.gitignore

@@ -1,4 +1,4 @@
-data
+Exp20201205_2myo_hard**
-docs
+Exp20201205_2myo_soft**
-logs
+Documents
-psf_lib
+python_speech_features**

Handle_emg_data.py

@@ -6,8 +6,8 @@ from pathlib import Path
 import numpy as np
 from pandas.core.frame import DataFrame
 import sys
-sys.path.insert(0, '/Users/Markus/Prosjekter git/Slovakia 2021/psf_lib/python_speech_features/python_speech_features')
+sys.path.insert(0, '/Users/Markus/Prosjekter git/Slovakia 2021/python_speech_features/python_speech_features')
-from psf_lib.python_speech_features.python_speech_features import mfcc
+from python_speech_features.python_speech_features import mfcc
 import json
 
 

Neural_Network_Analysis.py

@@ -1,19 +1,15 @@
 import json
-
+from python_speech_features.python_speech_features.base import mfcc
-from keras import callbacks
-from psf_lib.python_speech_features.python_speech_features.base import mfcc
 import numpy as np
 from sklearn.model_selection import train_test_split
 import tensorflow as tf
 import tensorflow.keras as keras
 from keras import backend as K
 from keras.regularizers import l2
-from keras.callbacks import Callback, CSVLogger, ModelCheckpoint
 from pathlib import Path
 import pandas as pd
 import matplotlib.pyplot as plt
 import statistics
-import csv
 
 # Path to json file that stores MFCCs and subject labels for each processed sample
 DATA_PATH_MFCC = str(Path.cwd()) + "/mfcc_data.json"
@@ -162,16 +158,12 @@ def prepare_datasets_sessions(X, y, session_lengths, test_session_index=4, nr_su
 # Trains the model 
 # Input: Keras.model, batch_size, nr epochs, training, and validation data
 # Ouput: History
-def train( model, X_train, y_train, verbose, batch_size=64, epochs=30, 
+def train(model, X_train, y_train, verbose, batch_size=64, epochs=30, X_validation=None, y_validation=None):
-            X_validation=None, y_validation=None):
 
     optimiser = keras.optimizers.Adam(learning_rate=0.0001)
     model.compile(optimizer=optimiser,
                   loss='categorical_crossentropy',
                   metrics=['accuracy'])
-    
-    #csv_path = str(Path.cwd()) + '/logs/{}/{}_train_log.csv'.format(MODEL_NAME, MODEL_NAME)
-    #csv_logger = CSVLogger(csv_path, append=False)
 
     if X_validation != None:
         history = model.fit(X_train, 
@@ -241,66 +233,24 @@ def batch_formatting(X_train, X_test, y_train, y_test, batch_size=64, nr_classes
 
     return X_train_batch, X_test_batch, y_train_batch, y_test_batch
 
-# Retrieves data sets for each session as test set and evalutes. DOES USE prediction_csv_logger as default
+# Retrieves data sets for each session as test set and evalutes 
 # the average of networks trained om them
 # Input: raw data, session_lengths list, total nr of sessions, batch_size, and nr of epochs 
 # Ouput: tuple(cross validation average, list(result for each dataset(len=nr_sessions)))
-def session_cross_validation(model_name:str, X, y, session_lengths, nr_sessions, log_to_csv=True, batch_size=64, epochs=30):
+def session_cross_validation_LSTM(X, y, session_lengths, nr_sessions, batch_size=64, epochs=30):
     session_training_results = []
     for i in range(nr_sessions):
-
+        model = LSTM(input_shape=(1, 208))
         X_train_session, X_test_session, y_train_session, y_test_session = prepare_datasets_sessions(X, y, session_lengths, i)
-    
+        train(model, X_train_session, y_train_session, verbose=0, batch_size=batch_size, epochs=epochs)
-        # Model:
-        if model_name == 'LSTM':
-            model = LSTM(input_shape=(1, 208))
-
-        elif model_name == 'GRU':
-            model = GRU(input_shape=(1, 208))
-
-        elif model_name == 'CNN_1D':
-            X_train_session = np.reshape(X_train_session, (X_train_session.shape[0], 208, 1))
-            X_test_session = np.reshape(X_test_session, (X_test_session.shape[0], 208, 1))
-            model = CNN_1D(input_shape=(208, 1))
-
-        elif model_name == 'FFN':
-            model = FFN(input_shape=(1, 208))
-
-        else:
-            raise Exception('Model not found')
-
-        #model.summary()
-
-        
-        train(model, X_train_session, y_train_session, verbose=1, batch_size=batch_size, epochs=epochs)
         test_loss, test_acc = model.evaluate(X_test_session, y_test_session, verbose=2)
         session_training_results.append(test_acc)
-        if log_to_csv:
-            prediction_csv_logger(X_test_session, y_test_session, model_name, model, i)
         del model
         K.clear_session()
-        #print('Session', i, 'as test data gives accuracy:', test_acc)
+        print('Session', i, 'as test data gives accuracy:', test_acc)
-
     average_result = statistics.mean((session_training_results))
-
     return average_result, session_training_results
 
