def gen_empty_ratings(folder):
split = os.path.basename(folder).split('_')
vp = split[-1]
session_id = split[-2]
wav_files = np.array(glob.glob(os.path.join(folder, '*.wav')))
wav_files = wav_files[snodgrass_audio_chronological_sort_indices(
wav_files)]
orders = snodgrass_audio2order(wav_files)
snodgrass_words = load_snodgrass_words()
if len(wav_files) != 233:
msg = 'Only {0} files in session folder {1} -- non-full sessions currently not supported for csv export'.format(
len(wav_files), folder)
print(msg)
raise RuntimeError(msg)
fake_ratings: List[SnodgrassWordRating] = []
for wav_file, order in zip(wav_files, orders):
word = snodgrass_words[order - 1]
fake_ratings.append(
SnodgrassWordRating(word, order, float('nan'), float('nan'),
float('nan'), '', '', vp, session_id,
wav_file))
return fake_ratings
def get_emu_word_counts(exclude_snodgrass=True,
cache_path='counts_per_db.pckl'):
def emu2word_counts_except(db, words):
db_path = os.path.join(raw_data_dir, db)
seq_rds_path = os.path.join(raw_data_dir, '{0}.rds'.format(db))
return emu2word_counts(db,
db_path,
seq_rds_path,
partial(exclude_words, words_to_exclude=words),
verbose=True)
snodgrass_words = load_snodgrass_words() if exclude_snodgrass else None
out_path = cache_path
if not os.path.exists(out_path):
emus = filter(
lambda x: os.path.isdir(os.path.join(raw_data_dir, x)) and x.
endswith('emuDB'), os.listdir(raw_data_dir))
skip_dbs = ['BROTHERS_emuDB']
counts_per_db = {}
for emu in emus:
if emu not in skip_dbs:
counts = emu2word_counts_except(emu, snodgrass_words)
counts_per_db[emu] = counts
with open(out_path, 'wb') as f:
pickle.dump(counts_per_db, f)
return counts_per_db
else:
return load_pickled(out_path)
def get_swc_word_counts(exclude_snodgrass=True):
if not os.path.exists(aligned_words_file):
collect_aligned_words(verbose=False)
with open(aligned_words_file, 'rb') as f:
words_dict = pickle.load(f)
snodgrass_words = load_snodgrass_words() if exclude_snodgrass else None
return filter_words_dict(words_dict,
word_filter=partial(
exclude_words,
words_to_exclude=snodgrass_words))
def __main():
if not os.path.exists(aligned_words_file):
collect_aligned_words(verbose=True)
with open(aligned_words_file, 'rb') as f:
words_dict = pickle.load(f)
snodgrass_words = load_snodgrass_words()
snodgrass_words_dict = filter_words_dict(
words_dict,
word_filter=partial(select_words, words_to_keep=snodgrass_words))
analyze_aligned_words(words_dict, snodgrass_words_dict)
analyze_aligned_words(words_dict, filter_words_dict(words_dict))
collect_swc_features('snodgrass_words_cleaned_v3', 'snodgrass_words',
snodgrass_words)
seq_rds_path,
feature_func,
partial(select_words, words_to_keep=words),
verbose=True)
emus = filter(
lambda x: os.path.isdir(os.path.join(raw_data_dir, x)) and x.endswith(
'emuDB'), os.listdir(raw_data_dir))
skip_dbs = ['BROTHERS_emuDB']
for emu in emus:
if emu not in skip_dbs:
_process_emu_db(emu, processed_sub_dir, collection_name, word_list)
# _process_emu_db('PD1_emuDB')
if __name__ == '__main__':
snodgrass_words = load_snodgrass_words()
collect_emu_features('snodgrass_words_cleaned_v3', 'snodgrass_words',
snodgrass_words)
# collect_emu_features('snodgrass_words_cleaned_v3_mfcc', 'snodgrass_words', snodgrass_words, feature_func=audio2mfcc)
# collect_emu_features('snodgrass_words_cleaned_v3_reverse', 'snodgrass_words', snodgrass_words,
# feature_func=audio2lmfe_reverse)
# collect_emu_features('snodgrass_words_cleaned_v3_pncc_mn_off_dct_off', 'snodgrass_words', snodgrass_words,
# feature_func=partial(audio2pncc, mean_norm=False, do_dct=False))
# collect_emu_features('test_ahm', 'test_ahm', ['<ähm>'], feature_func=None, debug=True)
# with open(os.path.join(res_dir, 'new_words.txt'), 'r') as f:
# new_words = [line.rstrip('\n') for line in f]
# collect_emu_features('new_words', 'new_words', new_words)