new gait event methods
This commit is contained in:
davidpagnon
parent
989967e2b2
commit
2160d47648
@ -7,43 +7,60 @@
|
|||||||
## GAIT EVENTS DETECTION ##
|
## GAIT EVENTS DETECTION ##
|
||||||
##################################################
|
##################################################
|
||||||
|
|
||||||
Determine gait events according to Zeni et al. (2008).
|
Determine gait on and off from a TRC file of point coordinates.
|
||||||
Write them in gaitevents.txt (append results if file already exists).
|
Three available methods, each of them with their own pros and cons:
|
||||||
|
|
||||||
t_HeelStrike = max(XHeel - Xsacrum)
|
- "forward_coordinates":
|
||||||
t_ToeOff = min(XToe - XSacrum)
|
on = max(XHeel - Xsacrum)
|
||||||
|
off = min(XToe - XSacrum)
|
||||||
|
++: Works well for walking (Zeni et al., 2008)
|
||||||
|
++: No argument nor tuning necessary
|
||||||
|
--: Not suitable for running
|
||||||
|
--: does not work if the person is not going on a straight line
|
||||||
|
|
||||||
Reference:
|
- "height_coordinates":
|
||||||
“Two simple methods for determining gait events during treadmill and
|
on = YToe < height_threshold
|
||||||
overground walking using kinematic data.”
|
off = YToe > height_threshold
|
||||||
Gait & posture vol. 27,4 (2008): 710-4. doi:10.1016/j.gaitpost.2007.07.007
|
++: Best results for running and walking
|
||||||
|
++: Works if the person is not going on a straight line
|
||||||
|
--: Does not work is the person is grazing the ground
|
||||||
|
--: Does not work if the field is not flat
|
||||||
|
--: height_threshold might need to be tuned
|
||||||
|
--: Heel point might be more accurate than toe point for walking
|
||||||
|
|
||||||
|
- "forward_velocity":
|
||||||
|
on = VToe < forward_velocity_threshold
|
||||||
|
off = VToe > forward_velocity_threshold
|
||||||
|
++: Works for running
|
||||||
|
--: More sensitive to noise
|
||||||
|
--: Tends to anticipate off if the marker is not at the tip of the toe
|
||||||
|
--: forward_velocity_threshold might need to be tuned
|
||||||
|
--: Does not work if the person is not going on a straight line
|
||||||
|
|
||||||
Usage:
|
Usage:
|
||||||
Replace constants with the appropriate marker names.
|
List of available arguments:
|
||||||
If direction is negative, you need to include an equal sign in the argument,
|
python -m trc_gaitevents -h
|
||||||
eg -d=-Z or --gait_direction=-Z
|
|
||||||
|
|
||||||
from Pose2Sim.Utilities import trc_gaitevents; trc_gaitevents.trc_gaitevents_func(r'<input_trc_file>', '<gait_direction>')
|
import trc_gaitevents; trc_gaitevents.trc_gaitevents_func(trc_path=r'<input_trc_file>')
|
||||||
OR python -m trc_gaitevents -i input_trc_file
|
OR import trc_gaitevents; trc_gaitevents.trc_gaitevents_func(trc_path=r'<input_trc_file>', method='forward_coordinates', gait_direction='-X', plot=True, save_output=True, output_file='gaitevents.txt')
|
||||||
OR python -m trc_gaitevents -i input_trc_file --gait_direction=-Z
|
OR import trc_gaitevents; trc_gaitevents.trc_gaitevents_func(trc_path=r'<input_trc_file>', method='height_coordinates', up_direction='Y', height_threshold=6, right_toe_marker='RBigToe', left_toe_marker='LBigToe')
|
||||||
|
OR import trc_gaitevents; trc_gaitevents.trc_gaitevents_func(trc_path=r'<input_trc_file>', method='forward_velocity', gait_direction='-Z', forward_velocity_threshold=1.5, right_toe_marker='RBigToe', left_toe_marker='LBigToe')
|
||||||
|
|
||||||
|
python -m trc_gaitevents -i input_trc_file
|
||||||
|
OR python -m trc_gaitevents -i input_trc_file --method forward_coordinates --gait_direction=-X --plot True --save_output True --output_file gaitevents.txt
|
||||||
|
OR python -m trc_gaitevents -i input_trc_file --method height_coordinates --up_direction=Y --height_threshold 6 --right_toe_marker RBigToe --left_toe_marker LBigToe
|
||||||
|
OR python -m trc_gaitevents -i input_trc_file --method forward_velocity --gait_direction=-Z --forward_velocity_threshold 1.5 --right_toe_marker RBigToe --left_toe_marker LBigToe
|
||||||
'''
|
'''
|
||||||
|
|
||||||
|
|
||||||
## CONSTANTS
|
|
||||||
R_SACRUM_MARKER = 'RHip'
|
|
||||||
R_HEEL_MARKER = 'RHeel'
|
|
||||||
R_TOE_MARKER = 'RBigToe'
|
|
||||||
L_SACRUM_MARKER = 'LHip'
|
|
||||||
L_HEEL_MARKER = 'LHeel'
|
|
||||||
L_TOE_MARKER = 'LBigToe'
|
|
||||||
|
|
||||||
|
|
||||||
## INIT
|
## INIT
|
||||||
import os
|
import os
|
||||||
import argparse
|
import argparse
|
||||||
import pandas as pd
|
import pandas as pd
|
||||||
import numpy as np
|
import numpy as np
|
||||||
from scipy import signal
|
from scipy import signal
|
||||||
|
from scipy.ndimage import gaussian_filter1d
|
||||||
|
import matplotlib.pyplot as plt
|
||||||
|
|
||||||
|
|
||||||
## AUTHORSHIP INFORMATION
|
## AUTHORSHIP INFORMATION
|
||||||
@ -58,124 +75,564 @@ __status__ = "Development"
|
|||||||
|
|
||||||
|
|
||||||
## FUNCTIONS
|
## FUNCTIONS
|
||||||
def df_from_trc(trc_path):
|
def start_end_true_seq(series):
|
||||||
'''
|
'''
|
||||||
Retrieve header and data from trc path.
