Python Librosa: 生成音频信号的语谱图、MFCC 图和对数梅尔谱图

这个任务可以使用 Python 中的 Librosa 库来完成。首先，我们需要导入所需的库和模块：

import os
import librosa
import librosa.display
import numpy as np
import matplotlib.pyplot as plt

然后，我们需要定义一个函数来生成语谱图、mfcc 图和对数梅尔谱图。该函数需要接受一个音频文件的路径作为参数，并返回生成的图像数据。

def generate_features(audio_path):
    # Load audio file
    y, sr = librosa.load(audio_path, sr=None)

    # Generate spectrogram
    spectrogram = np.abs(librosa.stft(y))
    spectrogram_db = librosa.amplitude_to_db(spectrogram, ref=np.max)

    # Generate MFCCs
    mfccs = librosa.feature.mfcc(y=y, sr=sr, n_mfcc=20)
    mfccs_db = librosa.amplitude_to_db(mfccs, ref=np.max)

    # Generate mel spectrogram
    mel_spectrogram = librosa.feature.melspectrogram(y=y, sr=sr)
    mel_spectrogram_db = librosa.amplitude_to_db(mel_spectrogram, ref=np.max)

    return spectrogram_db, mfccs_db, mel_spectrogram_db

接下来，我们需要遍历数据集中的每个音频文件，并生成对应的图像数据，并将它们保存到新的文件夹中。

# Dataset path
dataset_path = 'D:/论文代码/data/'

# Output path
output_path = 'D:/论文代码/outputs/'

# Get list of subdirectories in dataset path
subdirs = [os.path.join(dataset_path, d) for d in os.listdir(dataset_path) if os.path.isdir(os.path.join(dataset_path, d))]

# Loop through subdirectories
for subdir in subdirs:
    print('Processing directory:', subdir)

    # Get list of audio files in subdirectory
    audio_files = [os.path.join(subdir, f) for f in os.listdir(subdir) if f.endswith('.wav')]

    # Create output directory for subdirectory
    subdir_name = os.path.basename(subdir)
    output_dir = os.path.join(output_path, subdir_name)
    if not os.path.exists(output_dir):
        os.makedirs(output_dir)

    # Loop through audio files in subdirectory
    for audio_file in audio_files:
        print('Processing file:', audio_file)

        # Generate features
        spectrogram, mfccs, mel_spectrogram = generate_features(audio_file)

        # Save spectrogram
        spectrogram_file = os.path.join(output_dir, os.path.splitext(os.path.basename(audio_file))[0] + '_spectrogram.png')
        plt.figure(figsize=(10, 4))
        librosa.display.specshow(spectrogram, y_axis='log')
        plt.colorbar(format='%+2.0f dB')
        plt.title('Log-frequency power spectrogram')
        plt.savefig(spectrogram_file)
        plt.close()

        # Save MFCCs
        mfccs_file = os.path.join(output_dir, os.path.splitext(os.path.basename(audio_file))[0] + '_mfccs.png')
        plt.figure(figsize=(10, 4))
        librosa.display.specshow(mfccs, x_axis='time')
        plt.colorbar(format='%+2.0f dB')
        plt.title('Mel-frequency cepstral coefficients (MFCCs)')
        plt.savefig(mfccs_file)
        plt.close()

        # Save mel spectrogram
        mel_spectrogram_file = os.path.join(output_dir, os.path.splitext(os.path.basename(audio_file))[0] + '_mel_spectrogram.png')
        plt.figure(figsize=(10, 4))
        librosa.display.specshow(mel_spectrogram, y_axis='mel', x_axis='time')
        plt.colorbar(format='%+2.0f dB')
        plt.title('Mel spectrogram')
        plt.savefig(mel_spectrogram_file)
        plt.close()

print('Done')

以上代码将会生成语谱图、mfcc 图和对数梅尔谱图，并将它们保存到新的文件夹中，要求语谱图分类与数据集的子目录相同，音频文件在数据集下的子目录中。