def endurance_summary(FitFilePath, ConfigFile=None, OutStream=sys.stdout):
(FilePath, FitFileName) = os.path.split(FitFilePath)
if ConfigFile is None:
ConfigFile = FindConfigFile('', FilePath)
if (ConfigFile is None) or (not os.path.exists(ConfigFile)):
raise IOError('Configuration file not specified or found')
#
# Parse the configuration file
#
from ConfigParser import ConfigParser
config = ConfigParser()
config.read(ConfigFile)
print >> OutStream, 'reading config file ' + ConfigFile
ThresholdPower = config.getfloat( 'power', 'ThresholdPower' )
ThresholdHR = config.getfloat( 'power', 'ThresholdHR' )
print >> OutStream, 'ThresholdPower: ', ThresholdPower
print >> OutStream, 'ThresholdHR : ', ThresholdHR
# power zones from "Cyclist's Training Bible", 5th ed., by Joe Friel, p51
FTP = ThresholdPower
pZones = { 1 : [ 0 , 0.55*FTP ],
2 : [ 0.55*FTP, 0.75*FTP ],
3 : [ 0.75*FTP, 0.90*FTP ],
4 : [ 0.90*FTP, 1.05*FTP ],
5 : [ 1.05*FTP, 1.20*FTP ],
6 : [ 1.20*FTP, 1.50*FTP ],
7 : [ 1.50*FTP, 2.50*FTP ]}
# heart-rate zones from "Cyclist's Training Bible" 5th ed. by Joe Friel, p50
FTHR = ThresholdHR
hZones = { 1 : [ 0 , 0.82*FTHR ], # 1
2 : [ 0.82*FTHR, 0.89*FTHR ], # 2
3 : [ 0.89*FTHR, 0.94*FTHR ], # 3
4 : [ 0.94*FTHR, 1.00*FTHR ], # 4
5 : [ 1.00*FTHR, 1.03*FTHR ], # 5a
6 : [ 1.03*FTHR, 1.06*FTHR ], # 5b
7 : [ 1.07*FTHR, 1.15*FTHR ]} # 5c
# get zone bounds for plotting
p_zone_bounds = [ pZones[1][0],
pZones[2][0],
pZones[3][0],
pZones[4][0],
pZones[5][0],
pZones[6][0],
pZones[7][0],
pZones[7][1] ]
h_zone_bounds = [ 0.4*FTHR, # better plotting
hZones[2][0],
hZones[3][0],
hZones[4][0],
hZones[5][0],
hZones[6][0],
hZones[7][0],
hZones[7][1] ]
from datetime import datetime
from fitparse import Activity
from activity_tools import extract_activity_signals
required_signals = [ 'power',
'heart_rate' ]
# get the signals
activity = Activity(FitFilePath)
signals = extract_activity_signals(activity, resample='existing')
if not all( s in signals.keys() for s in required_signals ):
msg = 'required signals not in file'
print >> OutStream, msg
print >> OutStream, 'Signals required:'
for s in required_signals:
print >> OutStream, ' ' + s
print >> OutStream, 'Signals contained:'
for s in signals.keys():
print >> OutStream, ' ' + s
raise IOError(msg)
'''
####################
# Get Records of type 'lap'
# types: [ 'record', 'lap', 'event', 'session', 'activity', ... ]
records = activity.get_records_by_type('lap')
current_record_number = 0
elapsed_time = []
timer_time = []
avg_heart_rate = []
avg_power = []
avg_cadence = []
max_heart_rate = []
balance = []
lap_timestamp = []
lap_start_time = []
FirstIter = True
for record in records:
# Print record number
current_record_number += 1
#print (" Record #%d " % current_record_number).center(40, '-')
# Get the list of valid fields on this record
valid_field_names = record.get_valid_field_names()
for field_name in valid_field_names:
# Get the data and units for the field
field_data = record.get_data(field_name)
field_units = record.get_units(field_name)
## Print what we've got!
#if field_units:
# print >> OutStream, " * %s: %s %s" % (field_name, field_data, field_units)
#else:
# print >> OutStream, " * %s: %s" % (field_name, field_data)
if 'timestamp' in field_name:
lap_timestamp.append( field_data )
if 'start_time' in field_name:
lap_start_time.append( field_data )
if 'total_elapsed_time' in field_name:
elapsed_time.append( field_data )
if 'total_timer_time' in field_name:
timer_time.append( field_data )
if 'avg_power' in field_name:
avg_power.append( field_data )
# avg_heart_rate is in a lap record
if 'avg_heart_rate' in field_name:
avg_heart_rate.append(field_data)
if 'max_heart_rate' in field_name:
max_heart_rate.append(field_data)
if 'avg_cadence' in field_name:
avg_cadence.append(field_data)
if 'left_right_balance' in field_name:
balance.append(field_data)
#print
####################
'''
#
# extract lap results
#
from fitparse import Activity
from activity_tools import extract_activity_laps
import numpy as np
activity = Activity(FitFilePath)
laps = extract_activity_laps(activity)
avg_power = laps['power']
time = laps['time']
cadence = laps['cadence']
avg_heart_rate = laps['avg_hr']
max_heart_rate = laps['max_hr']
balance = laps['balance']
lap_start_time = laps['start_time']
lap_timestamp = laps['timestamp' ]
timer_time = laps['total_timer_time']
elapsed_time = laps['total_elapsed_time']
IntervalThreshold = 0.0 # get all laps (0.72*FTP)
from numpy import nonzero, array, arange, zeros, average, logical_and
# resample power to constant-increment (1 Hz) with zeros at missing samples
time_idx = signals['time'].astype('int')
power_vi = signals['power']
heart_rate_vi = signals['heart_rate']
nScans = time_idx[-1]+1
time_ci = arange(nScans)
power = zeros(nScans)
power[time_idx] = power_vi
heart_rate_ci = zeros(nScans)
heart_rate_ci[time_idx] = heart_rate_vi
t0 = signals['metadata']['timestamp']
print >> OutStream, 'signal timestamp: ', t0.time()
# plot lap results as continuous time signals
lap_avg_hr_c = zeros(nScans)
lap_avg_power_c = zeros(nScans)
lap_norm_power_c = zeros(nScans)