|
Find starts and ends of sequences of True values in a pandas Series
|
||||||
|
|
||||||
|
INPUTS:
|
||||||
|
- series: pandas Series of boolean values
|
||||||
|
|
||||||
|
OUTPUTS:
|
||||||
|
- start_indices: list of start indices
|
||||||
|
- end_indices: list of end indices
|
||||||
'''
|
'''
|
||||||
|
|
||||||
# DataRate CameraRate NumFrames NumMarkers Units OrigDataRate OrigDataStartFrame OrigNumFrames
|
diff = series.ne(series.shift())
|
||||||
df_header = pd.read_csv(trc_path, sep="\t", skiprows=1, header=None, nrows=2, encoding="ISO-8859-1")
|
start_indices = series.index[diff & series].tolist()
|
||||||
header = dict(zip(df_header.iloc[0].tolist(), df_header.iloc[1].tolist()))
|
end_indices = (series.index[diff & ~series]-1).tolist()
|
||||||
|
if end_indices[0] == -1: end_indices.pop(0)
|
||||||
|
|
||||||
# Label1_X Label1_Y Label1_Z Label2_X Label2_Y
|
return start_indices, end_indices
|
||||||
df_lab = pd.read_csv(trc_path, sep="\t", skiprows=3, nrows=1)
|
|
||||||
labels = df_lab.columns.tolist()[2:-1:3]
|
|
||||||
labels_XYZ = np.array([[labels[i]+'_X', labels[i]+'_Y', labels[i]+'_Z'] for i in range(len(labels))], dtype='object').flatten()
|
|
||||||
labels_FTXYZ = np.concatenate((['Frame#','Time'], labels_XYZ))
|
|
||||||
|
|
||||||
data = pd.read_csv(trc_path, sep="\t", skiprows=5, index_col=False, header=None, names=labels_FTXYZ)
|
|
||||||
|
|
||||||
return header, data
|
|
||||||
|
|
||||||
|
|
||||||
def gait_events(trc_path, gait_direction):
|
def read_trc(trc_path):
|
||||||
'''
|
'''
|
||||||
Determine gait events according to Zeni et al. (2008).
|
Read trc file
|
||||||
t_HellStrike = max(XHeel - Xsacrum)
|
|
||||||
t_ToeOff = min(XToe - XSacrum)
|
INPUTS:
|
||||||
|
- trc_path: path to the trc file
|
||||||
|
|
||||||
|
OUTPUTS:
|
||||||
|
- Q_coords: dataframe of coordinates
|
||||||
|
- frames_col: series of frames
|
||||||
|
- time_col: series of time
|
||||||
|
- markers: list of marker names
|
||||||
|
- header: list of header lines
|
||||||
'''
|
'''
|
||||||
|
|
||||||
# Read trc
|
with open(trc_path, 'r') as trc_file:
|
||||||
header, data = df_from_trc(trc_path)
|
header = [next(trc_file) for line in range(5)]
|
||||||
|
markers = header[3].split('\t')[2::3]
|
||||||
|
|
||||||
# In case of a sign in direction (eg -Z)
|
trc_df = pd.read_csv(trc_path, sep="\t", skiprows=4)
|
||||||
sign = ''
|
frames_col, time_col = pd.Series(trc_df.iloc[:,0], name='frames'), pd.Series(trc_df.iloc[:,1], name='time')
|
||||||
if any(x in gait_direction for x in ['-', '+']):
|
Q_coords = trc_df.drop(trc_df.columns[[0, 1]], axis=1)
|
||||||
sign = gait_direction[0]
|
|
||||||
gait_direction = gait_direction[-1]
|
|
||||||
|
|
||||||
# Retrieve data of interest
|
return Q_coords, frames_col, time_col, markers, header
|
||||||
XRSacrum = data['_'.join((R_SACRUM_MARKER, gait_direction))]
|
|
||||||
XRHeel = data['_'.join((R_HEEL_MARKER, gait_direction))]
|
|
||||||
XRToe = data['_'.join((R_TOE_MARKER, gait_direction))]
|
|
||||||
XLSacrum = data['_'.join((L_SACRUM_MARKER, gait_direction))]
|
|
||||||
XLHeel = data['_'.join((L_HEEL_MARKER, gait_direction))]
|
|
||||||
XLToe = data['_'.join((L_TOE_MARKER, gait_direction))]
|
|
||||||
|
|
||||||
# Prominence of the peaks
|
|
||||||
unit = header['Units']
|
def clean_gait_events(gait_events):
|
||||||
|
'''
|
||||||
|
Clean gait events
|
||||||
|
Remove consecutive on-off pairs if they are not alternating
|
||||||
|
|
||||||
|
'''
|
||||||
|
|
||||||
|
Ron, Lon, Roff, Loff = gait_events
|
||||||
|
|
||||||
|
# Remove on-off pairs if they are not alternating
|
||||||
|
|
||||||
|
|
||||||
|
# If first event is off, remove it
|
||||||
|
if Ron[0]>Roff[0]: Roff.pop(0)
|
||||||
|
if Lon[0]>Loff[0]: Loff.pop(0)
|
||||||
|
|
||||||
|
# If last event is on, remove it
|
||||||
|
if Ron[-1]>Roff[-1]: Ron.pop(-1)
|
||||||
|
if Lon[-1]>Loff[-1]: Lon.pop(-1)
|
||||||
|
|
||||||
|
# If there are several left onsets in a row between right onsets, remove all but the first one. Idem for the other side
|
||||||
|
merged_on = sorted(Ron + Lon)
|
||||||
|
merged_clean_on = [merged_on[0]]
|
||||||
|
# Keep alternating elements
|
||||||
|
for i in range(1, len(merged_on)):
|
||||||
|
if (merged_on[i] in Ron and merged_clean_on[-1] not in Ron) or \
|
||||||
|
(merged_on[i] in Lon and merged_clean_on[-1] not in Lon):
|
||||||
|
merged_clean_on.append(merged_on[i])
