def Train(save_model_name: str):
Config.save_model_name = save_model_name
x_train, x_test, y_train, y_test = Opensmile_Feature.get_data(Config.DATA_PATH, Config.TRAIN_FEATURE_PATH_OPENSMILE, train=True)
model = SVM_Model()
model.train(x_train, y_train)
model.evaluate(x_test, y_test)
model.save_model(save_model_name)
return model
def Train(save_model_name: str):
Config.save_model_name = save_model_name
x_train, x_test, y_train, y_test = Librosa_Feature.get_data(
Config.DATA_PATH, Config.TRAIN_FEATURE_PATH_LIBROSA, train=True)
model = SVM_Model()
model.train(x_train, y_train)
model.evaluate(x_test, y_test)
model.save_model(save_model_name)
return model
def Train(model_name: str,
save_model_name: str,
if_load: bool = True,
feature_method: str = 'opensmile'):
# 提取特征
print(f'if_load: {if_load}')
if (feature_method == 'o'):
# 使用openSmile提取特征
if (if_load == True):
x_train, x_test, y_train, y_test = of.load_feature(
feature_path=Config.TRAIN_FEATURE_PATH_OPENSMILE, train=True)
else:
x_train, x_test, y_train, y_test = of.get_data(
Config.DATA_PATH,
Config.TRAIN_FEATURE_PATH_OPENSMILE,
train=True)
elif (feature_method == 'l'):
if (if_load == True):
x_train, x_test, y_train, y_test = lf.load_feature(
feature_path=Config.TRAIN_FEATURE_PATH_LIBROSA, train=True)
else:
print('here-------------------------')
x_train, x_test, y_train, y_test = lf.get_data(
Config.DATA_PATH,
Config.TRAIN_FEATURE_PATH_LIBROSA,
train=True)
# 创建模型
if (model_name == 'svm'):
model = SVM_Model()
elif (model_name == 'mlp'):
model = MLP_Model()
else:
y_train = np_utils.to_categorical(y_train)
y_val = np_utils.to_categorical(y_test)
model = LSTM_Model(input_shape=x_train.shape[1],
num_classes=len(Config.CLASS_LABELS))
# 二维数组转三维（samples, time_steps, input_dim）
x_train = np.reshape(x_train, (x_train.shape[0], 1, x_train.shape[1]))
x_test = np.reshape(x_test, (x_test.shape[0], 1, x_test.shape[1]))
# 训练模型
print(
'-------------------------------- Start --------------------------------'
)
if (model_name == 'svm' or model_name == 'mlp'):
model.train(x_train, y_train)
else:
print(x_train.shape)
print(y_train.shape)
print(x_test.shape)
print(y_val.shape)
model.train(x_train, y_train, x_test, y_val, n_epochs=Config.epochs)
model.evaluate(x_test, y_test)
model.save_model(save_model_name)
print(
'---------------------------------- End ----------------------------------'
)
return model
# When using SVM, _svm = True, or _svm = False
#asdf = get_feature_svm("/Users/cemergin/PycharmProjects/Speech-Emotion-Recognition/Models/DataSet/RAVDESSson/angry/03-01-05-01-01-01-01.wav", 39)
#numpy.savetxt("asdf.txt",asdf)
#print("yazdım")
#print(asdf)
x_train, x_test, y_train, y_test = get_data(DATA_PATH, _svm=True)
#print(len(x_train))
#print(len(x_test))
#print(x_train.shape)
#print(x_test.shape)
#print(x_train)
#print(x_test)
numpy.savetxt("x_train.txt", x_train)
numpy.savetxt("x_test.txt", x_test)
numpy.savetxt("y_train.txt", y_train)
numpy.savetxt("y_test.txt", y_test)
model_svm = SVM_Model()
model_svm.save_model("svmmodeli")
model_svm.train(x_train, y_train)
model_svm.evaluate(x_test, y_test)
#model.recognize_one(get_feature("/Users/cemergin/PycharmProjects/Speech-Emotion-Recognition/Models/DataSet/RAVDESS/angry/03-01-05-01-01-01-01.wav"))
#a = get_feature("/Users/cemergin/PycharmProjects/audio-classification/data/03a01Nc.wav",39,flatten=1)
#print(a)
#print(type(a))
#print(len(a))
#numpy.savetxt("feature.txt",a)
#-36.04365339 03a01nc
#1134 -5.689264606224537779e-01 fa