# compute the 30-second, moving-average power signal.
p30 = BackwardMovingAverage( power )
#
# compute lap metrics
#
print >> OutStream, 'lap results:'
nLaps = len(elapsed_time)
vi_time_vector = signals['time']
lap_avg_power = zeros(nLaps)
lap_norm_power = zeros(nLaps)
lap_avg_hr = zeros(nLaps)
lap_if = zeros(nLaps) # intensity factor
lap_start_sec = zeros(nLaps) # lap start times in seconds
#time = array(elapsed_time)
#cadence = array(avg_cadence)
#avg_hr = array(avg_heart_rate)
#max_hr = array(max_heart_rate)
#balance = array(balance)
names1 = [ '', ' lap', ' avg', ' norm', 'avg', 'max', '' ]
names2 = [ 'lap', ' time', 'power', 'power', ' HR', ' HR', ' IF' ]
fmt = "%8s"+"%10s"+"%8s"*5
print >> OutStream, fmt % tuple(names1)
print >> OutStream, fmt % tuple(names2)
for i in range(nLaps):
# count samples in this lap
tBeg = (lap_start_time[i] - t0).total_seconds()
tEnd = (lap_timestamp[i] - t0).total_seconds()
ii = nonzero( logical_and( time_idx >= tBeg, \
time_idx < tEnd) )[0]
nPts = ii.size
lap_start_sec[i] = tBeg
lap_avg_hr[i] = average(heart_rate_vi[ii])
lap_avg_power[i] = average(power[time_idx[ii]])
lap_norm_power[i] = average( p30[time_idx[ii]]**4 )**(0.25)
lap_if[i] = lap_norm_power[i] / FTP
# duration from lap metrics
dur = (lap_timestamp[i] - lap_start_time[i]).total_seconds()
mm = timer_time[i] // 60
ss = timer_time[i] % 60
fmt = '%8d'+'%7i:%02i'+'%8i'*4 + '%8.2f'
print >> OutStream, fmt \
% ( i,
mm, ss,
avg_power[i],
lap_norm_power[i],
avg_heart_rate[i],
max_heart_rate[i],
lap_if[i] )
# plot lap results as continuous time signals
lap_avg_hr_c [time_idx[ii]] = lap_avg_hr[i]
lap_avg_power_c [time_idx[ii]] = lap_avg_power[i]
lap_norm_power_c[time_idx[ii]] = lap_norm_power[i]
#
# ride-level results
#
print >> OutStream, 'ride-level results:'
names1 = [ '', 'moving', ' avg', ' norm', 'avg', '', 'Pw:' ]
names2 = [ 'seg', ' time', 'power', 'power', ' HR', ' IF', ' HR' ]
fmt = "%8s"+"%10s"+"%8s"*5
print >> OutStream, fmt % tuple(names1)
print >> OutStream, fmt % tuple(names2)
# whole ride
tBeg = (lap_start_time[0] - t0).total_seconds()
tEnd = (lap_timestamp[-1] - t0).total_seconds()
ii = nonzero( logical_and( time_idx >= tBeg, \
time_idx < tEnd) )[0]
nPts = ii.size
dur = nPts # sample rate 1 Hz
hh = dur // 3600
mm = (dur % 3600) // 60
ss = (dur % 3600) % 60
all_avg_hr = average(heart_rate_vi[ii])
all_avg_power = average(power_vi[ii])
all_norm_power = average( p30[time_idx[ii]]**4 )**(0.25)
all_max_hr = max(heart_rate_vi[ii])
all_if = all_norm_power / FTP
# aerobic decoupling
iiH1 = ii[0:nPts/2]
h1_norm_power = average( p30[time_idx[iiH1]]**4 )**(0.25)
h1_avg_hr = average(heart_rate_vi[iiH1])
h1ef = h1_norm_power / h1_avg_hr
iiH2 = ii[nPts/2:]
h2_norm_power = average( p30[time_idx[iiH2]]**4 )**(0.25)
h2_avg_hr = average(heart_rate_vi[iiH2])
h2ef = h2_norm_power / h2_avg_hr
all_pw_hr = (h1ef-h2ef)/(h1ef)*100.0
fmt = '%8s'+'%4i:%02i:%02i'+'%8i'*3 + '%8.2f' + '%8.1f'
print >> OutStream, fmt \
% ( 'all',
hh, mm, ss,
all_avg_power,
all_norm_power,
all_avg_hr,
all_if,
all_pw_hr )
# without end laps
tBeg = (lap_start_time[1] - t0).total_seconds()
tEnd = (lap_timestamp[-2] - t0).total_seconds()
ii = nonzero( logical_and( time_idx >= tBeg, \
time_idx < tEnd) )[0]
nPts = ii.size
dur = nPts # sample rate 1 Hz
hh = dur // 3600
mm = (dur % 3600) // 60
ss = (dur % 3600) % 60
mid_avg_hr = average(heart_rate_vi[ii])
mid_avg_power = average(power_vi[ii])
mid_norm_power = average( p30[time_idx[ii]]**4 )**(0.25)
mid_max_hr = max(heart_rate_vi[ii])
mid_if = mid_norm_power / FTP
# aerobic decoupling
iiH1 = ii[0:nPts/2]
h1_norm_power = average( p30[time_idx[iiH1]]**4 )**(0.25)
h1_avg_hr = average(heart_rate_vi[iiH1])
h1ef = h1_norm_power / h1_avg_hr
iiH2 = ii[nPts/2:]
h2_norm_power = average( p30[time_idx[iiH2]]**4 )**(0.25)
h2_avg_hr = average(heart_rate_vi[iiH2])
h2ef = h2_norm_power / h2_avg_hr
mid_pw_hr = (h1ef-h2ef)/(h1ef)*100.0
fmt = '%5i-%02i'+'%4i:%02i:%02i'+'%8i'*3 + '%8.2f' + '%8.1f'
print >> OutStream, fmt \
% ( 1, nLaps-2,
hh, mm, ss,
mid_avg_power,
mid_norm_power,
mid_avg_hr,
mid_if,
mid_pw_hr )
print
print
#
# time plot
#
import matplotlib.pyplot as plt
import matplotlib.dates as md
from matplotlib.dates import date2num, DateFormatter
import datetime as dt
base = dt.datetime(2014, 1, 1, 0, 0, 0)
x = [base + dt.timedelta(seconds=t) for t in time_ci.astype('float')]
x = date2num(x) # Convert to matplotlib format
fig1, (ax0, ax1) = plt.subplots(nrows=2, sharex=True)
ax0.plot_date( x, heart_rate_ci, 'r-', linewidth=1 );
ax0.plot_date( x, lap_avg_hr_c, 'r-', linewidth=3 );
ax0.set_yticks( h_zone_bounds, minor=False)
x_laps = [ base + dt.timedelta(seconds=t) \
for t in lap_start_sec.astype('float') ]
x_laps = date2num(x_laps)
ax0.grid(True)
ax0.set_ylabel('heart rate, BPM')
ax1.plot_date( x, power, 'k-', linewidth=1 );
ax1.plot_date( x, p30, 'm-', linewidth=1);
ax1.plot_date( x, lap_avg_power_c, 'b-', linewidth=3);
ax1.plot_date( x, lap_norm_power_c, 'g-', linewidth=3);
ax1.xaxis.set_major_formatter(DateFormatter('%H:%M:%S'))
ax1.set_yticks( p_zone_bounds, minor=False)
ax1.grid(True)
ax1.set_ylabel('power, watts')
ax1.legend(['power', 'p30', 'lap_avg_power', 'lap_norm_power'],
loc='upper left');
for i in range(nLaps):
ax0.axvline( x_laps[i], label=str(i+1) )
ax1.axvline( x_laps[i], label=str(i+1) )
fig1.autofmt_xdate()
fig1.suptitle('Endurance Power Results', fontsize=20)
fig1.tight_layout()
fig1.subplots_adjust(hspace=0) # Remove horizontal space between axes
fig1.canvas.set_window_title(FitFilePath)
plt.show()
def ClosePlots():
plt.close('all')
return ClosePlots
fmt = '%20s:' + '%15s'*3
print >> OutStream, fmt % tuple(names1)
OutStream.flush()
for FitFile in fit_files:
TimerStart = time.time() # measure execution time
print >> OutStream, '#################################################'
print >> OutStream, '###%s###' % FitFile.center(43)
print >> OutStream, '#################################################'
# get the signals
FitFilePath = FilePath + FitFile
activity = Activity(FitFilePath)
signals = extract_activity_signals(activity, resample='existing')
if not all( s in signals.keys() for s in required_signals ):
msg = 'required signals not in file'
print >> OutStream, msg
print >> OutStream, 'Signals required:'
for s in required_signals:
print >> OutStream, ' ' + s
print >> OutStream, 'Signals contained:'
for s in signals.keys():
print >> OutStream, ' ' + s
raise IOError(msg)
# resample to constant-increment (1 Hz) with zeros at missing samples
time_idx = signals['time'].astype('int')
power_vi = signals['power']
def zone_detect(FitFilePath, ConfigFile=None, OutStream=sys.stdout):
(FilePath, FitFileName) = os.path.split(FitFilePath)
if ConfigFile is None:
ConfigFile = FindConfigFile('', FilePath)
if (ConfigFile is None) or (not os.path.exists(ConfigFile)):
raise IOError('Configuration file not specified or found')
#
# Parse the configuration file
#
from ConfigParser import ConfigParser
config = ConfigParser()
config.read(ConfigFile)
print >> OutStream, 'reading config file ' + ConfigFile
ThresholdPower = config.getfloat('power', 'ThresholdPower')
ThresholdHR = config.getfloat('power', 'ThresholdHR')
print >> OutStream, 'ThresholdPower: ', ThresholdPower
print >> OutStream, 'ThresholdHR : ', ThresholdHR
from datetime import datetime
from fitparse import Activity
from activity_tools import extract_activity_signals
required_signals = ['power'] # 'heart_rate' optional
# get the signals
activity = Activity(FitFilePath)
signals = extract_activity_signals(activity)
if not all(s in signals.keys() for s in required_signals):
msg = 'required signals not in file'
print >> OutStream, msg
print >> OutStream, 'Signals required:'
for s in required_signals:
print >> OutStream, ' ' + s
print >> OutStream, 'Signals contained:'
for s in signals.keys():
print >> OutStream, ' ' + s
raise IOError(msg)
hasHR = True if 'heart_rate' in signals.keys() else False
# up-sample by 5x so that zone-skipping is not needed
SampleRate = 5.0
from numpy import arange, interp
n = len(signals['power'])
old_time = arange(n)
nPts = int(n * SampleRate) # 32-bit integer
new_time = arange(nPts) / SampleRate
power = interp(new_time, old_time, signals['power'])
if hasHR:
heart_rate = interp(new_time, old_time, signals['heart_rate'])
# power zones from "Cyclist's Training Bible", 5th ed., by Joe Friel, p51
FTP = ThresholdPower
pZones = {
1: [0, 0.55 * FTP],
2: [0.55 * FTP, 0.75 * FTP],
3: [0.75 * FTP, 0.90 * FTP],
4: [0.90 * FTP, 1.05 * FTP],
5: [1.05 * FTP, 1.20 * FTP],
6: [1.20 * FTP, 1.50 * FTP],
7: [1.50 * FTP, 2.50 * FTP]
}
# heart-rate zones from "Cyclist's Training Bible" 5th ed. by Joe Friel, p50
FTHR = ThresholdHR
hZones = {
1: [0, 0.82 * FTHR], # 1
2: [0.82 * FTHR, 0.89 * FTHR], # 2
3: [0.89 * FTHR, 0.94 * FTHR], # 3
4: [0.94 * FTHR, 1.00 * FTHR], # 4
5: [1.00 * FTHR, 1.03 * FTHR], # 5a
6: [1.03 * FTHR, 1.06 * FTHR], # 5b
7: [1.07 * FTHR, 1.15 * FTHR]
} # 5c
def LocateZone(x, zones):