|
||||||
|
# Remove consecutive elements at the start if they are from the same list
|
||||||
|
while len(merged_clean_on) > 1 \
|
||||||
|
and (merged_clean_on[0] in Ron \
|
||||||
|
and merged_clean_on[1] in Ron or merged_clean_on[0] in Lon \
|
||||||
|
and merged_clean_on[1] in Lon):
|
||||||
|
merged_clean_on.pop(0)
|
||||||
|
# Remove consecutive elements at the end if they are from the same list
|
||||||
|
while len(merged_clean_on) > 1 \
|
||||||
|
and (merged_clean_on[-1] in Ron \
|
||||||
|
and merged_clean_on[-2] in Ron or merged_clean_on[-1] in Lon \
|
||||||
|
and merged_clean_on[-2] in Lon):
|
||||||
|
merged_clean_on.pop(-1)
|
||||||
|
# Make sure the output lists are in the right order
|
||||||
|
if merged_clean_on[0] in Ron:
|
||||||
|
Ron = merged_clean_on[::2]
|
||||||
|
Lon = merged_clean_on[1::2]
|
||||||
|
else:
|
||||||
|
Ron = merged_clean_on[1::2]
|
||||||
|
Lon = merged_clean_on[::2]
|
||||||
|
|
||||||
|
# If there are several left off in a row between right off, remove all but the last one. Idem for the other side
|
||||||
|
merged_off = sorted(Roff + Loff, reverse=True)
|
||||||
|
merged_clean_off = [merged_off[0]]
|
||||||
|
# Keep alternating elements
|
||||||
|
for i in range(1, len(merged_off)):
|
||||||
|
if (merged_off[i] in Roff and merged_clean_off[-1] not in Roff) or \
|
||||||
|
(merged_off[i] in Loff and merged_clean_off[-1] not in Loff):
|
||||||
|
merged_clean_off.append(merged_off[i])
|
||||||
|
# Remove consecutive elements at the start if they are from the same list
|
||||||
|
while len(merged_clean_off) > 1 \
|
||||||
|
and (merged_clean_off[0] in Roff \
|
||||||
|
and merged_clean_off[1] in Roff or merged_clean_off[0] in Loff \
|
||||||
|
and merged_clean_off[1] in Loff):
|
||||||
|
merged_clean_off.pop(0)
|
||||||
|
# Remove consecutive elements at the end if they are from the same list
|
||||||
|
while len(merged_clean_off) > 1 \
|
||||||
|
and (merged_clean_off[-1] in Roff \
|
||||||
|
and merged_clean_off[-2] in Roff or merged_clean_off[-1] in Loff \
|
||||||
|
and merged_clean_off[-2] in Loff):
|
||||||
|
merged_clean_off.pop(-1)
|
||||||
|
# Make sure the output lists are in the right order
|
||||||
|
if merged_clean_off[0] in Roff:
|
||||||
|
Roff = merged_clean_off[::2][::-1]
|
||||||
|
Loff = merged_clean_off[1::2][::-1]
|
||||||
|
else:
|
||||||
|
Roff = merged_clean_off[1::2][::-1]
|
||||||
|
Loff = merged_clean_off[::2][::-1]
|
||||||
|
|
||||||
|
|
||||||
|
return Ron, Lon, Roff, Loff
|
||||||
|
|
||||||
|
|
||||||
|
def gait_events_fwd_coords(trc_path, gait_direction, markers=['RHeel', 'RBigToe', 'LHeel', 'LBigToe', 'Hip'], plot=True):
|
||||||
|
'''
|
||||||
|
Determine gait on and off with "forward_coordinates" method
|
||||||
|
|
||||||
|
on = max(XHeel - Xsacrum)
|
||||||
|
off = min(XToe - XSacrum)
|
||||||
|
++: Works well for walking (Zeni et al., 2008)
|
||||||
|
++: No argument nor tuning necessary
|
||||||
|
--: Not suitable for running
|
||||||
|
--: does not work if the person is not going on a straight line
|
||||||
|
|
||||||
|
INPUTS:
|
||||||
|
- trc_path: path to the trc file
|
||||||
|
- gait_direction: tuple (sign, direction) with sign in {-1, 1} and direction in {'X', 'Y', 'Z'}
|
||||||
|
- markers: list of marker names in the following order: [right_heel_marker, right_toe_marker, left_heel_marker, left_toe_marker, sacrum_marker]
|
||||||
|
- plot: plot results or not (boolean)
|
||||||
|
|
||||||
|
OUTPUTS:
|
||||||
|
- t_Ron: list of right on times
|
||||||
|
- t_Lon: list of left on times
|
||||||
|
- t_Roff: list of right off times
|
||||||
|
- t_Loff: list of left off times
|
||||||
|
'''
|
||||||
|
|
||||||
|
# Retrieve gait direction
|
||||||
|
sign, direction = gait_direction
|
||||||
|
axis = ['X', 'Y', 'Z'].index(direction)
|
||||||
|
|
||||||
|
# Read trc file
|
||||||
|
Q_coords, _, time_col, trc_markers, header = read_trc(trc_path)
|
||||||
|
|
||||||
|
unit = header[2].split('\t')[4]
|
||||||
peak_prominence = .1 if unit=='m' else 1 if unit=='dm' else 10 if unit=='cm' else 100 if unit=='mm' else np.inf
|
peak_prominence = .1 if unit=='m' else 1 if unit=='dm' else 10 if unit=='cm' else 100 if unit=='mm' else np.inf
|
||||||
|
|
||||||
# Right and left heel strikes
|
RHeel_idx, RBigToe_idx, LHeel_idx, LBigToe_idx, Hip_idx = [trc_markers.index(m) for m in markers]
|
||||||
