def analyze_signal(data_path):
# calibrate zcr treshold
Frame.zcr_treshold = calibrate_param(CALIBRATE_PATH)
frames_processed = []
y_pred = []
# load data
X, y, mapping = load_data(data_path)
for(samples, c) in zip(X, y):
print("processing")
f = Frame(samples, c)
f.classify_zcr()
frames_processed.append(f)
y_pred.append(f.prediction)
cm = confusion_matrix(y, y_pred)
print(accuracy_score(y, y_pred))
print(cm)
plot_res(y, y_pred)
signal, noise, fec, msc, over, nds = analyze_objective_params(y, y_pred)
print("signal {}, noise {}, fec {}, msc {}, over {}, nds {}".format(
signal, noise, fec, msc, over, nds))
return frames_processed
def classify_by_feature(X_train, X_test, y_train, y_test):
# scale the features
sc = StandardScaler()
X_train = sc.fit_transform(X_train)
X_test = sc.transform(X_test)
# classifier and fitting
classifier = LogisticRegression(random_state=0)
classifier.fit(X_train, y_train)
# predictions
y_pred = classifier.predict(X_test)
# confusion matrix
cm = confusion_matrix(y_test, y_pred)
print(cm)
accuracy = accuracy_score(y_test, y_pred)
print(accuracy)
plot_res(y_test, y_pred)
signal, noise, fec, msc, over, nds = analyze_objective_params(
y_test, y_pred)
print("signal {}, noise {}, fec {}, msc {}, over {}, nds {}".format(
signal, noise, fec, msc, over, nds))
return classifier, sc
X_train = X_train[..., np.newaxis]
X_validation = X_validation[..., np.newaxis]
X_test = X_test[..., np.newaxis]
return X_train, X_validation, X_test, y_train, y_validation, y_test
if __name__ == "__main__":
# load dataset
X_train, X_validation, X_test, y_train, y_validation, y_test = prepare_dataset(
DATA_TRAIN, DATA_EVAL, DATA_TEST)
model = keras.models.load_model(PATH_TO_MODEL)
y_pred_val = model.predict(X_test)
# treshold
y_pred_val = [0 if x < 0.5 else 1 for x in y_pred_val]
cm = confusion_matrix(y_test, y_pred_val)
print(cm)
signal, noise, fec, msc, over, nds = analyze_objective_params(
y_test, y_pred_val)
print("signal {}, noise {}, fec {}, msc {}, over {}, nds {}".format(
signal, noise, fec, msc, over, nds))
plot_res(y_test, y_pred_val)
fig.savefig('plots_final/svm.png')
plt.show()
if __name__ == "__main__":
# analyze validation data
y_val, y_pred_val = analyze_signal_samples(
DATA_TRAIN_SAMPLES, DATA_TEST_SAMPLES)
# analyze test data
y_test, y_pred = analyze_signal_samples(
DATA_TRAIN_SAMPLES, DATA_STREAM)
# objective params for test data
signal, noise, fec, msc, over, nds = analyze_objective_params(
y_test, y_pred)
# objective params for validation data
signal_val, noise_val, fec_val, msc_val, over_val, nds_val = analyze_objective_params(
y_val, y_pred_val)
print("signal {}, noise {}, fec {}, msc {}, over {}, nds {}".format(
signal, noise, fec, msc, over, nds))
with open('svm/objective_stream.txt', 'w') as file:
file.write("signal {}, noise {}, fec {}, msc {}, over {}, nds {}".format(
signal, noise, fec, msc, over, nds))
print("signal {}, noise {}, fec {}, msc {}, over {}, nds {}".format(
signal_val, noise_val, fec_val, msc_val, over_val, nds_val))
with open('svm/objective_val.txt', 'w') as file:
ax.set_xlabel('numer ramki')
ax.set_ylabel('mowa/brak mowy')
ax.plot(y_pred, color='b', label='stan faktyczny')
ax.plot(y_pred, color='r', linestyle='--', label='decyzja algorytmu')
ax.set_yticks([0, 1])
ax.set_yticklabels(['brak mowy', 'mowa'])
ax.legend(loc="lower right")