-# Takes in test data and logs input data and the prediction from a model
-# Input: raw data, session_lengths list, total nr of sessions, batch_size, and nr of epochs 
-# Ouput: tuple(cross validation average, list(result for each dataset(len=nr_sessions)))
-def prediction_csv_logger(X, y, model_name, model, session_nr):
-    
-    csv_path = str(Path.cwd()) + '/logs/{}/{}_session{}_log.csv'.format(model_name, model_name, session_nr+1)
-
-    layerOutput = model.predict(X, verbose=0)
-
-    with open(csv_path, 'w') as csv_file:
-        writer = csv.writer(csv_file)
-        writer.writerow(['input', 'prediction', 'solution'])
-        data = zip(X, layerOutput, y)
-        writer.writerows(data)
-        csv_file.close()
-    
 
 # ----- MODELS ------
 
@@ -309,9 +259,9 @@ def prediction_csv_logger(X, y, model_name, model, session_nr):
 # Ouput: model:Keras.model
 def LSTM(input_shape, nr_classes=5):
 
-    model = keras.Sequential(name='LSTM_model')
+    model = keras.Sequential()
-    model.add(keras.layers.Bidirectional(keras.layers.LSTM(128), input_shape=input_shape, name='Bidirectional_LSTM'))
+    model.add(keras.layers.Bidirectional(keras.layers.LSTM(64), input_shape=input_shape, name='Bidirectional_LSTM'))
-    model.add(keras.layers.Dense(128, activation='relu', activity_regularizer=l2(0.005), name='Dense_relu'))
+    model.add(keras.layers.Dense(64, activation='relu', activity_regularizer=l2(0.001), name='Dense_relu'))
     model.add(keras.layers.Dropout(0.3, name='Dropout'))
     # Output layer
     model.add(keras.layers.Dense(nr_classes, activation='softmax', name='Dense_relu_output'))
@@ -323,44 +273,12 @@ def LSTM(input_shape, nr_classes=5):
 # Ouput: model:Keras.model
 def GRU(input_shape, nr_classes=5):
 