frame_RHS = signal.find_peaks(eval(sign+'(XRHeel-XRSacrum)'),prominence=peak_prominence)[0]
|
RHeel_df, RBigToe_df, LHeel_df, LBigToe_df, Hip_df = (Q_coords.iloc[:,axis+idx*3] for idx in [RHeel_idx, RBigToe_idx, LHeel_idx, LBigToe_idx, Hip_idx])
|
||||||
t_RHS = data.loc[frame_RHS, 'Time'].tolist()
|
|
||||||
|
|
||||||
frame_LHS = signal.find_peaks(eval(sign+'(XLHeel-XLSacrum)'),prominence=peak_prominence)[0]
|
# Find gait events
|
||||||
t_LHS = data.loc[frame_LHS, 'Time'].tolist()
|
max_r_heel_hip_proj = sign*(RHeel_df-Hip_df)
|
||||||
|
frame_Ron = signal.find_peaks(max_r_heel_hip_proj, prominence=peak_prominence)[0]
|
||||||
|
t_Ron = time_col[frame_Ron].tolist()
|
||||||
|
|
||||||
# Right and left toe offs
|
max_l_heel_hip_proj = sign*(LHeel_df-Hip_df)
|
||||||
frame_RTO = signal.find_peaks(eval(sign+'-(XRToe-XRSacrum)'),prominence=peak_prominence)[0]
|
frame_Lon = signal.find_peaks(max_l_heel_hip_proj, prominence=peak_prominence)[0]
|
||||||
t_RTO = data.loc[frame_RTO, 'Time'].tolist()
|
t_Lon = time_col[frame_Lon].tolist()
|
||||||
|
|
||||||
frame_LTO = signal.find_peaks(eval(sign+'-(XLToe-XLSacrum)'),prominence=peak_prominence)[0]
|
max_r_hip_toe_proj = sign*(Hip_df-RBigToe_df)
|
||||||
t_LTO = data.loc[frame_LTO, 'Time'].tolist()
|
frame_Roff = signal.find_peaks(max_r_hip_toe_proj, prominence=peak_prominence)[0]
|
||||||
|
t_Roff = time_col[frame_Roff].tolist()
|
||||||
|
|
||||||
return t_RHS, t_LHS, t_RTO, t_LTO
|
max_l_hip_toe_proj = sign*(Hip_df-LBigToe_df)
|
||||||
|
frame_Loff = signal.find_peaks(max_l_hip_toe_proj, prominence=peak_prominence)[0]
|
||||||
|
t_Loff = time_col[frame_Loff].tolist()
|
||||||
|
|
||||||
|
# # Clean gait events
|
||||||
|
# frame_Ron, frame_Lon, frame_Roff, frame_Loff = clean_gait_events((frame_Ron, frame_Lon, frame_Roff, frame_Loff))
|
||||||
|
# t_Ron, t_Lon, t_Roff, t_Loff = clean_gait_events((t_Ron, t_Lon, t_Roff, t_Loff))
|
||||||
|
|
||||||
|
# Plot
|
||||||
|
if plot:
|
||||||
|
plt.plot(time_col, max_r_heel_hip_proj, label='Right on')
|
||||||
|
plt.plot(time_col[frame_Ron], max_r_heel_hip_proj[frame_Ron], '+')
|
||||||
|
|
||||||
|
plt.plot(time_col, max_l_heel_hip_proj, label='Left on')
|
||||||
|
plt.plot(time_col[frame_Lon], max_l_heel_hip_proj[frame_Lon], '+')
|
||||||
|
|
||||||
|
plt.plot(time_col, max_r_hip_toe_proj, label='Right off')
|
||||||
|
plt.plot(time_col[frame_Roff], max_r_hip_toe_proj[frame_Roff], '+')
|
||||||
|
|
||||||
|
plt.plot(time_col, max_l_hip_toe_proj, label='Left off')
|
||||||
|
plt.plot(time_col[frame_Loff], max_l_hip_toe_proj[frame_Loff], '+')
|
||||||
|
|
||||||
|
plt.title('Gait events')
|
||||||
|
plt.xlabel('Time (s)')
|
||||||
|
plt.ylabel('Distance (cm)')
|
||||||
|
plt.legend()
|
||||||
|
plt.show()
|
||||||
|
|
||||||
|
print('Times:')
|
||||||
|
print('Right on:', t_Ron)
|
||||||
|
print('Right off:', t_Roff)
|
||||||
|
print('Left on:', t_Lon)
|
||||||
|
print('Left off:', t_Loff)
|
||||||
|
print('\nFrames:')
|
||||||
|
print('Right on:', frame_Ron)
|
||||||
|
print('Right off:', frame_Roff)
|
||||||
|
print('Left on:', frame_Lon)
|
||||||
|
print('Left off:', frame_Loff)
|
||||||
|
|
||||||
|
return (t_Ron, t_Lon, t_Roff, t_Loff), (frame_Ron, frame_Lon, frame_Roff, frame_Loff)
|
||||||
|
|
||||||
|
|
||||||
def trc_gaitevents_func(*args):
|
def gait_events_height_coords(trc_path, up_direction, height_threshold=6, filter_fs=10, markers=['RBigToe', 'LBigToe'], plot=True):
|
||||||
'''
|
'''
|
||||||
Determine gait events according to Zeni et al. (2008).
|
Determine gait on and off with "height_coordinates" method
|
||||||
Write them in gaitevents.txt (append results if file already exists).
|
|
||||||
|
|
||||||
t_HeelStrike = max(XHeel - Xsacrum)
|
on = YToe < height_threshold
|
||||||
t_ToeOff = min(XToe - XSacrum)
|
off = YToe > height_threshold
|
||||||
|
++: Best results for running and walking
|
||||||
|
++: Works if the person is not going on a straight line
|
||||||
|
--: Does not work is the person is grazing the ground
|
||||||
|
--: Does not work if the field is not flat
|
||||||
|
--: height_threshold might need to be tuned
|
||||||
|
--: Heel point might be more accurate than toe point for walking
|
||||||
|
|
||||||
Reference:
|
INPUTS:
|
||||||
“Two simple methods for determining gait events during treadmill and
|
- trc_path: path to the trc file
|
||||||
overground walking using kinematic data.”
