preprocess.py

import os
import json
import librosa
import math

DATASET_PATH = "./archive/Data/genres_original"
JSON_PATH = "DATA.json"

SAMPLE_RATE = 22050

DURATION = 30 # in seconds

SAMPLES_PER_TRACK = 22050 * DURATION

def save_mfcc(dataset_path, json_path, n_mfcc=13, n_fft=2048, hop_length=512, num_segments=5):
    
    # dict to store data
    data = {
        "mapping": [],
        "mfcc": [],
        "labels": []
    }
    
    num_samples_per_segment = SAMPLES_PER_TRACK // num_segments
    
    expected_num_mfcc_vectors_per_segment = math.ceil(num_samples_per_segment / hop_length) # 1.2 => 2
    
    # loop through all genres
    for i, (dirpath, dirnames, filenames) in enumerate(os.walk(dataset_path)):
        
        # ensure we are not at root directory
        if dirpath is not dataset_path:
            
            # save semantic label
            dirpath_components = dirpath.split("\\") # genre\\blues => ['genre', 'blues']
            sematic_label = dirpath_components[-1] # blues
            data["mapping"].append(sematic_label)
            
            print("Processing {}".format(sematic_label))
            
            # loop through all files in genre
            for f in filenames:
                
                # load audio file
                file_path = os.path.join(dirpath, f)
                try:
                    signal, sr = librosa.load(file_path, sr=SAMPLE_RATE)
                    
                    # process segments extracting mfcc and storing data
                    for s in range(num_segments):
                        start_sample = num_samples_per_segment * s # s=0 => 0
                        finish_sample = start_sample + num_samples_per_segment # s=0 => num_samples_per_segment
                        
                        
                        mfcc = librosa.feature.mfcc(signal[start_sample:finish_sample],
                                                    sr=sr,
                                                    n_fft=n_fft,
                                                    n_mfcc=n_mfcc,
                                                    hop_length=hop_length)
                        
                        mfcc = mfcc.T
                        
                        # store mfcc for segment if it has expected length
                        if len(mfcc) == expected_num_mfcc_vectors_per_segment:
                            data["mfcc"].append(mfcc.tolist())
                            data["labels"].append(i-1)
                except Exception as e:
                    print("Error processing file {}".format(file_path))
                    print(e)
                        
    # save data to json file
    with open(json_path,"w+", encoding="utf-8") as fp:
        json.dump(data, fp, indent=4)
        
if __name__ == "__main__":
    save_mfcc(DATASET_PATH, JSON_PATH, num_segments=10)