-    model = keras.Sequential(name='GRU_model')
+    model = keras.Sequential()
-    model.add(keras.layers.Bidirectional(keras.layers.GRU(128), input_shape=input_shape, name='Bidirectional_GRU'))
+    model.add(keras.layers.Bidirectional(keras.layers.GRU(64), input_shape=input_shape, name='Bidirectional_GRU'))
-    model.add(keras.layers.Dense(128, activation='relu', activity_regularizer=l2(0.005), name='Dense_relu'))
+    model.add(keras.layers.Dense(64, activation='relu', activity_regularizer=l2(0.001), name='Dense_relu'))
     model.add(keras.layers.Dropout(0.3, name='Dropout'))
     # Output layer:
-    model.add(keras.layers.Dense(nr_classes, activation='softmax', name='Softmax'))
+    model.add(keras.layers.Dense(nr_classes, activation='softmax', name='Dense_relu_output'))
-
-    return model
-
-# Creates a keras.model with a basic feed-forward-network
-# Input: input shape, classes of classification
-# Ouput: model:Keras.model
-def FFN(input_shape, nr_classes=5):
-
-    model = keras.Sequential(name='FFN_model')
-    model.add(keras.layers.Reshape((input_shape[-1],), input_shape=input_shape))
-    model.add(keras.layers.Dense(256, activation='relu', input_shape=input_shape, name='Dense_relu_1'))
-    model.add(keras.layers.Dense(128, activation='relu', activity_regularizer=l2(0.005), name='Dense_relu_2'))
-    model.add(keras.layers.Dense(64, activation='relu', activity_regularizer=l2(0.005), name='Dense_relu_3'))
-    model.add(keras.layers.Dropout(0.3, name='Dropout'))
-    # Output layer:
-    model.add(keras.layers.Dense(nr_classes, activation='softmax', name='Softmax'))
-
-    return model
-
-# Creates a keras.model with focus on Convulotion layers
-# Input: input shape, classes of classification
-# Ouput: model:Keras.model
-def CNN_1D(input_shape, nr_classes=5):
-
-    model = keras.Sequential(name='CNN_model')
-    model.add(keras.layers.Conv1D(32, kernel_size=5, activation='relu', input_shape=input_shape))
-    model.add(keras.layers.MaxPooling1D(pool_size=5))
-    model.add(keras.layers.Flatten())
-    model.add(keras.layers.Dense(128, activation='relu'))
-    model.add(keras.layers.Dropout(0.3))
-    # Ouput layer
-    model.add(keras.layers.Dense(nr_classes, activation='softmax', name='Softmax'))
 
     return model
 
@@ -377,12 +295,10 @@ if __name__ == "__main__":
     NR_SUBJECTS = 5
     NR_SESSIONS = 4
     BATCH_SIZE = 64
-    EPOCHS = 5
+    EPOCHS = 30
 
     TEST_SESSION_NR = 4
-    VERBOSE = 1
+    VERBOSE = 0 
-    MODEL_NAME = 'CNN_1D'
-    LOG = True
     
     # ----- Get prepared data: train, validation, and test ------
         # X_train.shape = (2806-X_test, 1, 208)
@@ -393,67 +309,29 @@ if __name__ == "__main__":
     X_train, X_test, y_train, y_test = prepare_datasets_sessions(X, y, session_lengths, TEST_SESSION_NR)
     
 
-    '''
+    #'''
     # ----- Make model ------
-    #model_GRU = GRU(input_shape=(1, 208)) # (timestep, 13*16 MFCC coefficients)
+    model_GRU = GRU(input_shape=(1, 208)) # (timestep, 13*16 MFCC coefficients)
-    #model_LSTM = LSTM(input_shape=(1, 208)) # (timestep, 13*16 MFCC coefficients)
+    model_LSTM = LSTM(input_shape=(1, 208)) # (timestep, 13*16 MFCC coefficients)
-    model_CNN_1D = CNN(input_shape=(208, 1)) # (timestep, 13*16 MFCC coefficients)
     
-    model_CNN_1D.summary()
+    model_GRU.summary()
-    #model_GRU.summary()
+    model_LSTM.summary()
-    #model_LSTM.summary()
     
-
     # ----- Train network ------
-    #history_GRU = train(model_GRU, X_train, y_train, verbose=VERBOSE, batch_size=BATCH_SIZE, epochs=EPOCHS)
+    history_GRU = train(model_GRU, X_train, y_train, verbose=VERBOSE, batch_size=BATCH_SIZE, epochs=EPOCHS)
-    #history_LSTM = train(model_LSTM, X_train, y_train, verbose=VERBOSE, batch_size=BATCH_SIZE, epochs=EPOCHS)
+    history_LSTM = train(model_LSTM, X_train, y_train, verbose=VERBOSE, batch_size=BATCH_SIZE, epochs=EPOCHS)
-    history_CNN_1D = train( model_CNN_1D, np.reshape(X_train, (X_train.shape[0], 208, 1)), 
+    #average = session_cross_validation_LSTM(X, y, session_lengths, NR_SESSIONS, BATCH_SIZE, EPOCHS)
-                            y_train, verbose=VERBOSE, batch_size=BATCH_SIZE, epochs=EPOCHS)
+    #print('\nCrossvalidated:', average)
     