|
- up_direction: tuple (sign, direction) with sign in {-1, 1} and direction in {'X', 'Y', 'Z'}
|
||||||
Gait & posture vol. 27,4 (2008): 710-4. doi:10.1016/j.gaitpost.2007.07.007
|
- height_threshold: height below which the foot is considered to have touched the ground (cm)
|
||||||
|
- filter_fs: butterworth filter cutoff frequency (Hz)
|
||||||
|
- markers: list of marker names in the following order: [right_toe_marker, left_toe_marker]
|
||||||
|
- plot: plot results or not (boolean)
|
||||||
|
|
||||||
|
OUTPUTS:
|
||||||
|
- t_Ron: list of right on times
|
||||||
|
- t_Lon: list of left on times
|
||||||
|
- t_Roff: list of right off times
|
||||||
|
- t_Loff: list of left off times
|
||||||
|
'''
|
||||||
|
|
||||||
|
# Retrieve gait direction
|
||||||
|
sign, direction = up_direction
|
||||||
|
axis = ['X', 'Y', 'Z'].index(direction)
|
||||||
|
|
||||||
|
# Read trc file
|
||||||
|
Q_coords, _, time_col, trc_markers, header = read_trc(trc_path)
|
||||||
|
unit = header[2].split('\t')[4]
|
||||||
|
unit_factor = 100 if unit=='m' else 10 if unit=='dm' else 1 if unit=='cm' else .1 if unit=='mm' else np.inf
|
||||||
|
Q_coords *= unit_factor
|
||||||
|
|
||||||
|
# Calculate height
|
||||||
|
Y_height = Q_coords.iloc[:,axis::3]
|
||||||
|
Y_height.columns = trc_markers
|
||||||
|
Rfoot_height, Lfoot_height = (Y_height[m] for m in markers)
|
||||||
|
|
||||||
|
dt = time_col.diff().mean()
|
||||||
|
b, a = signal.butter(4/2, filter_fs*dt*2, 'low', analog=False)
|
||||||
|
Rfoot_height_filtered = pd.Series(signal.filtfilt(b, a, Rfoot_height[1:]), name=Rfoot_height.name)
|
||||||
|
Lfoot_height_filtered = pd.Series(signal.filtfilt(b, a, Lfoot_height[1:]), name=Lfoot_height.name)
|
||||||
|
|
||||||
|
# Find gait events
|
||||||
|
low_Rfoot_height = Rfoot_height_filtered<height_threshold
|
||||||
|
frame_Ron, frame_Roff = start_end_true_seq(low_Rfoot_height)
|
||||||
|
if 0 in frame_Ron: frame_Ron.remove(0)
|
||||||
|
t_Ron, t_Roff = time_col[frame_Ron].tolist(), time_col[frame_Roff].tolist()
|
||||||
|
|
||||||
|
low_Lfoot_height = Lfoot_height_filtered<height_threshold
|
||||||
|
frame_Lon, frame_Loff = start_end_true_seq(low_Lfoot_height)
|
||||||
|
if 0 in frame_Lon: frame_Lon.remove(0)
|
||||||
|
t_Lon, t_Loff = time_col[frame_Lon].tolist(), time_col[frame_Loff].tolist()
|
||||||
|
|
||||||
|
# # Clean gait events
|
||||||
|
# frame_Ron, frame_Lon, frame_Roff, frame_Loff = clean_gait_events((frame_Ron, frame_Lon, frame_Roff, frame_Loff))
|
||||||
|
# t_Ron, t_Lon, t_Roff, t_Loff = clean_gait_events((t_Ron, t_Lon, t_Roff, t_Loff))
|
||||||
|
|
||||||
|
# Plot
|
||||||
|
if plot:
|
||||||
|
plt.plot(time_col[1:], Rfoot_height_filtered, label='Right foot height filtered')
|
||||||
|
plt.plot(time_col[1:][frame_Ron], Rfoot_height_filtered[frame_Ron], '+')
|
||||||
|
plt.plot(time_col[1:][frame_Roff], Rfoot_height_filtered[frame_Roff], '+')
|
||||||
|
|
||||||
|
plt.plot(time_col[1:], Lfoot_height_filtered, label='Left foot height filtered')
|
||||||
|
plt.plot(time_col[1:][frame_Lon], Lfoot_height_filtered[frame_Lon], '+')
|
||||||
|
plt.plot(time_col[1:][frame_Loff], Lfoot_height_filtered[frame_Loff], '+')
|
||||||
|
|
||||||
|
plt.title('Gait events')
|
||||||
|
plt.xlabel('Time (s)')
|
||||||
|
plt.ylabel('Height (cm)')
|
||||||
|
plt.legend()
|
||||||
|
plt.show()
|
||||||
|
|
||||||
|
print('Times:')
|
||||||
|
print('Right on:', t_Ron)
|
||||||
|
print('Right off:', t_Roff)
|
||||||
|
print('Left on:', t_Lon)
|
||||||
|
print('Left off:', t_Loff)
|
||||||
|
print('\nFrames:')
|
||||||
|
print('Right on:', frame_Ron)
|
||||||
|
print('Right off:', frame_Roff)
|
||||||
|
print('Left on:', frame_Lon)
|
||||||
|
print('Left off:', frame_Loff)
|
||||||
|
|
||||||
|
return (t_Ron, t_Lon, t_Roff, t_Loff), (frame_Ron, frame_Lon, frame_Roff, frame_Loff)
|
||||||
|
|
||||||
|
|
||||||
|
def gait_events_fwd_vel(trc_path, gait_direction, forward_velocity_threshold=1.5, filter_fs=10, markers=['RBigToe', 'LBigToe'], plot=True):
|
||||||
|
'''
|
||||||
|
Determine gait on and off with "forward_velocity" method
|
||||||
|
|
||||||
|
on = VToe < forward_velocity_threshold
|
||||||
|
off = VToe > forward_velocity_threshold
|
||||||
|
++: Works for running and walking
|
||||||
|
--: Tends to anticipate off if marker is not at the tip of the toe
|
||||||
|
--: forward_velocity_threshold might need to be tuned
|
||||||
|
--: does not work if the person is not going on a straight line
|
||||||
|
|
||||||
|
INPUTS:
|
||||||
|
- trc_path: path to the trc file
|
||||||
|
- gait_direction: tuple (sign, direction) with sign in {-1, 1} and direction in {'X', 'Y', 'Z'}
|
||||||
|
- forward_velocity_threshold: forward velocity below which the foot is considered to have touched the ground (m/s)
|
||||||
|
- filter_fs: butterworth filter cutoff frequency (Hz)
|
||||||
|
- markers: list of marker names in the following order: [right_toe_marker, left_toe_marker]
|
||||||
|
- plot: plot results or not (boolean)
|
||||||
|
|
||||||
|
OUTPUTS:
|
||||||
|
- t_Ron: list of right on times
|
||||||
|
- t_Lon: list of left on times
|
||||||
|
- t_Roff: list of right off times
|
||||||
|
- t_Loff: list of left off times
|
||||||
|
'''
|
||||||
|
|
||||||
|
# Retrieve gait direction
|
||||||
|
sign, direction = gait_direction
