def dtw_classify_character(test_char, train_char_list):
infinity = 2.0**32.0
score_list = []
count_list = []
for i in range(0, 10):
score_list.append(0)
count_list.append(0)
test_data_x = test_char._xseries
test_data_x = data_transform.xstretch_data(test_data_x)
test_data_x = data_transform.normalize_data(test_data_x)
test_data_y = test_char._yseries
test_data_y = data_transform.xstretch_data(test_data_y)
test_data_y = data_transform.normalize_data(test_data_y)
for i in range(train_char_list.shape[0]):
train_char = train_char_list[i]
train_data_x = train_char._xseries
train_data_x = data_transform.xstretch_data(train_data_x)
train_data_x = data_transform.normalize_data(train_data_x)
train_data_y = train_char._yseries
train_data_y = data_transform.xstretch_data(train_data_y)
train_data_y = data_transform.normalize_data(train_data_y)
print(train_data_y)
distance_matrix_x = pDTW.get_distance_matrix_DTW(test_data_x, train_data_x)
distance_matrix_y = pDTW.get_distance_matrix_DTW(test_data_y, train_data_y)
cost_matrix_x = pDTW.get_cost_matrix(distance_matrix_x)
cost_matrix_y = pDTW.get_cost_matrix(distance_matrix_y)
#path_x = pDTW.get_warp_path(cost_matrix_x)
#path_y = pDTW.get_warp_path(cost_matrix_y)
cost_matrix_x_area = 1.0*cost_matrix_x.shape[0]*cost_matrix_x.shape[1]
cost_matrix_y_area = 1.0*cost_matrix_y.shape[0]*cost_matrix_y.shape[1]
score = cost_matrix_x[-1,-1]/cost_matrix_x_area + cost_matrix_y[-1,-1]/cost_matrix_y_area
#print('identification = ' + str(train_char._classification))
#print('score = ' + str(score))
#if i < 3:
#p_plot.save_plot_matrix(test_char._data_bw)
#p_plot.save_plot_matrix(train_char._data_bw)
#
#number_of_lines = path_x.shape[0]
#nn = 1
#offset = 2
#plt.plot(train_data_x[:,0], train_data_x[:,1])
#plt.plot(test_data_x[:,0], test_data_x[:,1] +offset)
#for j in range(int(number_of_lines/nn) - 1):
#plt.plot((test_data_x[path_x[j*nn,0],0],\
#train_data_x[path_x[j*nn,1],0]),\
#(test_data_x[path_x[j*nn,0],1] + offset,\
#train_data_x[path_x[j*nn,1],1]))
#plt.show()
#
#p_plot.save_plot_matrix_line(distance_matrix_x, path_x)
#p_plot.save_plot_matrix_line(cost_matrix_x, path_x)
#
#number_of_lines = path_y.shape[0]
#nn = 1
#offset = 2
#plt.plot(train_data_y[:,0], train_data_y[:,1])
#plt.plot(test_data_y[:,0], test_data_y[:,1] +offset)
#for j in range(int(number_of_lines/nn) - 1):
#plt.plot((test_data_y[path_y[j*nn,0],0],\
#train_data_y[path_y[j*nn,1],0]),\
#(test_data_y[path_y[j*nn,0],1] + offset,\
#train_data_y[path_y[j*nn,1],1]))
#plt.show()
#
#p_plot.save_plot_matrix_line(distance_matrix_y, path_y)
#p_plot.save_plot_matrix_line(cost_matrix_y, path_y)
if train_char_list[i]._classification == 0:
score_list[0] = score_list[0] + score
count_list[0] = count_list[0] + 1
elif train_char_list[i]._classification == 1:
score_list[1] = score_list[1] + score
count_list[1] = count_list[1] + 1
elif train_char_list[i]._classification == 2:
score_list[2] = score_list[2] + score
count_list[2] = count_list[2] + 1
elif train_char_list[i]._classification == 3:
score_list[3] = score_list[3] + score
count_list[3] = count_list[3] + 1
elif train_char_list[i]._classification == 4:
score_list[4] = score_list[4] + score