Z = 1
if x >= zones[2][0]: Z = 2
if x >= zones[3][0]: Z = 3
if x >= zones[4][0]: Z = 4
if x >= zones[5][0]: Z = 5
if x >= zones[6][0]: Z = 6
if x >= zones[7][0]: Z = 7
return Z
# define boxcar averages used to test for upward transition out of
# indicated zone.
fpZ1 = ForwardBoxcarAverage(power, window=90, SampleRate=SampleRate)
fpZ2 = ForwardBoxcarAverage(power, window=60, SampleRate=SampleRate)
fpZ3 = ForwardBoxcarAverage(power, window=45, SampleRate=SampleRate)
fpZ4 = ForwardBoxcarAverage(power, window=30, SampleRate=SampleRate)
fpZ5 = ForwardBoxcarAverage(power, window=15, SampleRate=SampleRate)
fpZ6 = ForwardBoxcarAverage(power, window=5, SampleRate=SampleRate)
# fpZ7 not needed
# assemble these into a dictionary:
# so that I could test
# if LocateZone( FBoxCars[CurrentZone][i], pZones )
# > CurrentZone:
FBoxCars = {
1: fpZ1,
2: fpZ2,
3: fpZ3,
4: fpZ4,
5: fpZ5,
6: fpZ6
} # Z7 not needed
# define boxcar averages used to test for downward transition into
# indicated zone.
cpZ1 = CenteredBoxcarAverage(power, window=60, SampleRate=SampleRate)
cpZ2 = CenteredBoxcarAverage(power, window=45, SampleRate=SampleRate)
cpZ3 = CenteredBoxcarAverage(power, window=30, SampleRate=SampleRate)
cpZ4 = CenteredBoxcarAverage(power, window=15, SampleRate=SampleRate)
cpZ5 = CenteredBoxcarAverage(power, window=7, SampleRate=SampleRate)
cpZ6 = CenteredBoxcarAverage(power, window=3, SampleRate=SampleRate)
# fpZ7 not needed
# assemble these into a dictionary:
# so that I could test
# if LocateZone( CBoxCars[CurrentZone][i], pZones )
# > CurrentZone:
CBoxCars = {
1: cpZ1,
2: cpZ2,
3: cpZ3,
4: cpZ4,
5: cpZ5,
6: cpZ6
} # Z7 not needed
from numpy import array, arange, append, zeros, cumsum, average
cp2 = zeros(nPts)
fboxpower = zeros(nPts)
cboxpower = zeros(nPts)
zone = zeros(nPts)
zone_mid = zeros(nPts)
CurrentZone = 1
# create a phaseless, lowpass-filtered signal for downward transitions
# see
# https://docs.scipy.org/doc/scipy/reference/signal.html
from scipy import signal
poles = 4
cutoff = 0.1 # Hz
Wn = cutoff / (SampleRate / 2)
PadLen = int(SampleRate / cutoff)
b, a = signal.butter(poles, Wn, btype='lowpass')
# lpfpower = signal.filtfilt(b, a, power, padlen=PadLen)
# calculate zone midpoints for plotting
ZoneMidPoint = {} # empty dictionary
ZoneMidPoint[1] = (pZones[1][0] + pZones[1][1]) / 2
ZoneMidPoint[2] = (pZones[2][0] + pZones[2][1]) / 2
ZoneMidPoint[3] = (pZones[3][0] + pZones[3][1]) / 2
ZoneMidPoint[4] = (pZones[4][0] + pZones[4][1]) / 2
ZoneMidPoint[5] = (pZones[5][0] + pZones[5][1]) / 2
ZoneMidPoint[6] = (pZones[6][0] + pZones[6][1]) / 2
ZoneMidPoint[7] = (pZones[7][0] + pZones[7][1]) / 2
for i, p in zip(range(nPts), power):
# compute the centered 3-second power
#raise RuntimeError("need to account for sample rate in cp2")
sr = int(SampleRate)
if i == 0:
cp2[i] = power[i]
elif i < 2 * sr:
cp2[i] = average(power[0:i])
elif i > nPts - 2 * sr:
cp2[i] = average(power[i - sr:])
else:
cp2[i] = average(power[i - sr:i + sr + 1])
# upward transition
cz = CurrentZone # short name
if cz < 7:
tz = 7 # Test Zone
while tz > cz:
if (LocateZone( cp2[i], pZones ) >= tz) \
& (LocateZone( FBoxCars[tz-1][i], pZones ) >= tz):
CurrentZone = tz
zone[i] = CurrentZone
break
tz -= 1
# downward transition. Avoid 2nd test if in Z1.
# use centered-boxcar average to avoid getting "trapped"
# in low zones.
if cz > 1:
tz = cz - 1
while tz >= 1:
if (LocateZone( cp2[i], pZones ) <= tz) \
& (LocateZone( CBoxCars[tz][i], pZones ) <= tz) \
& (LocateZone( FBoxCars[tz][i], pZones ) <= tz):
CurrentZone = tz
zone[i] = CurrentZone
break
tz -= 1
# the filtered power comes from CurrentZone after any transition
fboxpower[i] = FBoxCars[min(CurrentZone, 6)][i]
cboxpower[i] = CBoxCars[max(CurrentZone - 1, 1)][i]
# calculate zone midpoints for plotting
zone_mid[i] = ZoneMidPoint[CurrentZone]
# get zone bounds for plotting
p_zone_bounds = [
pZones[1][0], pZones[2][0], pZones[3][0], pZones[4][0], pZones[5][0],
pZones[6][0], pZones[7][0], pZones[7][1]
]
h_zone_bounds = [
0.4 * FTHR, # better plotting
hZones[2][0],
hZones[3][0],
hZones[4][0],
hZones[5][0],
hZones[6][0],
hZones[7][0],
hZones[7][1]
]
############################################################
# plotting #
############################################################
#
# extract lap times
#
from activity_tools import extract_activity_laps
activity = Activity(FitFilePath)
laps = extract_activity_laps(activity)
lap_start_time = laps['start_time'] # datetime object
lap_timestamp = laps['timestamp']
nLaps = len(lap_start_time)
t0 = signals['metadata']['timestamp']
lap_start_sec = zeros(nLaps) # lap start times in seconds
for i in range(nLaps):
tBeg = (lap_start_time[i] - t0).total_seconds()
tEnd = (lap_timestamp[i] - t0).total_seconds()
lap_start_sec[i] = tBeg
# time plot
import matplotlib.pyplot as plt
import matplotlib.dates as md
from matplotlib.dates import date2num, DateFormatter
import datetime as dt
base = dt.datetime(2014, 1, 27, 0, 0, 0)
x = [base + dt.timedelta(seconds=t) for t in new_time]
x = date2num(x) # Convert to matplotlib format
x_laps = [ base + dt.timedelta(seconds=t) \
for t in lap_start_sec.astype('float') ]
x_laps = date2num(x_laps)
if hasHR:
fig1, (ax0, ax1) = plt.subplots(nrows=2, sharex=True)
ax0.plot_date(x, heart_rate, 'r-', linewidth=3)
ax0.set_yticks(h_zone_bounds, minor=False)
ax0.grid(True)
ax0.set_title('heart rate, BPM')
for i in range(nLaps):
ax0.axvline(x_laps[i], label=str(i + 1))
else:
fig1, ax1 = plt.subplots(nrows=1, sharex=True)
ax1.plot_date(x, power, 'k-', linewidth=1)
ax1.plot_date(x, fboxpower, 'm-', linewidth=1)
ax1.plot_date(x, cp2, 'r.', markersize=4)
ax1.plot_date(x, cboxpower, 'b-', linewidth=1)
ax1.plot_date(x, zone_mid, 'g-', linewidth=3)
ax1.xaxis.set_major_formatter(DateFormatter('%H:%M:%S'))
ax1.set_yticks(p_zone_bounds, minor=False)
ax1.grid(True)
ax1.set_title('power, watts')
ax1.legend(['power', 'FBoxCar', 'cp2', 'CBoxCar', 'zone mid'],
loc='upper left')
for i in range(nLaps):
ax1.axvline(x_laps[i], label=str(i + 1))
fig1.autofmt_xdate()
fig1.suptitle('Power Zone Detection', fontsize=20)
fig1.tight_layout()
fig1.canvas.set_window_title(FitFilePath)
plt.show()
# better histogram plot with control of counts
from numpy import histogram
PowerCounts, PowerBins = histogram(power, bins=p_zone_bounds)
ZoneCounts, ZoneBins = histogram(zone_mid, bins=p_zone_bounds)
fig2, ax = plt.subplots()
bar_width = 0.35
opacity = 0.4
#error_config = {'ecolor': '0.3'}
zone_ints = arange(7) + 1
LogY = True
rects1 = ax.bar(zone_ints,
PowerCounts / SampleRate / 60,
bar_width,
alpha=opacity,
color='b',
log=LogY,
label='raw power')
rects2 = ax.bar(zone_ints + bar_width,
ZoneCounts / SampleRate / 60,
bar_width,
alpha=opacity,
color='r',
log=LogY,
label='detected zone')
ax.set_xlabel('Zone')
ax.set_ylabel('minutes')
ax.set_title('Zone Detection Histogram')
ax.set_xticks(zone_ints + bar_width / 2)
ax.set_xticklabels(('Rec', 'End', 'Tmp', 'Thr', 'VO2', 'An', 'NM'))
ax.legend()
fig2.tight_layout()
fig2.canvas.set_window_title(FitFilePath)
plt.show()
# formatted print of histogram
print >> OutStream, 'Power Zone Histogram:'
for i in range(7):
dur = ZoneCounts[i] / SampleRate
pct = dur / sum(ZoneCounts / SampleRate) * 100
hh = dur // 3600
mm = (dur % 3600) // 60
ss = (dur % 3600) % 60
print >> OutStream, ' Zone %i: %2i:%02i:%02i (%2i%%)' \
% (i+1, hh, mm, ss, pct)
dur = sum(ZoneCounts) / SampleRate
hh = dur // 3600
mm = (dur % 3600) // 60
ss = (dur % 3600) % 60
print >> OutStream, ' total: %2i:%02i:%02i' % (hh, mm, ss)
def ClosePlots():
plt.close('all')