|
||||||
|
axis = ['X', 'Y', 'Z'].index(direction)
|
||||||
|
|
||||||
|
# Read trc file
|
||||||
|
Q_coords, _, time_col, trc_markers, header = read_trc(trc_path)
|
||||||
|
unit = header[2].split('\t')[4]
|
||||||
|
unit_factor = 1 if unit=='m' else 10 if unit=='dm' else 100 if unit=='cm' else 1000 if unit=='mm' else np.inf
|
||||||
|
forward_velocity_threshold *= unit_factor
|
||||||
|
Q_coords *= unit_factor
|
||||||
|
|
||||||
|
# Calculate speed
|
||||||
|
dt = time_col.diff().mean()
|
||||||
|
b, a = signal.butter(4/2, filter_fs*dt*2, 'low', analog=False)
|
||||||
|
X_speed = Q_coords.iloc[:,axis::3].diff()/dt
|
||||||
|
X_speed.columns = trc_markers
|
||||||
|
Rfoot_speed, Lfoot_speed = (X_speed[m] for m in markers)
|
||||||
|
|
||||||
|
Rfoot_speed = Rfoot_speed.where(Rfoot_speed<0, other=0) if sign==-1 else Rfoot_speed.where(Rfoot_speed>0, other=0)
|
||||||
|
Rfoot_speed = Rfoot_speed.abs()
|
||||||
|
# Rfoot_speed_filtered = pd.Series(signal.filtfilt(b, a, Rfoot_speed[1:]), name=Rfoot_speed.name)
|
||||||
|
Rfoot_speed_filtered = pd.Series(gaussian_filter1d(Rfoot_speed[1:], 5), name=Rfoot_speed.name)
|
||||||
|
|
||||||
|
Lfoot_speed = Lfoot_speed.where(Lfoot_speed<0, other=0) if sign==-1 else Lfoot_speed.where(Lfoot_speed>0, other=0)
|
||||||
|
Lfoot_speed = Lfoot_speed.abs()
|
||||||
|
# Lfoot_speed_filtered = pd.Series(signal.filtfilt(b, a, Lfoot_speed[1:]), name=Lfoot_speed.name)
|
||||||
|
Lfoot_speed_filtered = pd.Series(gaussian_filter1d(Lfoot_speed[1:], 5), name=Lfoot_speed.name)
|
||||||
|
|
||||||
|
# Find gait events
|
||||||
|
low_Rfoot_speed = Rfoot_speed_filtered<forward_velocity_threshold
|
||||||
|
frame_Ron, frame_Roff = start_end_true_seq(low_Rfoot_speed)
|
||||||
|
if 0 in frame_Ron: frame_Ron.remove(0)
|
||||||
|
t_Ron, t_Roff = time_col[frame_Ron].tolist(), time_col[frame_Roff].tolist()
|
||||||
|
|
||||||
|
low_Lfoot_speed = Lfoot_speed_filtered<forward_velocity_threshold
|
||||||
|
frame_Lon, frame_Loff = start_end_true_seq(low_Lfoot_speed)
|
||||||
|
if 0 in frame_Lon: frame_Lon.remove(0)
|
||||||
|
t_Lon, t_Loff = time_col[frame_Lon].tolist(), time_col[frame_Loff].tolist()
|
||||||
|
|
||||||
|
# # Clean gait events
|
||||||
|
# frame_Ron, frame_Lon, frame_Roff, frame_Loff = clean_gait_events((frame_Ron, frame_Lon, frame_Roff, frame_Loff))
|
||||||
|
# t_Ron, t_Lon, t_Roff, t_Loff = clean_gait_events((t_Ron, t_Lon, t_Roff, t_Loff))
|
||||||
|
|
||||||
|
# Plot
|
||||||
|
if plot:
|
||||||
|
plt.plot(time_col[1:], Rfoot_speed_filtered, label='Right foot speed filtered')
|
||||||
|
plt.plot(time_col[1:][frame_Ron], Rfoot_speed_filtered[frame_Ron], '+')
|
||||||
|
plt.plot(time_col[1:][frame_Roff], Rfoot_speed_filtered[frame_Roff], '+')
|
||||||
|
|
||||||
|
plt.plot(time_col[1:], Lfoot_speed_filtered, label='Left foot speed filtered')
|
||||||
|
plt.plot(time_col[1:][frame_Lon], Lfoot_speed_filtered[frame_Lon], '+')
|
||||||
|
plt.plot(time_col[1:][frame_Loff], Lfoot_speed_filtered[frame_Loff], '+')
|
||||||
|
|
||||||
|
plt.title('Gait events')
|
||||||
|
plt.xlabel('Time (s)')
|
||||||
|
plt.ylabel('Speed (m/s)')
|
||||||
|
plt.legend()
|
||||||
|
plt.show()
|
||||||
|
|
||||||
|
print('Times:')
|
||||||
|
print('Right on:', t_Ron)
|
||||||
|
print('Right off:', t_Roff)
|
||||||
|
print('Left on:', t_Lon)
|
||||||
|
print('Left off:', t_Loff)
|
||||||
|
print('\nFrames:')
|
||||||
|
print('Right on:', frame_Ron)
|
||||||
|
print('Right off:', frame_Roff)
|
||||||
|
print('Left on:', frame_Lon)
|
||||||
|
print('Left off:', frame_Loff)
|
||||||
|
|
||||||
|
return (t_Ron, t_Lon, t_Roff, t_Loff), (frame_Ron, frame_Lon, frame_Roff, frame_Loff)
|
||||||
|
|
||||||
|
|
||||||
|
def trc_gaitevents_func(**args):
|
||||||
|
'''
|
||||||
|
Determine gait on and off from a TRC file of point coordinates.
|
||||||
|
Three available methods, each of them with their own pros and cons:
|
||||||
|
|
||||||
|
- "forward_coordinates":
|
||||||
|
on = max(XHeel - Xsacrum)
|
||||||
|
off = min(XToe - XSacrum)
|
||||||
|
++: Works well for walking (Zeni et al., 2008)
|
||||||
|
++: No argument nor tuning necessary
|
||||||
|
--: Not suitable for running
|
||||||
|
--: does not work if the person is not going on a straight line
|
||||||
|
|
||||||
|
- "height_coordinates":
|
||||||
|
on = YToe < height_threshold
|
||||||
|
off = YToe > height_threshold
|
||||||
|
++: Best results for running and walking
|
||||||
|
++: Works if the person is not going on a straight line
|
||||||
|
--: Does not work is the person is grazing the ground
|
||||||
|
--: Does not work if the field is not flat
|
||||||
|
--: height_threshold might need to be tuned
|
||||||
|
--: Heel point might be more accurate than toe point for walking
|
||||||
|
|
||||||
|
- "forward_velocity":
|
||||||
|
on = VToe < forward_velocity_threshold
|
||||||
|
off = VToe > forward_velocity_threshold
|
||||||
|
++: Works for running
|
||||||
|
--: More sensitive to noise
|
||||||
|
--: Tends to anticipate off if the marker is not at the tip of the toe
|
||||||
|
--: forward_velocity_threshold might need to be tuned
|
||||||
|
--: Does not work if the person is not going on a straight line
|
||||||
|
|
||||||
Usage:
|
Usage:
|
||||||
Replace constants with the appropriate marker names in trc_gaitevents.py.
