Extract mel spectrogram

Simple extraction of mel spectrogram

import numpy as np
import audioflux as af

# Get a 220Hz's audio file path
sample_path = af.utils.sample_path('220')

# Read audio data and sample rate
audio_arr, sr = af.read(sample_path)

# Extract mel spectrogram
spec_arr, _ = af.mel_spectrogram(audio_arr, samplate=sr)

We recommend using the BFT class, you can use it more flexibly and efficiently.

import numpy as np
import audioflux as af
from audioflux.type import SpectralFilterBankScaleType

# Get a 220Hz's audio file path
sample_path = af.utils.sample_path('220')

# Read audio data and sample rate
audio_arr, sr = af.read(sample_path)

# Create BFT object and extract mel spectrogram
bft_obj = af.BFT(num=128, radix2_exp=12, samplate=sr,
                 scale_type=SpectralFilterBankScaleType.MEL)
spec_arr = bft_obj.bft(audio_arr)
spec_arr = np.abs(spec_arr)

# Display spectrogram
import matplotlib.pyplot as plt
from audioflux.display import fill_spec
audio_len = audio_arr.shape[-1]
fig, ax = plt.subplots()
img = fill_spec(spec_arr, axes=ax,
          x_coords=bft_obj.x_coords(audio_len),
          y_coords=bft_obj.y_coords(),
          x_axis='time', y_axis='log',
          title='Mel Spectrogram')
fig.colorbar(img, ax=ax)

Of course, BFT can not only extract the Mel spectrogram, but also extract the following spectrogram (just adjust the scale_type parameter):

• Linear(STFT) - Short-time Fourier transform.

• Linspace - Numbers evenly distributed over the specified interval.

• Mel - Mel-scale spectrogram.

• Bark - Bark-scale spectrogram.

• Erb - Erb-scale spectrogram.

• Octave - Octave-scale spectrogram.

• Log - Evenly distributed numbers on a logarithmic scale.