count_list[4] = count_list[4] + 1
elif train_char_list[i]._classification == 5:
score_list[5] = score_list[5] + score
count_list[5] = count_list[5] + 1
elif train_char_list[i]._classification == 6:
score_list[6] = score_list[6] + score
count_list[6] = count_list[6] + 1
elif train_char_list[i]._classification == 7:
score_list[7] = score_list[7] + score
count_list[7] = count_list[7] + 1
elif train_char_list[i]._classification == 8:
score_list[8] = score_list[8] + score
count_list[8] = count_list[8] + 1
elif train_char_list[i]._classification == 9:
score_list[9] = score_list[9] + score
count_list[9] = count_list[9] + 1
for i in range(0, 10):
if count_list[i] != 0:
score_list[i] = (1.0*score_list[i])/count_list[i]
else:
score_list[i] = infinity
minim = infinity
min_i = 11
for i in range(0, 10):
if score_list[i] < minim:
minim = score_list[i]
min_i = i
test_char._predicted_classification = min_i
for i in range(len(score_list)):
print('i = ' + str(i) + '\t\t\t' + str(score_list[i]))
print("test_char._predicted_classification = " + str(test_char._predicted_classification))
print("test_char._classification = " + str(test_char._classification))
return
def dtw_classify_character(test_char, train_char_list):
infinity = 2.0**32.0
score_list = []
count_list = []
for i in range(0, 10):
score_list.append(0)
count_list.append(0)
test_data_x = test_char._xseries
test_data_x = data_transform.xstretch_data(test_data_x)
test_data_x = data_transform.normalize_data(test_data_x)
test_data_y = test_char._yseries
test_data_y = data_transform.xstretch_data(test_data_y)
test_data_y = data_transform.normalize_data(test_data_y)
for i in range(train_char_list.shape[0]):
train_char = train_char_list[i]
train_data_x = train_char._xseries
train_data_x = data_transform.xstretch_data(train_data_x)
train_data_x = data_transform.normalize_data(train_data_x)
train_data_y = train_char._yseries
train_data_y = data_transform.xstretch_data(train_data_y)
train_data_y = data_transform.normalize_data(train_data_y)
print(train_data_y)
distance_matrix_x = pDTW.get_distance_matrix_DTW(
test_data_x, train_data_x)
distance_matrix_y = pDTW.get_distance_matrix_DTW(
test_data_y, train_data_y)
cost_matrix_x = pDTW.get_cost_matrix(distance_matrix_x)
cost_matrix_y = pDTW.get_cost_matrix(distance_matrix_y)
#path_x = pDTW.get_warp_path(cost_matrix_x)
#path_y = pDTW.get_warp_path(cost_matrix_y)
cost_matrix_x_area = 1.0 * cost_matrix_x.shape[
0] * cost_matrix_x.shape[1]
cost_matrix_y_area = 1.0 * cost_matrix_y.shape[
0] * cost_matrix_y.shape[1]
score = cost_matrix_x[-1, -1] / cost_matrix_x_area + cost_matrix_y[
-1, -1] / cost_matrix_y_area
#print('identification = ' + str(train_char._classification))
#print('score = ' + str(score))
#if i < 3:
#p_plot.save_plot_matrix(test_char._data_bw)
#p_plot.save_plot_matrix(train_char._data_bw)
#
#number_of_lines = path_x.shape[0]
#nn = 1
#offset = 2
#plt.plot(train_data_x[:,0], train_data_x[:,1])
#plt.plot(test_data_x[:,0], test_data_x[:,1] +offset)
#for j in range(int(number_of_lines/nn) - 1):
#plt.plot((test_data_x[path_x[j*nn,0],0],\
#train_data_x[path_x[j*nn,1],0]),\
#(test_data_x[path_x[j*nn,0],1] + offset,\