|
List of available arguments:
|
||||||
If direction is negative, you need to include an equal sign in the argument,
|
python -m trc_gaitevents -h
|
||||||
eg -d=-Z or --gait_direction=-Z
|
|
||||||
|
|
||||||
import trc_gaitevents; trc_gaitevents.trc_gaitevents_func(r'<input_trc_file>', '<gait_direction>')
|
import trc_gaitevents; trc_gaitevents.trc_gaitevents_func(trc_path=r'<input_trc_file>')
|
||||||
OR trc_gaitevents -i input_trc_file --gait_direction Z
|
OR import trc_gaitevents; trc_gaitevents.trc_gaitevents_func(trc_path=r'<input_trc_file>', method='forward_coordinates', gait_direction='-X', plot=True, save_output=True, output_file='gaitevents.txt')
|
||||||
OR trc_gaitevents -i input_trc_file --gait_direction=-Z
|
OR import trc_gaitevents; trc_gaitevents.trc_gaitevents_func(trc_path=r'<input_trc_file>', method='height_coordinates', up_direction='Y', height_threshold=6, right_toe_marker='RBigToe', left_toe_marker='LBigToe')
|
||||||
|
OR import trc_gaitevents; trc_gaitevents.trc_gaitevents_func(trc_path=r'<input_trc_file>', method='forward_velocity', gait_direction='-Z', forward_velocity_threshold=1.5, right_toe_marker='RBigToe', left_toe_marker='LBigToe')
|
||||||
|
|
||||||
|
python -m trc_gaitevents -i input_trc_file
|
||||||
|
OR python -m trc_gaitevents -i input_trc_file --method forward_coordinates --gait_direction=-X --plot True --save_output True --output_file gaitevents.txt
|
||||||
|
OR python -m trc_gaitevents -i input_trc_file --method height_coordinates --up_direction=Y --height_threshold 6 --right_toe_marker RBigToe --left_toe_marker LBigToe
|
||||||
|
OR python -m trc_gaitevents -i input_trc_file --method forward_velocity --gait_direction=-Z --forward_velocity_threshold 1.5 --right_toe_marker RBigToe --left_toe_marker LBigToe
|
||||||
'''
|
'''
|
||||||
|
|
||||||
try:
|
# Retrieve arguments
|
||||||
trc_path = args[0].get('input_file') # invoked with argparse
|
trc_path = args.get('trc_path')
|
||||||
gait_direction = args[0]['gait_direction']
|
gait_direction = args.get('gait_direction')
|
||||||
except:
|
up_direction = args.get('up_direction')
|
||||||
trc_path = args[0] # invoked as a function
|
method = args.get('method')
|
||||||
try:
|
forward_velocity_threshold = args.get('forward_velocity_threshold')
|
||||||
gait_direction = args[1]
|
height_threshold = args.get('height_threshold')
|
||||||
except:
|
sacrum_marker = args.get('sacrum_marker')
|
||||||
gait_direction = 'Z'
|
right_heel_marker = args.get('right_heel_marker')
|
||||||
|
right_toe_marker = args.get('right_toe_marker')
|
||||||
|
left_heel_marker = args.get('left_heel_marker')
|
||||||
|
left_toe_marker = args.get('left_toe_marker')
|
||||||
|
plot = args.get('plot')
|
||||||
|
save_output = args.get('save_output')
|
||||||
|
output_file = args.get('output_file')
|
||||||
|
|
||||||
|
filter_fs = 6
|
||||||
|
|
||||||
|
# If invoked via a function
|
||||||
|
if gait_direction == None: gait_direction = '+X'
|
||||||
|
if up_direction == None: up_direction = '+Y'
|
||||||
|
if method == None: method = 'height_coordinates'
|
||||||
|
if forward_velocity_threshold == None: forward_velocity_threshold = 1.5
|
||||||
|
if height_threshold == None: height_threshold = 6
|
||||||
|
if sacrum_marker == None: sacrum_marker = 'Hip'
|
||||||
|
if right_heel_marker == None: right_heel_marker = 'RHeel'
|
||||||
|
if right_toe_marker == None: right_toe_marker = 'RBigToe'
|
||||||
|
if left_heel_marker == None: left_heel_marker = 'LHeel'
|
||||||
|
if left_toe_marker == None: left_toe_marker = 'LBigToe'
|
||||||
|
if plot == None: plot = True
|
||||||
|
if save_output == None: save_output = True
|
||||||
|
if output_file == None: output_file = 'gaitevents.txt'
|
||||||
|
|
||||||
|
# In case of a sign in direction (eg -X)
|
||||||
|
if len(gait_direction)==1:
|
||||||
|
gait_direction = +1, gait_direction
|
||||||
|
elif len(gait_direction)==2:
|
||||||
|
gait_direction = int(gait_direction[0]+'1'), gait_direction[1]
|
||||||
|
if len(up_direction)==1:
|
||||||
|
up_direction = +1, up_direction
|
||||||
|
elif len(up_direction)==2:
|
||||||
|
up_direction = int(up_direction[0]+'1'), up_direction
|
||||||
|
|
||||||
|
# Retrieve gait events
|
||||||
|
if method == 'forward_coordinates':
|
||||||
|
print('Method: forward_coordinates')
|
||||||
|
markers = [right_heel_marker, right_toe_marker, left_heel_marker, left_toe_marker, sacrum_marker]
|
||||||
|
(t_Ron, t_Lon, t_Roff, t_Loff), (frame_Ron, frame_Lon, frame_Roff, frame_Loff) \
|
||||||
|
= gait_events_fwd_coords(trc_path, gait_direction, markers=markers, plot=plot)
|
||||||
|
elif method == 'height_coordinates':
|
||||||
|
print(f'Method: height_coordinates. Height threshold: {height_threshold} cm')
|
||||||
|
markers = [right_toe_marker, left_toe_marker]
|
||||||
|
(t_Ron, t_Lon, t_Roff, t_Loff), (frame_Ron, frame_Lon, frame_Roff, frame_Loff) \
|
||||||
|
= gait_events_height_coords(trc_path, up_direction, height_threshold=height_threshold, filter_fs=filter_fs, markers=markers, plot=plot)
|
||||||
|
elif method == 'forward_velocity':
|
||||||
|
print(f'Method: forward_velocity. Forward velocity threshold: {forward_velocity_threshold} m/s')
