def train_model_dir(quantity_file=5):
"""
function train_model_dir
"""
print("Bắt đầu training")
file_paths = open(c.FILE_NAME_TRAIN, 'r')
features = np.asarray(())
count = 1
for path in file_paths:
path = path.strip()
print(path)
sr, audio = read(c.TRAIN_SET + path)
vector = ef(audio, sr)
if features.size == 0:
features = vector
else:
features = np.vstack((features, vector))
if count == quantity_file:
gmm = GaussianMixture(n_components=6,
max_iter=200,
covariance_type='diag',
n_init=10)
gmm.fit(features)
picklefile = path.split('-')[0] + '.gmm'
pickle.dump(gmm, open(c.TRAINED_MODELS + picklefile, 'wb'))
features = np.asarray(())
count = 0
count += 1
def train_model(quantity_file=5, record=True):
if record == True:
record_audio(quantity_file, c.RECORD_SECONDS, c.FILE_NAME_TRAIN,
c.TRAIN_SET)
print("Bắt đầu training")
file_paths = open(c.FILE_NAME_TRAIN, 'r')
features = np.asarray(())
count = 1
for path in file_paths:
path = path.strip()
print(path)
sr, audio = read(c.TRAIN_SET + path)
vector = ef(audio, sr)
if features.size == 0:
features = vector
else:
features = np.vstack((features, vector))
if count == c.QUANTITY_TRAIN_FILE:
gmm = GaussianMixture(n_components=6,
max_iter=200,
covariance_type='diag',
n_init=10)
gmm.fit(features)
picklefile = path.split('-')[0] + '.gmm'
pickle.dump(gmm, open(c.TRAINED_MODELS + picklefile, 'wb'))
features = np.asarray(())
count = 0
count += 1
def test_list_file_from_dir():
"""
lấy tất cả các file từ thư mục config.TEST_SET == 'testing_set'
:return:
"""
# engine = pyttsx3.init()
# voices = engine.getProperty('voices')
# engine.setProperty("voice", voices[1].id)
print("Bắt đầu kiểm tra")
gmm_files = [
os.path.join(c.TRAINED_MODELS, file_name)
for file_name in os.listdir(c.TRAINED_MODELS)
if file_name.endswith('.gmm')
]
models = [pickle.load(open(file_name, 'rb')) for file_name in gmm_files]
speakers = [
file_name.split('\\')[-1].split(".gmm")[0] for file_name in gmm_files
]
cols = ['file_name', 'test_speaker', 'scores', 'speaker', 'result']
df_result = pd.DataFrame(columns=cols)
file_paths = os.listdir(c.TEST_SET)
for file_path in file_paths:
paths = os.listdir(c.TEST_SET + file_path)
print(file_path)
for path in paths:
sr, audio = read(c.TEST_SET + file_path + '/' + path)
vector = ef(audio, sr)
log_likelihood = np.zeros(len(models))
for i in range(len(models)):
gmm = models[i]
scores = np.array(gmm.score(vector))
log_likelihood[i] = scores.sum()
winner = np.argmax(log_likelihood) # trained_models/g
speaker = speakers[winner].split('/')[-1]
test_speaker = path.split('-')[0]
if speaker == test_speaker:
result = 1
else:
result = 0
# content = f"{path} la {speaker} nói"
# engine.say(content)
new_data = [path, test_speaker, scores, speaker, result]
new_df = pd.DataFrame([new_data], columns=cols)
df_result = pd.concat([df_result, new_df], ignore_index=True)
# engine.runAndWait()
df_result.to_csv('res/result.csv')
print("Sumit successfully!result to save dir res/result.csv")
def test_one_file_dir(file_name):
"""
chạy 1 file bất kỳ từ đường dẫn truyền vào
:param file_name:đường dẫn file
:return: void
"""
# file_path = record_one_file(file_name)
engine = pyttsx3.init()
voices = engine.getProperty('voices')
engine.setProperty("voice", voices[1].id)
print("bắt đầu kiểm tra")
gmm_files = [
os.path.join(c.TRAINED_MODELS, file_name)
for file_name in os.listdir(c.TRAINED_MODELS)
if file_name.endswith('.gmm')
]
models = [pickle.load(open(file_name, 'rb')) for file_name in gmm_files]
speakers = [
file_name.split('\\')[-1].split(".gmm")[0] for file_name in gmm_files
]
file_path = file_name.strip()
sr, audio = read(file_path)
vector = ef(audio, sr)
log_likelihood = np.zeros(len(models))
for i in range(len(models)):
gmm = models[i]
scores = np.array(gmm.score(vector))
log_likelihood[i] = scores.sum()
winner = np.argmax(log_likelihood)
speaker = speakers[winner].split('/')[-1]
content = f"file được truyền vào là {speaker} nói"
engine.say(content)
engine.runAndWait()
print(f"{file_name} la {speaker} nói")
def test_list_file(quantity_file=2, record=False):
if record == True:
record_audio(quantity_file, c.RECORD_SECONDS, c.FILE_NAME_TEST,
c.TEST_SET, True)
print("Bắt đầu kiểm tra")
gmm_files = [
os.path.join(c.TRAINED_MODELS, file_name)
for file_name in os.listdir(c.TRAINED_MODELS)
if file_name.endswith('.gmm')
]
models = [pickle.load(open(file_name, 'rb')) for file_name in gmm_files]
speakers = [
file_name.split('\\')[-1].split(".gmm")[0] for file_name in gmm_files
]
file_paths = open(c.FILE_NAME_TEST, 'r')
cols = ['file_name', 'test_speaker', 'scores', 'speaker']
df_result = pd.DataFrame(columns=cols)
# test_list_file(file_paths, models, speakers)
for path in file_paths:
path = path.strip()
sr, audio = read(c.TEST_SET + path)
vector = ef(audio, sr)
log_likelihood = np.zeros(len(models))
for i in range(len(models)):
gmm = models[i]
scores = np.array(gmm.score(vector))
log_likelihood[i] = scores.sum()
winner = np.argmax(log_likelihood)
speaker = speakers[winner].split('/')[-1]
# str = f"{path} la {speaker} nói"
test_speaker = path.split('-')[0]
new_data = [path, test_speaker, scores, speaker]
new_df = pd.DataFrame([new_data], columns=cols)
df_result = pd.concat([df_result, new_df], ignore_index=True)
df_result.to_csv('res/result.csv')
print("Sumit successfully!result to save dir res/result.csv")
def train_model_more_dir():
"""
function train model
"""
print("Bắt đầu training")
features = np.asarray(())
count = 1
paths = os.listdir(c.TRAIN_SET)
step = 0
for path in paths:
print(path)
list_file = os.listdir(c.TRAIN_SET + path)
for path_file in list_file:
sr, audio = read(c.TRAIN_SET + path + '/' + path_file)
vector = ef(audio, sr)
if features.size == 0:
features = vector
else:
features = np.vstack((features, vector))
if count == len(list_file):
gmm = GaussianMixture(n_components=6,
max_iter=200,
covariance_type='diag',
n_init=10)
gmm.fit(features)
picklefile = path_file.split('-')[0] + '.gmm'
pickle.dump(gmm, open(c.TRAINED_MODELS + picklefile, 'wb'))
features = np.asarray(())
count = 0
count += 1