python - 从音频文件中提取 db 频谱图，对频谱图进行去噪并将其转换回音频

Question

我正在尝试将一些图像处理技术应用于由音频文件创建的频谱图。在此示例中，我想对频谱图应用去噪算法，然后将其反演回音频。这将如何正确完成，以便我可以操纵频谱图，然后返回音频而不会丢失大部分信号的初始质量？显然我在这里做错了，所以任何帮助将不胜感激。

我使用了此处找到的部分代码： 如何将 Librosa 频谱图保存为特定尺寸的图像？

这是我正在处理的代码：

!pip install librosa --upgrade

import librosa
import matplotlib.pyplot as plt
import numpy as np
import librosa.display
from IPython.display import Audio,display
from scipy.io.wavfile import write
import skimage.io
from skimage.color import rgb2gray
import cv2



def scale_minmax(X, min=0.0, max=1.0):
    X_std = (X - X.min()) / (X.max() - X.min())
    X_scaled = X_std * (max - min) + min
    return X_scaled

def spectrogram_image(y, sr, out, hop_length, n_mels):

    stft = np.abs(librosa.stft(y=y,n_fft=hop_length*2, hop_length=hop_length))
    amp2db  = librosa.amplitude_to_db(stft, ref=np.max)

    # min-max scale to fit inside 8-bit range
    img = scale_minmax(amp2db, 0, 255).astype(numpy.uint8)
    img = numpy.flip(img, axis=0) # put low frequencies at the bottom in image
    img = 255-img # invert. make black==more energy

    # save as PNG
    skimage.io.imsave(out, img)

    return img


if __name__ == '__main__':
    # settings
    hop_length = 512 # number of samples per time-step in spectrogram
    n_mels = 128 # number of bins in spectrogram. Height of image
    time_steps = 384 # number of time-steps. Width of image

    # load audio. Using example from librosa
    path = librosa.util.example_audio_file()
    y, sr = librosa.load(path)
    out = 'out.png'

    # extract a fixed length window
    start_sample = 0 # starting at beginning
    length_samples = time_steps*hop_length
    window = y[start_sample:start_sample+length_samples]

    # convert to PNG
    img_png= spectrogram_image(window, sr=sr, out=out, hop_length=hop_length, n_mels=n_mels)
    print('wrote file', out)


    converted_img = cv2.cvtColor(img_png, cv2.COLOR_GRAY2BGR)

dst = cv2.fastNlMeansDenoisingColored(converted_img,None,10,10,7,21)

dst=img_gray = rgb2gray(dst)
#dst = scale_minmax(dst, 0, 1.0).astype(numpy.float64)

dst = numpy.flip(dst, axis=0) # do i need this???
fig= plt.figure(figsize=(32,16))

plt.subplot(211),plt.imshow(img_png)
plt.subplot(212),plt.imshow(dst)
plt.show()


y=librosa.amplitude_to_db(dst)
y_hat = librosa.istft(y)
#y_hat = librosa.griffinlim(y)

audio1= Audio(y_hat,rate=sr)
display(audio1)

write("/content/XXX.wav", sr,y_hat)

标签： pythonnumpyaudiospectrogram