return ClosePlots
# Delay range in seconds.
# for finding the delay in Edge relative to Zwift:
# a positive delay means Edge looks like a delayed version of Zwift.
MinDelay = 40 * 60
MaxDelay = 50 * 60
#EdgeFilePath = r'2017-01-02-16-12-43_edge.fit'
#ZwiftFilePath = r'2017-01-02-16-07-51_zwift.fit'
#EdgeFilePath = r'2017-01-07-10-20-12_edge_quarq.fit'
#ZwiftFilePath = r'2017-01-07-10-15-57_zwift_fluid2.fit'
EdgeActivity = Activity(EdgeFilePath)
ZwiftActivity = Activity(ZwiftFilePath)
EdgeSignals = extract_activity_signals(EdgeActivity)
ZwiftSignals = extract_activity_signals(ZwiftActivity)
# save signals for faster analysis
SignalMap = {'EdgeSignals': EdgeSignals, 'ZwiftSignals': ZwiftSignals}
import pickle
SignalsFile = open('signals.pkl', 'wb')
pickle.dump(SignalMap, SignalsFile)
SignalsFile.close()
from pylab import arange, interp, array, zeros, sqrt, average
#from numpy import array, zeros, arange, interp, sqrt, average
# plot heart rate and calories
edge_hr = EdgeSignals['heart_rate']
edge_t = arange(len(edge_hr))
edge_power = EdgeSignals['power']
def pwhr_transfer_function(FitFilePath, ConfigFile=None, OutStream=sys.stdout):
# this needs to stay INSIDE the function or bad things happen
import matplotlib.pyplot as plt
(FilePath, FitFileName) = os.path.split(FitFilePath)
if ConfigFile is None:
ConfigFile = FindConfigFile('', FilePath)
if (ConfigFile is None) or (not os.path.exists(ConfigFile)):
raise IOError('Configuration file not specified or found')
#
# Parse the configuration file
#
if type(ConfigFile) != type(ConfigParser()):
if (ConfigFile is None) or (not os.path.exists(ConfigFile)):
raise IOError('Configuration file not specified or found')
config = ConfigParser()
config.read(ConfigFile)
print >> OutStream, 'reading config file ' + ConfigFile
else:
config = ConfigFile
WeightEntry = config.getfloat('user', 'weight')
WeightToKg = config.getfloat('user', 'WeightToKg')
weight = WeightEntry * WeightToKg
age = config.getfloat('user', 'age')
EndurancePower = config.getfloat('power', 'EndurancePower')
ThresholdPower = config.getfloat('power', 'ThresholdPower')
EnduranceHR = config.getfloat('power', 'EnduranceHR')
ThresholdHR = config.getfloat('power', 'ThresholdHR')
HRTimeConstant = config.getfloat('power', 'HRTimeConstant')
HRDriftRate = config.getfloat('power', 'HRDriftRate')
print >> OutStream, 'WeightEntry : ', WeightEntry
print >> OutStream, 'WeightToKg : ', WeightToKg
print >> OutStream, 'weight : ', weight
print >> OutStream, 'age : ', age
print >> OutStream, 'EndurancePower : ', EndurancePower
print >> OutStream, 'ThresholdPower : ', ThresholdPower
print >> OutStream, 'EnduranceHR : ', EnduranceHR
print >> OutStream, 'ThresholdHR : ', ThresholdHR
print >> OutStream, 'HRTimeConstant : ', HRTimeConstant
print >> OutStream, 'HRDriftRate : ', HRDriftRate
# power zones from "Cyclist's Training Bible", 5th ed., by Joe Friel, p51
FTP = ThresholdPower
pZones = {
1: [0, 0.55 * FTP],
2: [0.55 * FTP, 0.75 * FTP],
3: [0.75 * FTP, 0.90 * FTP],
4: [0.90 * FTP, 1.05 * FTP],
5: [1.05 * FTP, 1.20 * FTP],
6: [1.20 * FTP, 1.50 * FTP],
7: [1.50 * FTP, 2.50 * FTP]
}
# heart-rate zones from "Cyclist's Training Bible" 5th ed. by Joe Friel, p50
FTHR = ThresholdHR
hZones = {
1: [0, 0.82 * FTHR], # 1
2: [0.82 * FTHR, 0.89 * FTHR], # 2
3: [0.89 * FTHR, 0.94 * FTHR], # 3
4: [0.94 * FTHR, 1.00 * FTHR], # 4
5: [1.00 * FTHR, 1.03 * FTHR], # 5a
6: [1.03 * FTHR, 1.06 * FTHR], # 5b
7: [1.07 * FTHR, 1.15 * FTHR]
} # 5c
# get zone bounds for plotting
p_zone_bounds = [
pZones[1][0], pZones[2][0], pZones[3][0], pZones[4][0], pZones[5][0],
pZones[6][0], pZones[7][0], pZones[7][1]
]
h_zone_bounds = [
0.4 * FTHR, # better plotting
hZones[2][0],
hZones[3][0],
hZones[4][0],
hZones[5][0],
hZones[6][0],
hZones[7][0],
hZones[7][1]
]
from fitparse import Activity
from activity_tools import extract_activity_signals
required_signals = ['power', 'heart_rate']
# get the signals
activity = Activity(FitFilePath)
signals = extract_activity_signals(activity, resample='existing')
if not all(s in signals.keys() for s in required_signals):
msg = 'required signals not in file'
print >> OutStream, msg
print >> OutStream, 'Signals required:'
for s in required_signals:
print >> OutStream, ' ' + s
print >> OutStream, 'Signals contained:'
for s in signals.keys():
print >> OutStream, ' ' + s
raise IOError(msg)
# resample to constant-increment (1 Hz) with zeros at missing samples
time_idx = signals['time'].astype('int')
power_vi = signals['power']
heart_rate_vi = signals['heart_rate']
nScans = time_idx[-1] + 1
time_ci = np.arange(nScans)
power = np.zeros(nScans)
power[time_idx] = power_vi
heart_rate_ci = np.zeros(nScans)
heart_rate_ci[time_idx] = heart_rate_vi
# compute the 30-second, moving-average power signal.
p30 = BackwardMovingAverage(power)
'''
# compute moving time
moving_time would contain data in the gaps--repeats of the "stopped"
time until it resumes:
>>> time_idx
array([ 0, 1, 2, 5, 6, 7, 10, 11, 12])
>>> time_ci
array([ 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12])
>>> moving_time
array([ 0, 1, 2, 2, 2, 3, 4, 5, 5, 5, 6, 7, 8])
'''
moving_time = np.zeros(nScans).astype('int')
idx = 0
for i in time_ci[0:-1]:
moving_time[i] = idx
if time_idx[idx + 1] == i + 1:
idx += 1
moving_time[nScans - 1] = idx
# Calculate running normalized power and TSS.
norm_power = np.zeros(nScans)
TSS = np.zeros(nScans)
for i in range(1, nScans):
ii = np.nonzero(time_idx <= i)[0]
norm_power[i] = np.average(p30[time_idx[ii]]**4)**(0.25)
TSS[i] = moving_time[i] / 36 * (norm_power[i] / FTP)**2
#
# simulate the heart rate
#
SampleRate = 1.0
tau = HRTimeConstant # 63.0 seconds
PwHRTable = np.array([
[0, 0.50 * FTHR], # Active resting HR
[0.55 * FTP, 0.70 * FTHR], # Recovery
[0.70 * FTP, 0.82 * FTHR], # Aerobic threshold
[1.00 * FTP, FTHR], # Functional threshold
[1.20 * FTP, 1.03 * FTHR], # Aerobic capacity
[1.50 * FTP, 1.06 * FTHR]
]) # Max HR
def heartrate_dot(HR, t):
i = min(int(t * SampleRate), nScans - 1)
HRp = np.interp(power[i], PwHRTable[:, 0], PwHRTable[:, 1])
HRt = HRp + HRDriftRate * TSS[i]
return (HRt - HR) / tau
heart_rate_sim = odeint(heartrate_dot, heart_rate_ci[0], time_ci)
err = np.squeeze( heart_rate_sim ) \
- np.squeeze( heart_rate_ci )
RMSError = np.sqrt(np.average(err[time_idx]**2))
print >> OutStream, 'Average measured HR : %7i BPM' \
% np.average(heart_rate_ci[time_idx])
print >> OutStream, 'Average simulated HR : %7i BPM' \
% np.average(heart_rate_sim[time_idx])
print >> OutStream, 'RMS error : %7i BPM' % RMSError
#
# Estimate better values for FTHR and HRDriftRate
#
coef = np.polyfit(TSS[time_idx],
-err[time_idx],
deg=1,
w=heart_rate_ci[time_idx] - 0.50 * FTHR)
slope = coef[0]
offset = coef[1]
NewThresholdHR = offset + ThresholdHR
NewHRDriftRate = slope + HRDriftRate
print >> OutStream, 'Estimated ThresholdHR : %7.1f BPM' \
% NewThresholdHR
print >> OutStream, 'Estimated HRDriftRate : %7.4f BPM/TSS' \
% NewHRDriftRate
# -------------- debug ---------------
print 'coef = ', coef
print 'TSS = ', TSS[-1]
hh = moving_time[-1] // 3600
mm = (moving_time[-1] % 3600) // 60
ss = (moving_time[-1] % 3600) % 60
print 'moving time: %4i:%02i:%02i' % (hh, mm, ss)
print 'normalized power:', norm_power[-1], norm_power.max()
CrossPlotFig = plt.figure()
sc = plt.scatter(TSS[time_idx], -err[time_idx], s=5)
plt.title('Simulation Error Vs TSS')
plt.xlabel('TSS')
plt.ylabel('BPM')
plt.grid(b=True, which='major', axis='both')
a = plt.axis()
#plt.axis([ 0, a[1], 0, a[3] ])
y_fit = slope * TSS[time_idx] + offset
plt.plot(TSS[time_idx], y_fit, 'k-')
plt.show()
#
# extract lap results
#
from activity_tools import extract_activity_laps
activity = Activity(FitFilePath)
laps = extract_activity_laps(activity)
lap_start_time = laps['start_time'] # datetime object
lap_timestamp = laps['timestamp']
nLaps = len(lap_start_time)
t0 = signals['metadata']['timestamp']
lap_start_sec = np.zeros(nLaps) # lap start times in seconds
for i in range(nLaps):
tBeg = (lap_start_time[i] - t0).total_seconds()
tEnd = (lap_timestamp[i] - t0).total_seconds()
lap_start_sec[i] = tBeg
#
# time plot
#
import matplotlib.dates as md
from matplotlib.dates import date2num, DateFormatter
import datetime as dt
base = dt.datetime(2014, 1, 1, 0, 0, 0)
x = [base + dt.timedelta(seconds=t) for t in time_ci.astype('float')]
x = date2num(x) # Convert to matplotlib format
x_laps = [ base + dt.timedelta(seconds=t) \
for t in lap_start_sec.astype('float') ]
x_laps = date2num(x_laps)
fig1, (ax0, ax1) = plt.subplots(nrows=2, sharex=True)
ax0.plot_date(x, heart_rate_ci, 'r-', linewidth=1)
ax0.plot_date(x, heart_rate_sim, 'm-', linewidth=3)
ax0.set_yticks(h_zone_bounds, minor=False)
ax0.grid(True)
ax0.legend(['measured', 'simulated'], loc='upper left')
ax0.set_title('heart rate, BPM')
ax1.plot_date(x, power, 'k-', linewidth=1)
ax1.plot_date(x, p30, 'b-', linewidth=3)
ax1.xaxis.set_major_formatter(DateFormatter('%H:%M:%S'))
ax1.set_yticks(p_zone_bounds, minor=False)
ax1.grid(True)
ax1.set_title('power, watts')
for i in range(nLaps):
ax0.axvline(x_laps[i], label=str(i + 1))
ax1.axvline(x_laps[i], label=str(i + 1))
fig1.autofmt_xdate()
ax1.legend(['power', 'p30'], loc='upper left')
fig1.suptitle('Pw:HR Transfer Function', fontsize=20)
fig1.tight_layout()
fig1.canvas.set_window_title(FitFilePath)
fig1.subplots_adjust(hspace=0) # Remove horizontal space between axes
plt.show()
def ClosePlots():
plt.close('all')
return ClosePlots
def saddle_endurance_anls(FitFilePath, ConfigFile=None, OutStream=sys.stdout):
(FilePath, FitFileName) = os.path.split(FitFilePath)
# no config file needed
from datetime import datetime
from fitparse import Activity
from activity_tools import extract_activity_signals
import numpy as np
required_signals = ['power', 'cadence']
# get the signals
activity = Activity(FitFilePath)
signals = extract_activity_signals(activity)
if not all(s in signals.keys() for s in required_signals):
msg = 'required signals not in file'
print >> OutStream, msg
print >> OutStream, 'Signals required:'
for s in required_signals:
print >> OutStream, ' ' + s
print >> OutStream, 'Signals contained:'
for s in signals.keys():
print >> OutStream, ' ' + s
raise IOError(msg)
SampleRate = 1.0
cadence = signals['cadence']
elapsed_time = signals['time']
nPts = len(cadence)
cad_fwd_min_3_8 = FwdRunningMinimum(cadence, wBeg=3, wEnd=8)
cad_fwd_min_1_8 = FwdRunningMinimum(cadence, wBeg=1, wEnd=8)
STANDING = 0
SEATED = 1
state = STANDING
CThr = 40.0
seated_state = np.zeros(nPts)
for i in range(nPts):
if state == STANDING:
if cadence[i] >= CThr and cad_fwd_min_1_8[i] >= CThr:
state = SEATED
else: # state==SEATED:
if cadence[i] < CThr and cad_fwd_min_3_8[i] < CThr:
state = STANDING
seated_state[i] = state
#
# Determine seated segment durations
#
iiUp = np.nonzero(seated_state[1:] - seated_state[0:-1] == 1)[0]
iiDn = np.nonzero(seated_state[1:] - seated_state[0:-1] == -1)[0]
if iiUp[-1] < iiDn[-1]:
seg_durtns = np.zeros(len(iiDn))
seg_starts = np.zeros(len(iiDn)).astype('int')
seg_stops = np.zeros(len(iiDn)).astype('int')
else:
seg_durtns = np.zeros(len(iiDn) + 1)
seg_starts = np.zeros(len(iiDn) + 1).astype('int')
seg_stops = np.zeros(len(iiDn) + 1).astype('int')
if iiDn[0] < iiUp[0]: # begins seated
seg_durtns[0] = iiDn[0]
seg_starts[0] = 0
seg_stops[0] = iiDn[0]
if iiUp[-1] > iiDn[-1]: # Ends seated
seg_durtns[1:-1] = iiDn[1:] - iiUp[0:-1]
seg_starts[1:-1] = iiUp[0:-1]
seg_stops[1:-1] = iiDn[1:]
seg_durtns[-1] = nPts - iiUp[-1]
seg_starts[-1] = iiUp[-1]
seg_stops[-1] = nPts
else: # ends standing
seg_durtns[1:] = iiDn[1:] - iiUp
seg_starts[1:] = iiUp
seg_stops[1:] = iiDn[1:]
elif iiUp[0] < iiDn[0]: # begins standing
if iiUp[-1] > iiDn[-1]: # ends seated
seg_durtns[0:-1] = iiDn - iiUp[0:-1]
seg_starts[0:-1] = iiUp[0:-1]
seg_stops[0:-1] = iiDn
seg_durtns[-1] = nPts - iiUp[-1]
seg_starts[-1] = iiUp[-1]
seg_stops[-1] = nPts
else: # ends standing
seg_durtns = iiDn - iiUp
seg_starts = iiUp
seg_stops = iiDn
else:
raise RuntimeError("shouldn't be able to reach this code.")
