毕业设计：多音频文件频谱图批量提取代码求助

Hey there! Let's get your batch spectrogram extraction working smoothly for your thesis—you've got the core single-file logic down, we just need to tweak the loop to handle multiple files properly. Here's what's missing in your current code and how to fix it:

Key Issues in Your Current Code #

• You're looping over an index range, but not actually loading each individual audio file (your x and sr variables are still tied to a single audio file outside the loop)
• There's no logic to save the generated spectrograms (critical for batch processing—you don't want to just display them one by one!)
• No cleanup of matplotlib figures, which will eat up memory as you process more files

Fixed & Annotated Code #

First, make sure you've imported all required libraries:

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

Then, here's the updated batch processing code:

# First, make sure audio_path is a list of ALL your audio file paths
# Example to get all .wav/mp3 files in a folder:
# audio_path = librosa.util.find_files("path/to/your/audio_directory", ext=["wav", "mp3"])

# Loop through each audio file path directly (no need for range indexes)
for file_path in audio_path:
    try:
        # Load the current audio file (sr=None preserves original sample rate)
        x, sr = librosa.load(file_path, sr=None)
        
        # Compute STFT and convert to dB scale (same as your single-file logic)
        X = librosa.stft(x)
        Xdb = librosa.amplitude_to_db(abs(X))
        
        # Set up the figure
        plt.figure(figsize=(14, 5))
        librosa.display.specshow(Xdb, sr=sr, x_axis='time', y_axis='hz')
        plt.colorbar(format='%+2.0f dB')
        plt.title(f"Spectrogram: {os.path.basename(file_path)}")
        
        # Save the spectrogram to a file (use original filename to stay organized)
        output_filename = f"{os.path.splitext(os.path.basename(file_path))[0]}_spectrogram.png"
        # Optional: Specify a dedicated output folder
        # output_path = os.path.join("path/to/save/spectrograms", output_filename)
        plt.savefig(output_filename, bbox_inches='tight')
        
        # Close the figure to free up memory (essential for large batches!)
        plt.close()
        
        print(f"Processed: {file_path}")
    
    except Exception as e:
        # Handle errors (like corrupted files) so the loop doesn't break
        print(f"Failed to process {file_path}: {str(e)}")

Quick Tips #

• Path Organization: Use os.path.join to handle file paths seamlessly across Windows/macOS/Linux
• Filter Formats: Use librosa.util.find_files with ext=["wav", "mp3"] to target specific audio types if your folder has mixed files
• Logarithmic Axis: Uncomment the y_axis='log' line in specshow if you want a log-scale frequency axis—this is common for music analysis
• Progress Tracking: For large datasets, add a progress bar with the tqdm library to keep tabs on processing speed

内容的提问来源于stack exchange，提问作者Atharva Athaley