def __modify_acc_test(self,
simulated_marker,
x_train,
x_test,
R_cylinder=[]):
segment_data = SegmentData(self.__subject_data, self.__moved_segment)
walking_data_1_df = segment_data.get_segment_walking_1_data(
self.__speed)
train_index = x_train.index
test_index = x_test.index
R_standing_to_ground = segment_data.get_segment_R()
marker_cali_matrix = segment_data.get_marker_cali_matrix(self.__speed)
virtual_marker, R_IMU_transform = Processor.get_virtual_marker(
simulated_marker, walking_data_1_df, marker_cali_matrix,
R_standing_to_ground)
# Processor.check_virtual_marker(virtual_marker, walking_data_1_df, self.__moved_segment) # for check
acc_IMU = Processor.get_acc(virtual_marker, R_IMU_transform)
# if it was simulated on cylinder, a rotation around the cylinder surface is necessary
if len(R_cylinder) != 0:
acc_IMU = np.matmul(R_cylinder, acc_IMU.T).T
changed_columns = []
for acc_name in ['_acc_x', '_acc_y', '_acc_z']:
column = self.__moved_segment + acc_name
changed_columns.append(column)
acc_IMU_df = pd.DataFrame(acc_IMU)
# x_test is just a quote of the original x_test, deep copy so that no warning will show up
x_train_changed = x_train.copy()
x_test_changed = x_test.copy()
x_train_changed[changed_columns] = acc_IMU_df.loc[train_index]
x_test_changed[changed_columns] = acc_IMU_df.loc[test_index]
return x_train_changed, x_test_changed
def check_acc_difference(self, x, R_cylinder=[]):
point_list, x_range, z_range = self.__get_plane_position()
for point in point_list:
simulated_marker = point[0]
segment_data = SegmentData(self.__subject_data,
self.__moved_segment)
walking_data_1_df = segment_data.get_segment_walking_1_data(
self.__speed)
test_index = x.index
R_standing_to_ground = segment_data.get_segment_R()
marker_cali_matrix = segment_data.get_marker_cali_matrix(
self.__speed)
virtual_marker, R_IMU_transform = Processor.get_virtual_marker(
simulated_marker, walking_data_1_df, marker_cali_matrix,
R_standing_to_ground)
# Processor.check_virtual_marker(virtual_marker, walking_data_1_df, self.__moved_segment) # for check
acc_IMU = Processor.get_acc(virtual_marker, R_IMU_transform)
# if it was simulated on cylinder, a rotation around the cylinder surface is necessary
if len(R_cylinder) != 0:
acc_IMU = np.matmul(R_cylinder, acc_IMU.T).T
changed_columns = []
for acc_name in ['_acc_x', '_acc_y', '_acc_z']:
column = self.__moved_segment + acc_name
changed_columns.append(column)
plt.plot(acc_IMU[:, 0])
plt.plot(x[changed_columns].as_matrix()[:, 0])
score = r2_score(acc_IMU[:, 0], x[changed_columns].as_matrix()[:,
0])
plt.title('pelvis, x += ' + str(point[1]) + ', z += ' +
str(point[2]) + ', score = ' + str(score))
plt.show()
def get_xy(self):
speed = self.__speed
walking_data_1_df = self.__subject_data.get_walking_1_data(speed)
# for now we only use walking data 1
data_len = walking_data_1_df.shape[0]
# get x
if self.__gyr_data:
x = np.zeros([data_len, 48])
else:
x = np.zeros([data_len, 24])
i_segment = 0
for segment_name in SEGMENT_NAMES:
segment_data = SegmentData(self.__subject_data, segment_name)
segment_data_walking_1_df = segment_data.get_segment_walking_1_data(
speed)
center_marker = segment_data.get_center_point_mean(speed)
R_standing_to_ground = segment_data.get_segment_R()
marker_cali_matrix = segment_data.get_marker_cali_matrix(speed)
virtual_marker, R_IMU_transform = Processor.get_virtual_marker(
center_marker, segment_data_walking_1_df, marker_cali_matrix,
R_standing_to_ground)
# Processor.check_virtual_marker(virtual_marker, walking_data_1_df, segment_name) # for check
acc_IMU = Processor.get_acc(virtual_marker, R_IMU_transform)
x[:, 3 * i_segment:3 * (i_segment + 1)] = acc_IMU
if self.__gyr_data:
gyr_IMU = Processor.get_gyr(segment_data_walking_1_df,
R_IMU_transform)
x[:, 24 + 3 * i_segment:24 + 3 * (i_segment + 1)] = gyr_IMU
i_segment += 1
x = pd.DataFrame(x)
if self.__gyr_data:
x.columns = ALL_ACC_GYR_NAMES
else:
x.columns = ALL_ACC_NAMES
# get y
y = walking_data_1_df[self.__output_names]
return x, y