helperFunctions.py

import os
import mutagen
import mutagen.wave
import pandas as pd

def create_dataset_df(csv_file):
    dataset_df = pd.read_csv(csv_file)
    filepaths = []
    for i, row in dataset_df.iterrows():
        filepaths.append(os.path.join('UrbanSound8K/audio', 'fold'+str(row['fold']), row['slice_file_name']))
    dataset_df['filepath'] = filepaths
    return dataset_df


def compute_audio_statistics(dataset_df):
    metadata_dict = {'length': [], 'bitrate': [], 'channels': [], 'sample_rate': [], 'bits_per_sample': []}
    # Extract metadata
    for filepath in dataset_df['filepath']:
        metadata = get_audio_metadata_mutagen(filepath)
        for key in metadata_dict.keys():
            metadata_dict[key].append(metadata[key])
    # Add new columns to dataframe
    for key in metadata_dict.keys():
        dataset_df[key] = metadata_dict[key]

    return dataset_df


def get_audio_metadata_mutagen(filepath):
    metadata = {}
    f = mutagen.wave.WAVE(filepath)
    metadata['length'] = f.info.length
    metadata['bitrate'] = f.info.bitrate
    metadata['channels'] = f.info.channels
    metadata['sample_rate'] = f.info.sample_rate
    metadata['bits_per_sample'] = f.info.bits_per_sample
    return metadata