#train_data_x[path_x[j*nn,1],1]))
#plt.show()
#
#p_plot.save_plot_matrix_line(distance_matrix_x, path_x)
#p_plot.save_plot_matrix_line(cost_matrix_x, path_x)
#
#number_of_lines = path_y.shape[0]
#nn = 1
#offset = 2
#plt.plot(train_data_y[:,0], train_data_y[:,1])
#plt.plot(test_data_y[:,0], test_data_y[:,1] +offset)
#for j in range(int(number_of_lines/nn) - 1):
#plt.plot((test_data_y[path_y[j*nn,0],0],\
#train_data_y[path_y[j*nn,1],0]),\
#(test_data_y[path_y[j*nn,0],1] + offset,\
#train_data_y[path_y[j*nn,1],1]))
#plt.show()
#
#p_plot.save_plot_matrix_line(distance_matrix_y, path_y)
#p_plot.save_plot_matrix_line(cost_matrix_y, path_y)
if train_char_list[i]._classification == 0:
score_list[0] = score_list[0] + score
count_list[0] = count_list[0] + 1
elif train_char_list[i]._classification == 1:
score_list[1] = score_list[1] + score
count_list[1] = count_list[1] + 1
elif train_char_list[i]._classification == 2:
score_list[2] = score_list[2] + score
count_list[2] = count_list[2] + 1
elif train_char_list[i]._classification == 3:
score_list[3] = score_list[3] + score
count_list[3] = count_list[3] + 1
elif train_char_list[i]._classification == 4:
score_list[4] = score_list[4] + score
count_list[4] = count_list[4] + 1
elif train_char_list[i]._classification == 5:
score_list[5] = score_list[5] + score
count_list[5] = count_list[5] + 1
elif train_char_list[i]._classification == 6:
score_list[6] = score_list[6] + score
count_list[6] = count_list[6] + 1
elif train_char_list[i]._classification == 7:
score_list[7] = score_list[7] + score
count_list[7] = count_list[7] + 1
elif train_char_list[i]._classification == 8:
score_list[8] = score_list[8] + score
count_list[8] = count_list[8] + 1
elif train_char_list[i]._classification == 9:
score_list[9] = score_list[9] + score
count_list[9] = count_list[9] + 1
for i in range(0, 10):
if count_list[i] != 0:
score_list[i] = (1.0 * score_list[i]) / count_list[i]
else:
score_list[i] = infinity
minim = infinity
min_i = 11
for i in range(0, 10):
if score_list[i] < minim:
minim = score_list[i]
min_i = i
test_char._predicted_classification = min_i
for i in range(len(score_list)):
print('i = ' + str(i) + '\t\t\t' + str(score_list[i]))
print("test_char._predicted_classification = " +
str(test_char._predicted_classification))
print("test_char._classification = " + str(test_char._classification))
return
five_0_x = five_0._xseries five_0_x = data_transform.stretch_data_x(five_0_x) five_0_x = data_transform.normalize_data(five_0_x) five_1_x = five_1._xseries five_1_x = data_transform.stretch_data_x(five_1_x) five_1_x = data_transform.normalize_data(five_1_x) distance_matrix_x = pDTW.get_distance_matrix_DTW(five_0_x, five_1_x) cost_matrix_x = pDTW.get_cost_matrix(distance_matrix_x) path_x = pDTW.get_warp_path(cost_matrix_x) p_plot.save_plot_matrix_line(distance_matrix_x, path_x) p_plot.save_plot_matrix_line(cost_matrix_x, path_x) offset_1 = 1 for i in range(five_1_x.shape[0]): five_1_x[i][1] = five_1_x[i][1] + offset_1 fig = plt.figure() for i in range(path_x.shape[0]): x0 = five_0_x[path_x[i][0]][0] x1 = five_1_x[path_x[i][1]][0] y0 = five_0_x[path_x[i][0]][1]