#
# Formatted print of results for all segments
#
# overall results
dur = nPts / SampleRate
hh = dur // 3600
mm = (dur % 3600) // 60
ss = (dur % 3600) % 60
print >> OutStream, 'total time : %2i:%02i:%02i' % (hh, mm, ss)
iSeat = np.nonzero(seated_state == 1)[0]
pct = len(iSeat) / float(nPts) * 100.0
dur = len(iSeat) / SampleRate
hh = dur // 3600
mm = (dur % 3600) // 60
ss = (dur % 3600) % 60
print >> OutStream, 'seated time : %2i:%02i:%02i (%2i%%))' % (hh, mm,
ss, pct)
iStnd = np.nonzero(seated_state == 0)[0]
pct = len(iStnd) / float(nPts) * 100.0
dur = len(iStnd) / SampleRate
hh = dur // 3600
mm = (dur % 3600) // 60
ss = (dur % 3600) % 60
print >> OutStream, 'standing time : %2i:%02i:%02i (%2i%%))' % (hh, mm,
ss, pct)
# segment results
nSeg = len(seg_durtns)
print >> OutStream, 'standing segments:'
names = ['segment', 'start', 'stop', 'duration']
fmt = "%12s" + "%10s" * 3
print >> OutStream, fmt % tuple(names)
for i in range(nSeg):
Beg = elapsed_time[seg_starts[i]]
hhBeg = Beg // 3600
mmBeg = (Beg % 3600) // 60
ssBeg = (Beg % 3600) % 60
End = elapsed_time[seg_stops[i]]
hhEnd = End // 3600
mmEnd = (End % 3600) // 60
ssEnd = (End % 3600) % 60
dur = seg_durtns[i] / SampleRate
hhDur = dur // 3600
mmDur = (dur % 3600) // 60
ssDur = (dur % 3600) % 60
DurPlus = '' if hhDur == 0 else '+%ih' % (hhDur)
fmt = '%12d' + '%4i:%02i:%02i' + '%4i:%02i:%02i' + '%7i:%02i' + '%s'
print >> OutStream, fmt \
% (i, hhBeg, mmBeg, ssBeg,
hhEnd, mmEnd, ssEnd,
mmDur, ssDur, DurPlus)
#
# best hour saddle endurance
#
'''
Find the longest segments that together total one hour,
and compute the average duration to serve as a metric for
the ride.
'''
# time limit: 15 minutes less than end for short rides
TimeLimit = min(3600.0, (nPts / SampleRate - 15 * 60))
# list of indices for longest durations
def GetSegment(i):
return seg_durtns[i]
indx = range(nSeg)
indx.sort(reverse=True, key=GetSegment)
# compute average and print
print >> OutStream, 'best-hour segments:'
names = ['segment', 'start', 'stop', 'duration']
fmt = "%12s" + "%10s" * 3
print >> OutStream, fmt % tuple(names)
TotalTime = 0.0
i = 0
while TotalTime < TimeLimit and i < nSeg:
Beg = elapsed_time[seg_starts[indx[i]]]
hhBeg = Beg // 3600
mmBeg = (Beg % 3600) // 60
ssBeg = (Beg % 3600) % 60
End = elapsed_time[seg_stops[indx[i]]]
hhEnd = End // 3600
mmEnd = (End % 3600) // 60
ssEnd = (End % 3600) % 60
dur = seg_durtns[indx[i]] / SampleRate
hhDur = dur // 3600
mmDur = (dur % 3600) // 60
ssDur = (dur % 3600) % 60
DurPlus = '' if hhDur == 0 else '+%ih' % (hhDur)
fmt = '%12d' + '%4i:%02i:%02i' + '%4i:%02i:%02i' + '%7i:%02i' + '%s'
print >> OutStream, fmt \
% (indx[i], hhBeg, mmBeg, ssBeg,
hhEnd, mmEnd, ssEnd,
mmDur, ssDur, DurPlus)
TotalTime += dur
i += 1
BestAve = TotalTime / float(i)
hh = BestAve // 3600
mm = (BestAve % 3600) // 60
ss = (BestAve % 3600) % 60
DurPlus = '' if hh == 0 else '+%ih' % (hh)
print >> OutStream, ' BEST ONE-HOUR AVERAGE: %7i:%02i%s' \
% (mm, ss, DurPlus)
############################################################
# plotting #
############################################################
#
# extract lap times
#
from activity_tools import extract_activity_laps
activity = Activity(FitFilePath)
laps = extract_activity_laps(activity)
lap_start_time = laps['start_time'] # datetime object
lap_timestamp = laps['timestamp']
nLaps = len(lap_start_time)
t0 = signals['metadata']['timestamp']
lap_start_sec = np.zeros(nLaps) # lap start times in seconds
for i in range(nLaps):
tBeg = (lap_start_time[i] - t0).total_seconds()
tEnd = (lap_timestamp[i] - t0).total_seconds()
lap_start_sec[i] = tBeg
# time plot
import matplotlib.pyplot as plt
import matplotlib.dates as md
from matplotlib.dates import date2num, DateFormatter
import datetime as dt
base = dt.datetime(2014, 1, 27, 0, 0, 0)
x = [base + dt.timedelta(seconds=t) for t in elapsed_time]
x = date2num(x) # Convert to matplotlib format
x_laps = [ base + dt.timedelta(seconds=t) \
for t in lap_start_sec.astype('float') ]
x_laps = date2num(x_laps)
fig, (ax0, ax1, ax2) = plt.subplots(nrows=3, sharex=True)
ax0.plot_date(x, signals['power'], 'b-', linewidth=1)
ax0.grid(True)
ax0.set_ylabel('power, W')
ax0.set_title('Saddle Endurance')
ax1.plot_date(x, signals['cadence'], 'g-', linewidth=1)
ax1.plot_date(x, cad_fwd_min_3_8, 'm-', linewidth=1)
ax1.plot_date(x, cad_fwd_min_1_8, 'brown', linestyle='-', linewidth=1)
ax1.grid(True)
ax1.set_ylabel('cadence, RPM')
ax1.legend(['cadence', 'cad_fwd_min_3_8', 'cad_fwd_min_1_8'],
loc='upper left')
ax2.plot_date(x, seated_state, 'r-', linewidth=3)
ax2.xaxis.set_major_formatter(DateFormatter('%H:%M:%S'))
ax2.grid(True)
ax2.set_ylabel('seated')
ax2.set_yticks([0, 1])
ax2.set_yticklabels(('standing', 'seated'))
for i in range(nLaps):
ax0.axvline(x_laps[i], label=str(i + 1))
ax1.axvline(x_laps[i], label=str(i + 1))
ax2.axvline(x_laps[i], label=str(i + 1))
fig.canvas.set_window_title(FitFilePath)
fig.tight_layout()
fig.subplots_adjust(hspace=0) # Remove horizontal space between axes
plt.show()
def ClosePlots():
plt.close('all')
return ClosePlots
def force_analysis(FitFilePath, ConfigFile=None, OutStream=sys.stdout):
verbose = False
(FilePath, FitFileName) = os.path.split(FitFilePath)
if ConfigFile is None:
ConfigFile = FindConfigFile('', FilePath)
if (ConfigFile is None) or (not os.path.exists(ConfigFile)):
raise IOError('Configuration file not specified or found')
#
# Parse the configuration file
#
from ConfigParser import ConfigParser
config = ConfigParser()
config.read(ConfigFile)
print >> OutStream, 'reading config file ' + ConfigFile
CrankRadius = config.getfloat( 'power', 'CrankRadius' ) \
/ 25.4 # mm -> inches
print >> OutStream, 'CrankRadius: ', CrankRadius, ' inches'
from datetime import datetime
from fitparse import Activity
from activity_tools import extract_activity_signals
required_signals = ['power', 'cadence']
# get the signals
activity = Activity(FitFilePath)
signals = extract_activity_signals(activity)
if not all(s in signals.keys() for s in required_signals):
msg = 'required signals not in file'
print >> OutStream, msg
print >> OutStream, 'Signals required:'
for s in required_signals:
print >> OutStream, ' ' + s
print >> OutStream, 'Signals contained:'
for s in signals.keys():
print >> OutStream, ' ' + s
raise IOError(msg)
SampleRate = 1.0 # Hz
time_signal = signals['time']
power = signals['power']
cadence = signals['cadence']
nPts = len(power)
# Calculate leg force and reps
from math import pi
torque = np.zeros(nPts)
ii = cadence.nonzero()[0]
torque[ii] = power[ii] / cadence[ii] / (2*pi/60) \
* 8.8507 # N*m -> in*lb
leg_force = torque / CrankRadius
'''
Reference from squat:
At the end of the Anatomic Adaptation
phase, I could squat 160lbx4x25. With 75% of my body weight
(190 lbs), the total leg force was 300 lbs, or 150 lbs per leg.
