def get_desired_features():
print("With which features do you want to train the model? (e.g. MFCC, FBANK, MFCC and pitch, FBANK and deltas)")
features = input()
if "exit" in features.lower():
raise ExitApp()
print("\nYou entered: \n\n{}\n\nIs this correct? (Y/N)".format(features))
correct = input()
if "exit" in correct.lower():
raise ExitApp()
elif "y" not in correct.lower():
features = get_desired_features()
return features
def get_label_column():
print("What is the column name for the label? (e.g. word, gender, class)")
label_column = input()
if "exit" in label_column.lower():
raise ExitApp()
print("You entered ' {} '. Is that correct? (Y/N)".format(label_column))
correct = input()
if "y" in correct.lower():
pass
elif "exit" in correct.lower():
raise ExitApp()
else:
label_column = get_label_column()
return label_column
def get_feature_type():
print("What kind of features will you extract? (e.g. mfcc, fbank, fbank with deltas, mfcc with pitch values)")
features = input()
if "exit" in features.lower():
raise ExitApp()
print("You entered {}".format(features))
return features
def go():
print("Press ENTER to continue.")
cont = input()
if "exit" in cont.lower():
raise ExitApp("Have a good day!")
elif cont != "":
print("Type 'exit' to end the program.")
cont = go()
return True
def login(self,username):
print("\nEnter your password to access your lists.\n")
password = self.get_password()
if 'exit' == password.lower():
raise ExitApp("Aw man. You're leaving already? Come back soon!")
if password:
match = self.check_password(username, password)
return match
return None
def get_label_data_type():
print("What type of data will the label be saved in the table? (i.e. 'INT', or 'TEXT')")
data_type = input()
type_options = ["INT","INTEGER","TEXT"]
if data_type.upper() not in type_options:
print("Please enter one of the following: {}".format(", ").join(type_options))
data_type = label_data_type()
elif "exit" in data_type.lower():
raise ExitApp()
return data_type
def add_noise():
print("Will you add noise? (Y/N)")
noise = input()
if "y" in noise.lower():
noise = True
elif "n" in noise.lower():
noise = False
elif "exit" in noise.lower():
raise ExitApp()
return noise
def set_limit():
print("Is there a limit for the data? If YES: enter an INTEGER.")
limit = input()
if "exit" in limit.lower():
raise ExitApp()
elif limit.isdigit():
limit = int(limit)
else:
limit = None
return limit
def get_num_features():
print("How many feature columns do you need? (If you want the derivatives, DON'T INCLUDE THE COLUMNS FOR THOSE VALUSE HERE)")
num_features = input()
if num_features.isdigit():
pass
elif "exit" in num_features.lower():
raise ExitApp()
else:
print("Please enter an integer")
num_features = get_num_features()
return int(num_features)
def get_username(self):
print("\nUsername: "******"Spaces and special characters will be removed: ")
username = rem_space_specialchar(username)
if 'exit' == username.lower():
raise ExitApp("Aw man. I didn't even catch your name. Come back soon!")
if username:
return username
else:
print("Not enough alphanumeric characters used. Try again.")
self.get_username()
return None
def load_data(database,table,columns=None):
if columns is None:
columns = "all"
print("Loading data from {} columns from {} in database: {}".format(columns,table,database))
limit = set_limit()
print("Press ENTER to continue")
cont = input()
if "exit" in cont.lower():
raise ExitApp()
elif cont.lower() == "":
pass
else:
cont = load_data(database,table)
data = select_data(database,table,limit)
return data
from errors import ExitApp
from user_account import LoginUser
from user_data_babyname import UserData
if __name__ == "__main__":
database = "babynames_USA.db"
num_clusters = 30
num_features = 5 #selected features to define clusters
try:
login = None
ud = None
login = LoginUser(database)
login.sign_in()
if login.is_user != True:
print("Login failed.")
raise ExitApp("UserId not found.")
user_id = login.get_userid(login.username)
if user_id is None:
print("User ID Error")
raise SystemExit
ud = UserData(database,
user_id,
num_clusters,
num_features=num_features)
ud.main_menu()
except ExitApp:
print("Have a good day!")
except SystemExit as e:
print("Problem occurred: {}. Had to close app.".format(e))
except Exception as e:
print(e)
def main(script_purpose, split=False):
current_filename = os.path.basename(__file__)
session_name = get_date(
) #make sure this session has a unique identifier - link to model name and logging information
path_to_data = "./data_3word"
start = time.time()
try:
start_logging(script_purpose)
logging.info("Running script: {}".format(current_filename))
logging.info("Session: {}".format(session_name))
tablename, feature_type, num_features, num_feature_columns, noise = user_input.set_variables(
)
paths, labels = featfun.collect_audio_and_labels(path_to_data)
noise_path = "./_background_noise_/doing_the_dishes.wav"
label_list = [i[1] for i in labels]
class_labels = list(set(label_list))
print("The labels found: {}".format(class_labels))
#find out number of recordings from each class
#ask user if they want to balance out the data?