|
||||||
|
markers = [right_toe_marker, left_toe_marker]
|
||||||
|
(t_Ron, t_Lon, t_Roff, t_Loff), (frame_Ron, frame_Lon, frame_Roff, frame_Loff) \
|
||||||
|
= gait_events_fwd_vel(trc_path, gait_direction, forward_velocity_threshold=forward_velocity_threshold, filter_fs=filter_fs, markers=markers, plot=plot)
|
||||||
|
|
||||||
|
if save_output or save_output==None:
|
||||||
trc_dir = os.path.dirname(trc_path)
|
trc_dir = os.path.dirname(trc_path)
|
||||||
trc_name = os.path.basename(trc_path)
|
trc_name = os.path.basename(trc_path)
|
||||||
|
if output_file == None: output_file = 'gaitevents.txt'
|
||||||
t_RHS, t_LHS, t_RTO, t_LTO = gait_events(trc_path, gait_direction)
|
with open(os.path.join(trc_dir, output_file), 'a') as gaitevents:
|
||||||
|
|
||||||
with open(os.path.join(trc_dir, 'gaitevents.txt'), 'a') as gaitevents:
|
|
||||||
L = trc_name + '\n'
|
L = trc_name + '\n'
|
||||||
L += 'Right Heel strikes: ' + str(t_RHS) + '\n'
|
L += f'Method: {method}. '
|
||||||
L += 'Left Heel strikes: ' + str(t_LHS) + '\n'
|
L+= f'Height threshold: {height_threshold}\n' if method=='height_coordinates' else f'Forward velovity threshold: {forward_velocity_threshold}\n' if method == 'forward_velocity' else '\n'
|
||||||
L += 'Right Toe off: ' + str(t_RTO) + '\n'
|
L += 'Times:\n'
|
||||||
L += 'Left Toe off: ' + str(t_LTO) + '\n\n'
|
L += '\tRight on: ' + str(t_Ron) + '\n'
|
||||||
|
L += '\tLeft on: ' + str(t_Lon) + '\n'
|
||||||
|
L += '\tRight off: ' + str(t_Roff) + '\n'
|
||||||
|
L += '\tLeft off: ' + str(t_Loff) + '\n'
|
||||||
|
L += 'Frames:\n'
|
||||||
|
L += '\tRight on: ' + str(frame_Ron) + '\n'
|
||||||
|
L += '\tLeft on: ' + str(frame_Lon) + '\n'
|
||||||
|
L += '\tRight off: ' + str(frame_Roff) + '\n'
|
||||||
|
L += '\tLeft off: ' + str(frame_Loff) + '\n\n'
|
||||||
|
|
||||||
gaitevents.write(L)
|
gaitevents.write(L)
|
||||||
|
|
||||||
|
return (t_Ron, t_Lon, t_Roff, t_Loff), (frame_Ron, frame_Lon, frame_Roff, frame_Loff)
|
||||||
|
|
||||||
|
|
||||||
if __name__ == '__main__':
|
if __name__ == '__main__':
|
||||||
parser = argparse.ArgumentParser()
|
parser = argparse.ArgumentParser(description='Determine gait on and off with "forward_coordinates", "height_coordinates", or "forward_velocity" method. More details in the file description.')
|
||||||
parser.add_argument('-i', '--input_file', required = True, help='trc input file')
|
parser.add_argument('-i', '--trc_path', required = True, help='Trc input file')
|
||||||
parser.add_argument('-d', '--gait_direction', default = 'Z', required = False, help='direction of the gait. If negative, you need to include an equal sign in the argument, eg -d=-Z')
|
parser.add_argument('-g', '--gait_direction', default = 'X', required = False, help='Direction of the gait. "X", "Y", "Z", "-X", "-Y", or "-Z". Default: "X". Requires an equal sign if negative, eg -g=-X')
|
||||||
|
parser.add_argument('-u', '--up_direction', default = 'Y', required = False, help='Up direction. "X", "Y", or "Z", "-X", "-Y", or "-Z". Default: "Y"')
|
||||||
|
parser.add_argument('-m', '--method', default = 'height_coordinates', required = False, help='Method to determine gait events. "forward_coordinates", "height_coordinates", or "forward_velocity". Default:"height_coordinates"')
|
||||||
|
parser.add_argument('-V', '--forward_velocity_threshold', default = 1, type=float, required = False, help='Forward velocity below which the foot is considered to have touched the ground (m/s). Used if method is forward_velocity. Default: 1.5')
|
||||||
|
parser.add_argument('-H', '--height_threshold', default = 6, type=float, required = False, help='Height below which the foot is considered to have touched the ground (cm). Used if method is height_coordinates. Default: 6')
|
||||||
|
parser.add_argument('--sacrum_marker', default = 'Hip', required = False, help='Hip marker name. Default: "Hip"')
|
||||||
|
parser.add_argument('--right_heel_marker', default = 'RHeel', required = False, help='Right heel marker name. Default: "RHeel"')
|
||||||
|
parser.add_argument('--right_toe_marker', default = 'RBigToe', required = False, help='Right toe marker name. Default: "RBigToe"')
|
||||||
|
parser.add_argument('--left_heel_marker', default = 'LHeel', required = False, help='Left heel marker name. Default: "LHeel"')
|
||||||
|
parser.add_argument('--left_toe_marker', default = 'LBigToe', required = False, help='Left toe marker name. Default: "LBigToe"')
|
||||||
|
parser.add_argument('-p', '--plot', default = True, required = False, help='Plot results. Default: True')
|
||||||
|
parser.add_argument('-s', '--save_output', default = True, required = False, help='Save output in csv file. Default: True')
|
||||||
|
parser.add_argument('-o', '--output_file', default = 'gaitevents.txt', required = False, help='Output file name. Default: "gaitevents.txt"')
|
||||||
|
|
||||||
args = vars(parser.parse_args())
|
args = vars(parser.parse_args())
|
||||||
|
|
||||||
trc_gaitevents_func(args)
|
trc_gaitevents_func(**args)
|
||||||
|
|
||||||
|
|||||||
Loading…
Reference in New Issue
Block a user