This went through a depth of 16 inches. So each leg performed
work of 100 reps times 150 lbs times 16 inches, or
240,000 in*lb, which is 27.117 kJ at 150 lb.
'''
SquatLegForce = (160 + 0.75 * 190) / 2 # lb
SquatDepth = 16.0 # inches
SquatReps = 100.0
SquatWork = SquatLegForce * SquatDepth * SquatReps \
* 0.000112985416 # in*lb -> kJ
MaxForce = max(SquatLegForce, max(leg_force))
# Histogram bin edges. The first bin is underflow (includes data
# down to the minimum regardless of the first edge), and the last
# is overflow (includes data up to maximum regardless of the last edge).
force_bins = np.arange(0, 82, 2) # (0, 85, 5)
cadence_bins = np.arange(26, 126, 2) # (25, 125, 5)
# compute the force-work histogram
# Instead of counts, we have revs and work as a function of force.
# So we have to perform the histogram manually.
nFBins = len(force_bins) - 1
force_width = (force_bins[1:nFBins + 1] - force_bins[0:nFBins]) * 0.95
revs = np.zeros(nFBins)
force_work = np.zeros(nFBins)
for i in range(nFBins):
FBinLo = force_bins[i] if i > 1 else leg_force.min()
FBinHi = force_bins[i + 1] if i < nFBins else leg_force.max() * 1.1
ii = np.nonzero(np.logical_and(leg_force >= FBinLo,
leg_force < FBinHi))[0]
# Compute the revs and work at these indices:
revs[i] = sum(cadence[ii]) / 60 / SampleRate
force_work[i] = sum(power[ii]) / SampleRate \
/ 1000.0 # J -> kJ
# compute the cadence-work histogram
nCBins = len(cadence_bins) - 1
cadence_width = (cadence_bins[1:nCBins + 1] -
cadence_bins[0:nCBins]) * 0.95
cadence_work = np.zeros(nCBins)
for i in range(nCBins):
CBinLo = cadence_bins[i] if i > 1 else leg_force.min()
CBinHi = cadence_bins[i + 1] if i < nCBins else cadence.max() * 1.1
ii = np.nonzero(np.logical_and(cadence >= CBinLo, cadence < CBinHi))[0]
cadence_work[i] = sum(power[ii]) / SampleRate \
/ 1000.0 # J -> kJ
#
# Exposure Histogram
#
# I want the plot to display force on the X axis, but this is
# the column (2nd) index, and the Y axis is the row (1st) index.
# So I need to transpose work2d at some point; it would probably
# be best to do so right up front since np.meshgrid() formats
# the coordinates this way.
work2d = np.zeros([nCBins, nFBins]) # note shape!
power_grid = np.zeros([nCBins + 1, nFBins + 1])
force_grid, cadence_grid = np.meshgrid(force_bins, cadence_bins)
for i in range(nFBins):
FBinLo = force_bins[i] if i > 1 else leg_force.min()
FBinHi = force_bins[i + 1] if i < nFBins else leg_force.max() * 1.1
ii = np.nonzero(np.logical_and(leg_force >= FBinLo,
leg_force < FBinHi))[0]
for j in range(nCBins):
CBinLo = cadence_bins[j] if j > 1 else cadence.min()
CBinHi = cadence_bins[j + 1] if j < nCBins else cadence.max() * 1.1
jj = np.nonzero(
np.logical_and(cadence[ii] >= CBinLo, cadence[ii] < CBinHi))[0]
work2d[j,i] = sum(power[ii[jj]]) \
/ SampleRate / 1000.0 # J -> kJ
TorqueNM = force_bins[i] * CrankRadius / 8.8507
power_grid[j, i] = TorqueNM * cadence_bins[j] * (2 * pi / 60)
# last grid column
i += 1
TorqueNM = force_bins[i] * CrankRadius / 8.8507
for j in range(nCBins + 1):
power_grid[j, i] = TorqueNM * cadence_bins[j] * (2 * pi / 60)
# last grid row. Leave j = nCBins
for i in range(nFBins + 1):
TorqueNM = force_bins[i] * CrankRadius / 8.8507
power_grid[j, i] = TorqueNM * cadence_bins[j] * (2 * pi / 60)
power_lines = np.arange(50, power_grid.max(), 50)
###########################################################
### plotting ###
###########################################################
# this needs to stay INSIDE the function or bad things happen
import matplotlib.pyplot as plt
# force & cadence time plot with grids at valid bin edges so that
# underflow and overflow bins are shown accurately.
import matplotlib.dates as md
from matplotlib.dates import date2num, DateFormatter
import datetime as dt
base = dt.datetime(2014, 1, 1, 0, 0, 0)
x = [base + dt.timedelta(seconds=t) for t in time_signal.astype('float')]
x = date2num(x) # Convert to matplotlib format
fig1, (ax0, ax1) = plt.subplots(nrows=2, sharex=True)
ax0.plot_date(x, leg_force, 'r-+', linewidth=1)
ax0.grid(True)
ax0.legend(['leg_force'], loc='upper left')
ax0.set_title('Leg Force, lb')
ax0.set_yticks(force_bins[1:-1], minor=False)
ax1.plot_date(x, cadence, 'g-+', linewidth=1)
ax1.legend(['cadence'], loc='upper left')
ax1.xaxis.set_major_formatter(DateFormatter('%H:%M:%S'))
ax1.grid(True)
ax1.set_title('cadence, RPM')
ax1.set_yticks(cadence_bins[1:-1], minor=False)
fig1.autofmt_xdate()
fig1.suptitle('Force Analysis', fontsize=20)
fig1.tight_layout()
fig1.canvas.set_window_title(FitFilePath)
plt.show()
# force_work histogram plot
fig2, ax2 = plt.subplots()
opacity = 0.8
LogY = False
rects1 = ax2.bar(force_bins[:-1],
force_work,
force_width,
align='edge',
alpha=opacity,
color='r',
log=LogY,
label='force')
ax2.set_xlabel('Pedal Force, lb')
ax2.set_ylabel('work, kJ')
ax2.set_title('Pedal Force Work Histogram')
ax2.legend()
fig2.tight_layout()
fig2.canvas.set_window_title(FitFilePath)
plt.show()
# cadence_work histogram plot
fig4, ax4 = plt.subplots()
opacity = 0.8
rects4 = ax4.barh(cadence_bins[:-1],
cadence_work,
cadence_width,
align='edge',
alpha=opacity,
color='g',
log=False,
label='cadence')
ax4.set_ylabel('cadence, RPM')
ax4.set_xlabel('work, kJ')
ax4.set_title('Cadence-Work Histogram')
ax4.legend()
fig4.tight_layout()
fig4.canvas.set_window_title(FitFilePath)
plt.show()