dict_class_distribution = featfun.get_class_distribution(
class_labels, label_list)
min_val = (1000000, None)
for key, value in dict_class_distribution.items():
if value < min_val[0]:
min_val = (value, key)
print("Number of wave files for each class:\n\n{}\n\n".format(
dict_class_distribution))
print("Chosen max number of files from each class = {}".format(
min_val[0]))
print("\n\nDo you approve this? (Y/N)")
approve = input()
if 'exit' in approve.lower():
raise ExitApp()
elif 'y' in approve.lower():
pass
elif 'n' in approve.lower():
print("Okay... woring on that functionality")
raise ExitApp()
else:
raise ExitApp()
max_num_per_class = min_val[0]
#create dictionary w indices to labels and paths for each class
dict_class_index_list = {}
for label in class_labels:
dict_class_index_list[label] = []
for i, label_item in enumerate(label_list):
if label == label_item:
dict_class_index_list[label].append(i)
#get num per training/validation/test datasets
max_nums_train_val_test = get_max_nums_train_val_test(
max_num_per_class)
#randomly assign indices to train, val, test datasets:
dict_class_dataset_index_list = {}
for label in class_labels:
tot_indices = dict_class_index_list[label]
tot_indices_copy = tot_indices.copy()
random.shuffle(tot_indices_copy)
train_indices = tot_indices_copy[:max_nums_train_val_test[0]]
val_indices = tot_indices_copy[
max_nums_train_val_test[0]:max_nums_train_val_test[0] +
max_nums_train_val_test[1]]
test_indices = tot_indices_copy[
max_nums_train_val_test[0] +
max_nums_train_val_test[1]:max_nums_train_val_test[0] +
max_nums_train_val_test[1] + max_nums_train_val_test[2]]
dict_class_dataset_index_list[label] = [
train_indices, val_indices, test_indices
]
print()
print("Name for directory to save feature images:")
new_directory = input()
if 'exit' in new_directory.lower():
raise ExitApp()
train_val_test_dirs = []
for i in ["train", "val", "test"]:
train_val_test_dirs.append(new_directory + "_{}".format(i))
start_feature_extraction = time.time()
for label in class_labels:
for i, directory in enumerate(train_val_test_dirs):
if i == 0:
train = True
else:
train = False
try:
os.makedirs(directory)
except OSError as e:
if e.errno != errno.EEXIST:
raise
dict_new_paths = {}
new_path = './{}/{}/'.format(directory, label)
try:
os.makedirs(new_path)
dict_new_paths[label] = new_path
except OSError as e:
if e.errno != errno.EEXIST:
raise
#limit = int(max_nums_train_val_test[i]*.3)
limit = None
num_pics = max_nums_train_val_test[i]
msg = "\nExtracting features from {} samples. \nImages will be saved in the directory {}".format(
num_pics, new_path)
print(msg)
logging.info(msg)
frame_width = 11
time_step = 6
logging.info(
"extracting features from wavefiles. Limit = {}".format(
limit))
paths_list_dataset = []
labels_list_dataset = []
train_val_test_index_list = dict_class_dataset_index_list[
label]
for k in train_val_test_index_list[i]:
paths_list_dataset.append(paths[k])
labels_list_dataset.append(label_list[k])
print("Extracting features from class: {}".format(label))
for j, wav in enumerate(paths_list_dataset):
if limit:
if j <= limit:
featfun.save_chroma(
wav,
split,
frame_width,
time_step,
feature_type,
num_features,
num_feature_columns,
noise,
dict_new_paths[labels_list_dataset[j]],
train,
noise_path,
vad=True,
add_noise=True)
else:
featfun.save_chroma(
wav,
split,
frame_width,
time_step,
feature_type,
num_features,
num_feature_columns,
noise,
dict_new_paths[labels_list_dataset[j]],
train,
noise_path,
vad=True,
add_noise=True)
end_feature_extraction = time.time()
logging.info("Duration setup: {} minutes".format(
round((start_feature_extraction - start) / 60, 2)))
print("Duration of feature extraction: {} minutes".format(
round((end_feature_extraction - start_feature_extraction) / 60,
2)))
except ExitApp:
print("Have a good day!")
logging.info("User exited app.")
except FeatureExtractionError as e:
logging.exception("Error occurred in feature extraction: {}".format(e))
except Exception as e:
logging.exception("Error occurred: {}".format(e))
finally:
end = time.time()
duration = round((end - start) / 60, 2)
logging.info("Duration: {} minutes".format(duration))
color_scale = 1
sampling_rate = 16000
noise_filename = "background_noise.wav"
if sex == "female":
speech_filename = "speaker_aislyn_satz.wav"
speech_nr_filename = "speech_modelready_aislyn11.wav"
else:
speech_filename = "speaker_thomas_satz.wav"
speech_nr_filename = "speech_modelready_thomas9.wav"
print("Testing classifier with {} speech:".format(sex))
print("Press ENTER to start")
ready = input()
if ready != "":
raise ExitApp()
#background_noise = speech_collection.record(5,sampling_rate)
#speech_collection.save_recording("noise_trash.wav",background_noise,sampling_rate)
#print("Please press Enter and then say loudly and clearly: \n\n'Hallo, wie geht es Ihnen?'\n")
#start = input()
#if start != "":
# raise ExitApp()
#speech = speech_collection.record(5,sampling_rate)
#print("\nNow saving..")
#speech_collection.save_recording("speech_trash.wav",speech,sampling_rate)
print("\nNow extracting features..\n")
y_speech, sr = librosa.load(speech_filename, sr=sampling_rate)
#y_noise, sr = librosa.load("background_noise.wav",sr=sampling_rate)