From 967a89cc22a896470102252764f35a27928502cd Mon Sep 17 00:00:00 2001
From: wuliaozhiji <youquzhiji@gmail.com>
Date: Sat, 12 Mar 2022 21:15:40 -0500
Subject: [PATCH] [+] Calculate formants on vox_celeb

---
 src/formant.py | 80 ++++++++++++++++++++++++++++++++++++++++++++++++--
 1 file changed, 78 insertions(+), 2 deletions(-)

diff --git a/src/formant.py b/src/formant.py
index 87817c2..54c5eb1 100644
--- a/src/formant.py
+++ b/src/formant.py
@@ -1,5 +1,12 @@
 from __future__ import annotations
 
+import csv
+import os
+from multiprocessing import Pool
+from os import PathLike
+from pathlib import Path
+
+import tqdm
 import matplotlib.pyplot as plt
 import numpy
 import parselmouth
@@ -31,6 +38,75 @@ def calculate_freq_info(audio: parselmouth.Sound, show_plot=False) -> numpy.ndar
     return result
 
 
+def load_vox_celeb_asab_dict(path: PathLike) -> dict[str, str]:
+    """
+    Load voxCeleb 1 or 2's metadata to gather a dictionary mapping id to assigned sex at birth.
+
+    :param path: CSV path (Tab separated)
+    :return: {id: ASAB}
+    """
+    with open(path, 'r', newline='') as f:
+        return {row[0]: row[2] for row in csv.reader(f, delimiter='\t') if row[0].startswith('id')}
+
+
+def compute_vox_celeb_helper(aud_dir: str):
+    """
+    Compute one audio file
+
+    :param aud_dir: Audio file path
+    :return: None
+    """
+    array = calculate_freq_info(parselmouth.Sound(aud_dir))
+    numpy.save(aud_dir, array)
+
+
+def compute_vox_celeb():
+    vox_celeb_dir = Path('C:/Workspace/EECS 6414/Datasets/VoxCeleb1/wav')
+    audio_suffix = 'wav'
+
+    print('Finding audio files...')
+
+    asab = load_vox_celeb_asab_dict(vox_celeb_dir.joinpath('../vox1_meta.csv'))
+    queue: list[str] = []
+
+    # Loop through all ids
+    for id in asab:
+        id_dir = vox_celeb_dir.joinpath(id)
+
+        # Check if directory exists
+        if not id_dir.is_dir():
+            continue
+
+        # Loop through all videos
+        for vid in os.listdir(id_dir):
+            vid_dir = id_dir.joinpath(vid)
+
+            # Check if it's a video directory
+            if not vid_dir.is_dir():
+                continue
+
+            # Loop through all audios
+            for aud in os.listdir(vid_dir):
+                aud_dir = vid_dir.joinpath(aud)
+
+                # Check if end with suffix
+                if not aud.endswith(audio_suffix):
+                    continue
+
+                # Add to queue
+                queue.append(str(aud_dir))
+
+    print(f'There are {len(queue)} audio files to process.')
+    print('Starting processing...')
+
+    # Compute audio files in a cpu pool
+    with Pool(8) as pool:
+        for _ in tqdm.tqdm(pool.imap(compute_vox_celeb_helper, queue), total=len(queue)):
+            pass
+
+
 if __name__ == '__main__':
-    print(calculate_freq_info(parselmouth.Sound('D:/Downloads/Vowels-Extract-Z-44kHz.flac')))
-    print(calculate_freq_info(parselmouth.Sound('D:/Downloads/Vowels-Azalea.flac')))
+    compute_vox_celeb()
+    # print(calculate_freq_info(parselmouth.Sound('D:/Downloads/Vowels-Extract-Z-44kHz.flac')))
+    # print(calculate_freq_info(parselmouth.Sound('D:/Downloads/Vowels-Azalea.flac')))
+