def get_wav_and_labels(filename, data_dir):
# folder = filename[:-9]
wav_path = data_dir + "/audio/" + filename
label_path = data_dir + "/Annotations/" + filename[:-9] + ".txt"
with open(label_path, 'r') as label_file:
category = ""
dimensions = ""
speaker = ""
for row in label_file:
if row[0] == '[':
split = row.split("\t")
if split[1] == filename[:-4]:
category = get_emotion_from_label(split[2])
dimensions = cont2list(split[3])
dimensions_dis = cont2list(split[3], binned=True)
speaker = get_speaker_from_filename(filename)
audio = audio_utils.load_wav(wav_path)
audio = np.array(audio, dtype=np.float32)
labels = concatenate_labels(category, speaker, dimensions, dimensions_dis)
return audio, labels
def get_samples_and_labels(filename, config):
# config = yaml.load(open('./config.yaml', 'r'))
wav_path = config['data']['sample_set_dir'] + "/" + filename
folder = filename[:-9]
label_path = config['data'][
'dataset_dir'] + "/Annotations/" + folder + ".txt"
with open(label_path, 'r') as label_file:
category = ""
dimensions = ""
speaker = ""
for row in label_file:
if row[0] == '[':
split = row.split("\t")
if split[1] == filename[:-4]:
category = get_emotion_from_label(split[2])
dimensions = cont2list(split[3])
dimensions_dis = cont2list(split[3], binned=True)
speaker = get_speaker_from_filename(filename)
audio = audio_utils.load_wav(wav_path)
audio = np.array(audio, dtype=np.float32)
labels = concatenate_labels(category, speaker, dimensions, dimensions_dis)
return audio, labels
# for f in filenames:
# print(f[:-4])
# in_dir = '../data/labels'
# files = find_files(in_dir, ext = 'npy')
# filenames = [os.path.basename(f)[:-4] + ".wav" for f in files]
# print("Found", len(filenames), " files.")
#
# filenames = [f for f in filenames if pp.get_wav_and_labels(f, config['data']['dataset_dir'])[1][1] in range(0,6)]
# random.shuffle(filenames)
# filenames = filenames[:10]
# print(filenames)
# print("Number of files to be converted = ", len(filenames))
filenames = ["../data/stairway.wav"]
wav = audio_utils.load_wav(filenames[0])
labels = [3, 1, 0,0,0,0,0,0]
########################################
# BASELINE SYNTHESIS LOOP #
########################################
# for f in filenames:
#
# wav, labels = pp.get_wav_and_labels(f, config['data']['dataset_dir'])
# wav = np.array(wav, dtype = np.float64)
# labels = np.array(labels)
# f0, ap, sp, coded_sp = preprocess_world.cal_mcep(wav)
# coded_sp = coded_sp.T
# # coded_sp_temp = np.copy(coded_sp).T
# # print(coded_sp_temp.shape)
# filename_wav = f[0:-4] + "_" + str(int(labels[0].item())) + ".wav"
# print(coded_sp.shape)
import librosa
import audio_utils
import numpy as np
dir = "../datasets/IEMOCAP/All"
filenames = librosa.util.find_files(dir)
filename = filenames[0]
audio = audio_utils.load_wav(filename)
print(audio.shape)
audio = np.array(audio, dtype=np.float32)
mel = audio_utils.wav2melspectrogram(audio)
np.save('./samples/Test/mel.npy', mel)