fig.savefig('plots_final/rfc.png')
plt.show()
if __name__ == "__main__":
# analyze signal
y_prediction, y_fact = analyze_signal_samples(DATA_TRAIN_SAMPLES,
DATA_TEST_SAMPLES)
# calculate objective params
signal, noise, fec, msc, over, nds = analyze_objective_params(
y_fact, y_prediction)
print("signal {}, noise {}, fec {}, msc {}, over {}, nds {}".format(
signal, noise, fec, msc, over, nds))
with open('random_forest/objective.txt', 'w') as file:
file.write(
"signal {}, noise {}, fec {}, msc {}, over {}, nds {}".format(
signal, noise, fec, msc, over, nds))
# plot vad's decisions
plot_res(y_fact, y_prediction)
def analyze_signal(data_path):
# ! new conf matrix from sklearn
y_pred = []
# Set thresholds
p = 0.1
e_silence = 0
e_r_old, unvoiced_buffer = calibrate_param(CALIBRATE_PATH)
var_old = np.var(unvoiced_buffer)
# set init treshold
Frame.full_band_treshold = e_r_old
X, y, mapping = load_data(data_path)
# array with classified frames
frames_processed = []
# ! Delete it, it's global
# unvoiced_buffer = []
vector_threshold = []
buffor_index = N - 1
#frame by frame processing
for (samples, c) in zip(X, y):
f = Frame(samples, c)
f.classify_aled_frame(f.samples)
vector_threshold.append([Frame.full_band_treshold] * FRAME_SIZE)
# save frame
frames_processed.append(f)
# save the prediction value
y_pred.append(f.prediction)
# check frame and update threshold
if f.prediction is 1:
continue
# set index at 0, when it is the end of the buffer
if buffor_index == N:
buffor_index = 0
unvoiced_buffer[buffor_index] = f.full_band_energy
var_new = np.var(unvoiced_buffer)
e_silence = f.full_band_energy
p = p_actualize(var_new, var_old)
Frame.full_band_treshold = calc_new_threshold(e_r_old, e_silence, p)
var_old = var_new
buffor_index += 1
cm = confusion_matrix(y, y_pred)
print(accuracy_score(y, y_pred))
plot_res(y, y_pred)
signal, noise, fec, msc, over, nds = analyze_objective_params(y, y_pred)
print("signal {}, noise {}, fec {}, msc {}, over {}, nds {}".format(
signal, noise, fec, msc, over, nds))
return frames_processed, vector_threshold, cm
print(f'Train Accuracy - : {rf_Grid.score(X_train,y_train):.3f}')
print(f'Test Accuracy - : {rf_Grid.score(X_test,y_test):.3f}')
print(f'Best params - :{rf_Grid.best_params_}')
best_grid = RandomForestClassifier(n_estimators= rf_Grid.best_params_['n_estimators'],
max_features= rf_Grid.best_params_['max_features'],
max_depth= rf_Grid.best_params_['max_depth'],
min_samples_split= rf_Grid.best_params_['min_samples_split'],
min_samples_leaf= rf_Grid.best_params_['min_samples_leaf'],
bootstrap= rf_Grid.best_params_['bootstrap'])
print("confussion matrix")
y_pred = best_grid.predict(X_test, y_test)
cm = confusion_matrix(y_test, y_pred)
print("calculate test stream")
y_pred_stream = best_grid.predict(X_stream, y_stream)
cm = confusion_matrix(y_stream, y_pred_stream)
with open('random_forest/conf_test.txt', 'w') as file:
file.write("True Neg {}, False Pos {}, False neg {}, True Pos {}".format(cm[0][0], cm[0][1], cm[1][0], cm[1][1]))
signal, noise, fec, msc, over, nds = analyze_objective_params(y_stream,y_pred_stream)
print("signal {}, noise {}, fec {}, msc {}, over {}, nds {}".format(signal, noise, fec, msc, over, nds))
with open('random_forest/objective.txt', 'w') as file:
file.write("signal {}, noise {}, fec {}, msc {}, over {}, nds {}".format(signal, noise, fec, msc, over, nds))