-
     # ----- Plot train accuracy/error -----
     #plot_train_history(history)
 
-
     # ----- Evaluate model on test set ------
-
+    test_loss, test_acc = model_GRU.evaluate(X_test, y_test, verbose=VERBOSE)
-    #test_loss, test_acc = model_GRU.evaluate(X_test, y_test, verbose=VERBOSE)
+    print('\nTest accuracy GRU:', test_acc, '\n')
-    #print('\nTest accuracy GRU:', test_acc, '\n')
+    test_loss, test_acc = model_LSTM.evaluate(X_test, y_test, verbose=VERBOSE)
-    #test_loss, test_acc = model_LSTM.evaluate(X_test, y_test, verbose=VERBOSE)
+    print('\nTest accuracy LSTM:', test_acc, '\n')
-    #print('\nTest accuracy LSTM:', test_acc, '\n')
+    #'''
-    test_loss, test_acc = model_CNN_1D.evaluate(np.reshape(X_test, (X_test.shape[0], 208, 1)), y_test, verbose=0)
-    print('\nTest accuracy CNN_1D:', test_acc, '\n')
     
 
-    # ----- Store test predictions in CSV ------
-    prediction_csv_logger(np.reshape(X_test, (X_test.shape[0], 208, 1)), y_test, MODEL_NAME, model_CNN_1D, TEST_SESSION_NR)
-    '''
-
-
-    #'''
-    # ----- Cross validation ------
-    average_GRU = session_cross_validation('GRU', X, y, session_lengths, nr_sessions=NR_SESSIONS, 
-                                                                        log_to_csv=LOG,
-                                                                        batch_size=BATCH_SIZE, 
-                                                                        epochs=EPOCHS)
-    average_LSTM = session_cross_validation('LSTM', X, y, session_lengths, nr_sessions=NR_SESSIONS,
-                                                                        log_to_csv=LOG,
-                                                                        batch_size=BATCH_SIZE, 
-                                                                        epochs=EPOCHS)
-    average_FFN = session_cross_validation('FFN', X, y, session_lengths, nr_sessions=NR_SESSIONS,
-                                                                        log_to_csv=LOG,
-                                                                        batch_size=BATCH_SIZE, 
-                                                                        epochs=EPOCHS)
-    average_CNN = session_cross_validation('CNN_1D', X, y, session_lengths, nr_sessions=NR_SESSIONS,
-                                                                        log_to_csv=LOG,
-                                                                        batch_size=BATCH_SIZE, 
-                                                                        epochs=EPOCHS)
-
-    print('\n')
-    print('Crossvalidated GRU:', average_GRU)
-    print('Crossvalidated LSTM:', average_LSTM)
-    print('Crossvalidated FFN:', average_FFN)
-    print('Cross-validated CNN_1D:', average_CNN)
-    print('\n')
-    #'''
 

python_speech_features.egg-info/PKG-INFO

@@ -0,0 +1,10 @@
+Metadata-Version: 1.0
+Name: python-speech-features
+Version: 0.6.1
+Summary: Python Speech Feature extraction
+Home-page: https://github.com/jameslyons/python_speech_features
+Author: James Lyons
+Author-email: james.lyons0@gmail.com
+License: MIT
+Description: UNKNOWN
+Platform: UNKNOWN

python_speech_features.egg-info/SOURCES.txt

@@ -0,0 +1,7 @@
+python_speech_features/example.py
+python_speech_features/setup.py
+python_speech_features.egg-info/PKG-INFO
+python_speech_features.egg-info/SOURCES.txt
+python_speech_features.egg-info/dependency_links.txt
+python_speech_features.egg-info/requires.txt
+python_speech_features.egg-info/top_level.txt

python_speech_features.egg-info/dependency_links.txt

@@ -0,0 +1 @@
+

python_speech_features.egg-info/requires.txt

@@ -0,0 +1,2 @@
+numpy
+scipy

python_speech_features.egg-info/top_level.txt

@@ -0,0 +1 @@
+python_speech_features