# Segmentation
for n in range(nSignals):
segmented_emg.append(fex.segmentation(emg[n]))
# Feature calculation
feature_list = [fex.mav, fex.rms, fex.var, fex.ssi, fex.zc, fex.wl, fex.ssc, fex.wamp]
nSegments = len(segmented_emg[0][0])
nFeatures = len(feature_list)
feature_matrix = np.zeros((nGestures*nIterations*nSegments,nFeatures*nChannels))
n = 0
for i in range(0,nSignals,nChannels):
for j in range(nSegments):
feature_matrix[n] = fex.features((segmented_emg[i][:,j],segmented_emg[i+1][:,j]),feature_list)
n = n + 1
# Target matrix generation
y = fex.gestures(nIterations*nSegments,nGestures)
# Dimensionality reduction and feature scaling
[X,reductor,scaler] = fex.feature_scaling(feature_matrix, y)
# Split dataset into training and testing datasets
X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.4, random_state=42)
# Classifier training
classifier = SVC(kernel='rbf',C=10,gamma=10)
classifier.fit(X_train,y_train)
emg.append(emg_data['motion'+str(m)+'_ch'+str(c)][:,i]) #motion1_ch1_i1, motion1_ch2_i1, motion1_ch1_i2, motion1_ch2_i2
for n in range(len(emg)):
emg_seg.append(fex.segmentation(emg[n]))
feature_list = [fex.mav, fex.rms, fex.var, fex.ssi, fex.zc, fex.wl, fex.ssc, fex.wamp]
nSegments = len(emg_seg[0][0])
nFeatures = len(feature_list)
feature_matrix = np.zeros((nSegments*nIterations*nMotions,nFeatures*nChannels))
n = 0
for i in range(0,len(emg_seg),nChannels):
for j in range(nSegments):
feature_matrix[n] = fex.features((emg_seg[i][:,j],emg_seg[i+1][:,j]),feature_list)
n += 1
from sklearn.svm import SVC
import timeit
target = fex.gestures(nIterations*nSegments,nMotions)
#target = np.concatenate((0*np.ones((128,1)),1*np.ones((128,1)),2*np.ones((128,1)),3*np.ones((128,1)),4*np.ones((128,1))),axis=0).ravel()
tic = timeit.default_timer()
[feat_scaled,reductor,scaler] = fex.feature_scaling(feature_matrix, target)
segmented_emg.append(fex.segmentation(emg[n]))
# Feature calculation
feature_list = [
fex.mav, fex.rms, fex.var, fex.ssi, fex.zc, fex.wl, fex.ssc, fex.wamp
]
nSegments = len(segmented_emg[0][0])
nFeatures = len(feature_list)
feature_matrix = np.zeros(
(nGestures * nIterations * nSegments, nFeatures * nChannels))
n = 0
for i in range(0, nSignals, nChannels):
for j in range(nSegments):
feature_matrix[n] = fex.features(
(segmented_emg[i][:, j], segmented_emg[i + 1][:, j]), feature_list)
n = n + 1
# Target matrix generation
y = fex.gestures(nIterations * nSegments, nGestures)
# Dimensionality reduction and feature scaling
[X, reductor, scaler] = fex.feature_scaling(feature_matrix, y)
# Split dataset into training and testing datasets
X_train, X_test, y_train, y_test = train_test_split(X,
y,
test_size=0.4,
random_state=42)
# Classifier training
feature_list = [fex.mav, fex.rms, fex.var, fex.ssi, fex.zc, fex.wl, fex.ssc, fex.wamp]
n_segments = len(segmented_emg[0][0])
for i in range(0,n_signals,n_channels):
if len(segmented_emg[i][0]) < n_segments :
n_segments = len(segmented_emg[i][0])
n_features = len(feature_list)
feature_matrix = np.zeros((n_classes*n_iterations*n_segments,n_features*n_channels))
n = 0
for i in range(0,n_signals,n_channels):
for j in range(n_segments):
feature_matrix[n] = fex.features((segmented_emg[i][:,j],
segmented_emg[i+1][:,j],
segmented_emg[i+2][:,j],
segmented_emg[i+3][:,j],
segmented_emg[i+4][:,j],
segmented_emg[i+5][:,j],
segmented_emg[i+6][:,j],
segmented_emg[i+7][:,j]),feature_list)
n = n + 1
# Target matrix generation
y = fex.generate_target(n_iterations*n_segments,class_labels)
# Dimensionality reduction and feature scaling
[X,reductor,scaler] = fex.feature_scaling(feature_matrix, y)
# Split dataset into training and testing datasets
X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.4, random_state=42)
# Classifier training
def process_emg(data):
global count, emg, classifier, class_labels
if count < 4:
count += 1
emg.append(np.array(data[0]))
emg.append(np.array(data[1]))
else:
count = 0
n_classes = 1
n_iterations = 1
n_channels = 8
#n_signals = 8
n_signals = n_classes * n_iterations * n_channels
segmented_emg = list()
print("EMG: ")
print(emg)
# Segmentation
for n in range(n_signals):
segmented_emg.append(fex.segmentation(emg[n], n_samples=1))
# Feature calculation
feature_list = [
fex.mav, fex.rms, fex.var, fex.ssi, fex.zc, fex.wl, fex.ssc,
fex.wamp
]
n_segments = len(segmented_emg[0][0])
print("\nN_Segments: " + str(n_segments) + "\n")
n_features = len(feature_list)
feature_matrix = np.zeros(
(n_classes * n_iterations * n_segments, n_features * n_channels))
n = 0
print("Feature_Matrix_SHAPE: ")
print(feature_matrix.shape)
print("")
print("Segmented_EMG: ")
print(segmented_emg)
print("")
for i in range(0, n_classes):
for j in range(n_segments):
#print("i: " + str(i))
#print("j: " + str(j))
feature_matrix[n] = fex.features(
(segmented_emg[i][:, j], segmented_emg[i + 1][:, j],
segmented_emg[i + 2][:, j], segmented_emg[i + 3][:, j],
segmented_emg[i + 4][:, j], segmented_emg[i + 5][:, j],
segmented_emg[i + 6][:, j], segmented_emg[i + 7][:, j]),
feature_list)
n = n + 1
# Target matrix generation
print(feature_matrix)
print(feature_matrix.shape)
y = fex.generate_target(n_iterations * n_segments, class_labels)
print(y)
print(y.shape)
# Dimensionality reduction and feature scaling
[X, reductor, scaler] = fex.feature_scaling(feature_matrix, y)
# Classification
print(X)
print(X.shape)
predict = classifier.predict(feature_matrix)
print(predict)
emg = list()