample_path('220')
>>> audio_arr, sr = af.read(audio_path)
>>> audio_len = 4096 * 5
>>> audio_arr = audio_arr[:audio_len]

Create SWT object

>>> from audioflux.type import WaveletDiscreteType
>>> obj = af.SWT(num=5, fft_length=audio_len,
>>>              wavelet_type=WaveletDiscreteType.DB, t1=4, t2=0)

Get SWT data

>>> data_arr1, data_arr2 = obj.swt(audio_arr)

Save SWT data

>>> save_data_arr1 = data_arr1[-1]  # index `0～num-1` is the SWT data of different series
>>> save_data_arr2 = data_arr2[-1]  # index `0～num-1` is the SWT data of different series
>>> # You can save SWT to audio file
>>> # af.write('SAVE_PATH_1.wav', save_data_arr1)
>>> # af.write('SAVE_PATH_2.wav', save_data_arr2)

Show wave plot

>>> import matplotlib.pyplot as plt
>>> from audioflux.display import fill_wave
>>>
>>> n_num, n_time = data_arr2.shape
>>> fig, ax = plt.subplots(nrows=n_num + 1, figsize=(8, 16), sharex=True, layout='tight')
>>> _ax0 = fill_wave(audio_arr[:n_time], samplate=sr, axes=ax[0])
>>> _ax0.set_title('SWT arr2 origin')
>>> for i in range(n_num):
>>>     _ax = fill_wave(data_arr2[i], samplate=sr, axes=ax[i + 1])
>>>     _ax.set_title(f'SWT arr2 num={i}')
>>>
>>> n_num, n_time = data_arr1.shape
>>> fig, ax = plt.subplots(nrows=n_num + 1, figsize=(8, 16), sharex=True, layout='tight')
>>> _ax0 = fill_wave(audio_arr[:n_time], samplate=sr, axes=ax[0])
>>> _ax0.set_title('SWT arr1 origin')
>>> for i in range(n_num):
>>>     _ax = fill_wave(data_arr1[i], samplate=sr, axes=ax[i + 1])
>>>     _ax.set_title(f'SWT arr1 num={i}')

Methods

swt(data_arr)

Get swt matrix

swt(data_arr)

Get swt matrix

Parameters

data_arr: np.ndarray [shape=(…, n)]

Input audio data

Returns

m_data_arr1: np.ndarray [shape=(…, fre, time)]

m_data_arr2: np.ndarray [shape=(…, fre, time)]