Librosa未能绘制生成的mfcc

Question

我是能够生成MFCC从系统捕获的音频和绘图，但经过一些重构和配置Tensorflow与CUDA。我使用Librosa生成mfcc，matplotlib.pyplot和librosa.display绘制MFCC和声音设备从立体声混合捕捉声音从窗口。当前的配置可以根据示例.wav文件创建和绘制MFCC，但是当使用系统捕获的声音时，它无法绘制它，因为它生成一个3D数组，而不是在运行MFCC时生成一个2D。下面是生成和绘制的代码

N_MFCC = 40 
N_MELS = 40 
N_FFT = 512 
HOP_LENGTH = 160 
MIN_FREQ = 0 
MAX_FREQ = None

def create_mfcc(record, sample_rate):
  features = librosa.feature.mfcc(record, sample_rate, n_fft=N_FFT,n_mfcc=N_MFCC, 
              n_mels=N_MELS,hop_length=HOP_LENGTH,fmin=MIN_FREQ, fmax=MAX_FREQ, htk=False)
  return features

def plot_and_save_mfcc(mfcc_data, file_name, sample_rate):
    plt.figure(figsize=(10, 8))
    plt.title('Current audio MFCC', fontsize=18)
    plt.xlabel('Time [s]', fontsize=18)
    librosa_display.specshow(mfcc_data, sr=sample_rate)
    plt.savefig(file_name)
    plt.cla()

这将生成此堆栈跟踪。

Traceback (most recent call last):
  File "main.py", line 68, in <module>
    main()
  File "main.py", line 63, in main
    start_listening_and_creating_mfcc()
  File "main.py", line 48, in start_listening_and_creating_mfcc
    plot_and_save_mfcc(mfcc_data, conf.DEFAULT_MFCC_IMAGE_NAME.format(image_count), conf.SAMPLE_RATE)
  File "main.py", line 38, in plot_and_save_mfcc
    librosa_display.specshow(mfcc_data, sr=sample_rate)
  File anaconda3\lib\site-packages\librosa\util\decorators.py", line 88, in inner_f
    return f(*args, **kwargs)
  File anaconda3\lib\site-packages\librosa\display.py", line 879, in specshow
    out = axes.pcolormesh(x_coords, y_coords, data, **kwargs)
  File anaconda3\lib\site-packages\matplotlib\__init__.py", line 1361, in inner
    return func(ax, *map(sanitize_sequence, args), **kwargs)
  File anaconda3\lib\site-packages\matplotlib\axes\_axes.py", line 6183, in pcolormesh
    X, Y, C, shading = self._pcolorargs('pcolormesh', *args,
  File anaconda3\lib\site-packages\matplotlib\axes\_axes.py", line 5671, in _pcolorargs
    nrows, ncols = C.shape
ValueError: too many values to unpack (expected 2)

我试过调试它并更改mfcc配置，但没有成功。也尝试重新配置我的环境，但这也没有帮助。

编辑:这是系统音频(48000，40，1)和.wav示例文件(40,122个)的.wav

如前所述，我忽略了一个函数，但这里是它和用于为.wav文件加载和创建mfcc的函数

def create_mfcc_from_file(file_path):
    (signal, sample_rate) = librosa.load(file_path)
    librosa_features = create_mfcc(signal, sample_rate)
    plot_and_save_mfcc(librosa_features, 'mfcc-librosa', sample_rate)

def start_listening_and_creating_mfcc():
    image_count = 0
    while True:
        my_recording = record_window()
        mfcc_data = create_mfcc(my_recording, conf.SAMPLE_RATE)
        plot_and_save_mfcc(mfcc_data, conf.DEFAULT_MFCC_IMAGE_NAME.format(image_count), conf.SAMPLE_RATE)
        wav.write(conf.DEFAULT_MFCC_IMAGE_NAME.format(image_count) + '.wav', conf.SAMPLE_RATE, my_recording)
        image_count += 1

Answer 1

def delta(feat, N):
    """Compute delta features from a feature vector sequence.
    :param feat: A numpy array of size (NUMFRAMES by number of features) containing features. Each row holds 1 feature vector.
    :param N: For each frame, calculate delta features based on preceding and following N frames
    :returns: A numpy array of size (NUMFRAMES by number of features) containing delta features. Each row holds 1 delta feature vector.
    """
    if N < 0:
        raise ValueError('N must be an integer >0')
    NUMFRAMES = len(feat)
    denominator = 2 * sum([i**2 for i in range(1, N+1)])
    delta_feat = numpy.empty_like(feat)
    padded = numpy.pad(feat, ((N, N), (0, 0)), mode='edge')   # padded version of feat
    for t in range(NUMFRAMES):
      
        delta_feat[t] = numpy.dot(numpy.arange(-N, N+1), padded[t : t+2*N+1]) / denominator 
        plt.plot(signal, c='c')# [t : t+2*N+1] == [(N+t)-N : (N+t)+N+1]
    return delta_feat