# create a custom segmented colormap. I want zero to be a cool color
# on which I can see the power contours; I want a quick transition that
# exposes low work values with a subtle cool color; then I want the
# map to gradually transition through "hotter" colors to red.
from matplotlib import colors as mcolors
colors = dict(mcolors.BASE_COLORS, **mcolors.CSS4_COLORS)
segment_data = [
# X color
(0, 'darkviolet'),
(0.005, 'purple'),
(0.1, 'darkgreen'),
(0.3, 'blue'),
(0.5, 'cyan'),
(0.7, 'yellow'),
(1.0, 'red')
]
cdict = {'red': [], 'green': [], 'blue': []}
for x, cName in segment_data:
rgba = mcolors.to_rgba(colors[cName])
cdict['red'].append((x, rgba[0], rgba[0]))
cdict['green'].append((x, rgba[1], rgba[1]))
cdict['blue'].append((x, rgba[2], rgba[2]))
newcmp = mcolors.LinearSegmentedColormap('WillsCmap',
segmentdata=cdict,
N=256)
# exposure histogram plot
# clone from
# https://matplotlib.org/gallery/images_contours_and_fields/pcolor_demo.html
fig3, ax3 = plt.subplots()
c = ax3.pcolor(force_grid,
cadence_grid,
work2d,
cmap=newcmp,
vmin=work2d.min(),
vmax=work2d.max())
ax3.axis([
force_bins.min(),
force_bins.max(),
cadence_bins.min(),
cadence_bins.max()
])
cbar = fig3.colorbar(c, ax=ax3)
cbar.ax.set_ylabel('work, kJ')
CS = ax3.contour(force_grid,
cadence_grid,
power_grid,
power_lines,
colors='k')
ax3.clabel(CS, fontsize=9, inline=1)
ax3.set_xlabel('Pedal Force, lb')
ax3.set_ylabel('cadence, RPM')
ax3.set_title('Pedal Exposure Histogram')
fig3.tight_layout()
fig3.canvas.set_window_title(FitFilePath)
plt.show()
def ClosePlots():
plt.close('all')
return ClosePlots
print 'weight : ', weight print 'age : ', age print 'EndurancePower: ', EndurancePower print 'ThresholdPower: ', ThresholdPower print 'EnduranceHR : ', EnduranceHR print 'ThresholdHR : ', ThresholdHR from datetime import datetime from fitparse import Activity from activity_tools import extract_activity_signals # typ 'S:\\will\\documents\\bike\\fitfiles\\2017-01-08-15-41-48_zwift.fit' fitfilepath = sys.argv[1] activity = Activity(fitfilepath) signals = extract_activity_signals(activity) ######################################################################## ### Compute Calories ### ######################################################################## from numpy import array, arange, append, zeros, cumsum, average #from pylab import * # Formula widely available. One site: # https://www.easycalculation.com/formulas/heart-rate-calorie-burn.html #weight = 188.0*0.45359237 # lb->kg #age = 50.0 # calibration at endurance
def plot_heartrate(FitFilePath, ConfigFile=None, OutStream=sys.stdout):
verbose = False
(FilePath, FitFileName) = os.path.split(FitFilePath)
if ConfigFile is None:
ConfigFile = FindConfigFile('', FilePath)
if (ConfigFile is None) or (not os.path.exists(ConfigFile)):
raise IOError('Configuration file not specified or found')
#
# Parse the configuration file
#
from ConfigParser import ConfigParser
config = ConfigParser()
config.read(ConfigFile)
print >> OutStream, 'reading config file ' + ConfigFile
WeightEntry = config.getfloat('user', 'weight')
WeightToKg = config.getfloat('user', 'WeightToKg')
weight = WeightEntry * WeightToKg
age = config.getfloat('user', 'age')
sex = config.get('user', 'sex')
EndurancePower = config.getfloat('power', 'EndurancePower')
ThresholdPower = config.getfloat('power', 'ThresholdPower')
EnduranceHR = config.getfloat('power', 'EnduranceHR')
ThresholdHR = config.getfloat('power', 'ThresholdHR')
HRTimeConstant = config.getfloat('power', 'HRTimeConstant')
HRDriftRate = config.getfloat('power', 'HRDriftRate')
print >> OutStream, 'WeightEntry : ', WeightEntry
print >> OutStream, 'WeightToKg : ', WeightToKg
print >> OutStream, 'weight : ', weight
print >> OutStream, 'age : ', age
print >> OutStream, 'sex : ', sex
print >> OutStream, 'EndurancePower: ', EndurancePower
print >> OutStream, 'ThresholdPower: ', ThresholdPower
print >> OutStream, 'EnduranceHR : ', EnduranceHR
print >> OutStream, 'ThresholdHR : ', ThresholdHR
print >> OutStream, 'HRTimeConstant : ', HRTimeConstant
print >> OutStream, 'HRDriftRate : ', HRDriftRate
from datetime import datetime
from fitparse import Activity
from activity_tools import extract_activity_signals
required_signals = ['heart_rate']
# get the signals
activity = Activity(FitFilePath)
signals = extract_activity_signals(activity)
if not all(s in signals.keys() for s in required_signals):
msg = 'required signals not in file'
print >> OutStream, msg
print >> OutStream, 'Signals required:'
for s in required_signals:
print >> OutStream, ' ' + s
print >> OutStream, 'Signals contained:'
for s in signals.keys():
print >> OutStream, ' ' + s
raise IOError(msg)
time_signal = signals['time']
heartrate_signal = signals['heart_rate']
# plot the heart rate
import numpy as np
########################################################################
### Compute Calories ###
########################################################################
'''
Formula widely available. One site:
https://www.easycalculation.com/formulas/heart-rate-calorie-burn.html
For Male,
Calorie Burned = ( ( -55.0969
+ (0.6309 x HR)
+ (0.1988 x W )
+ (0.2017 x A ) )
/ 4.184) x 60 x T
For Female,
Calorie Burned = ( ( -20.4022
+ (0.4472 x HR)
+ (0.1263 x W )
+ (0.0740 x A ) )
/ 4.184) x 60 x T
Where,
HR = Heart Rate
W = Weight in kilograms
A = Age
T = Exercise duration time in hours
However, the formula is adapted to the user's power capacity:
: At EnduranceHR, CalPerMin is set to EndurancePower
(assuming efficiency of 1/4.184 so that calories burned
equal kJ expended).
: At ThresholdHR, CalPerMin is set to ThresholdPower.
: Between EnduranceHR and ThresholdHR, CalPerMin is
interpolated.
: Below EnduranceHR, CalPerMin follows the formula, but
it is scaled onto [EnduranceHR, EndurancePower].
: Above ThresholdHR, CalPerMin follows the formula, but
it is scaled onto [ThresholdHR, ThresholdPower].
'''
hr_sig = signals['heart_rate']
t_sig = signals['time']
dt_sig = np.append(np.array([1.0]), t_sig[1:] - t_sig[0:-1])
nPts = t_sig.size
calories = np.zeros(nPts)
EnduranceBurn = EndurancePower * 3600 / 1e3 / 60 # Cal/min
print >> OutStream, 'EnduranceBurn = %5.2f cal/min' % EnduranceBurn
ThresholdBurn = ThresholdPower * 3600 / 1e3 / 60 # Cal/min
print >> OutStream, 'ThresholdBurn = %5.2f cal/min' % ThresholdBurn
if sex == 'male':
# calibration at endurance
EnduranceCoef = EnduranceBurn \
/ ( -55.0969 \
+ 0.6309*EnduranceHR \
+ 0.1988*weight \
+ 0.2017*age) \
* 4.184
# calibration at threshold
ThresholdCoef = ThresholdBurn \
/ ( -55.0969 \
+ 0.6309*EnduranceHR \
+ 0.1988*weight \
+ 0.2017*age) \
* 4.184
else: # female
# calibration at endurance
EnduranceCoef = EnduranceBurn \
/ ( -20.4022 \
+ 0.4472*EnduranceHR \
+ 0.1263*weight \
+ 0.0740*age) \
* 4.184
# calibration at threshold
ThresholdCoef = ThresholdBurn \
/ ( -20.4022 \
+ 0.4472*EnduranceHR \
+ 0.1263*weight \
+ 0.0740*age) \
* 4.184
for i, dt, HR in zip(range(nPts), dt_sig, hr_sig):
# calories per minute
if HR >= EnduranceHR and HR <= ThresholdHR:
CalPerMin = EnduranceBurn \
+ (HR-EnduranceHR) \
* (ThresholdBurn-EnduranceBurn) \
/ (ThresholdHR-EnduranceHR)
else:
coef = EnduranceCoef if (HR < EnduranceHR) else ThresholdCoef
if sex == 'male':
CalPerMin = ( -55.0969 \
+ 0.6309*HR \
+ 0.1988*weight \
+ 0.2017*age) \
/ 4.184 \
* coef
else:
CalPerMin = ( -20.4022 \
+ 0.4472*HR \
+ 0.1263*weight \
+ 0.0740*age) \
/ 4.184 \
* coef
calories[i] = dt * CalPerMin / 60
running_calories = np.cumsum(calories)
print >> OutStream, 'total calories = %i' % running_calories[nPts - 1]
########################################################################
### Zone Histogram ###
########################################################################
# heart-rate zones from "Cyclist's Training Bible" 5th ed. by Joe Friel, p50
FTHR = ThresholdHR
hZones = {
1: ([0, 0.82 * FTHR], ' 1'),
2: ([0.82 * FTHR, 0.89 * FTHR], ' 2'),
3: ([0.89 * FTHR, 0.94 * FTHR], ' 3'),
4: ([0.94 * FTHR, 1.00 * FTHR], ' 4'),
5: ([1.00 * FTHR, 1.03 * FTHR], '5a'),
6: ([1.03 * FTHR, 1.06 * FTHR], '5b'),
7: ([1.07 * FTHR, 1.15 * FTHR], '5c')
}
h_zone_bounds = [
0.4 * FTHR, # 1 lo
hZones[2][0][0], # 2 lo
hZones[3][0][0], # 3 lo
hZones[4][0][0], # 4 lo
hZones[5][0][0], # 5a lo
hZones[6][0][0], # 5b lo
hZones[7][0][0], # 5c lo
hZones[7][0][1]
] # 5c hi
h_zone_labels = [hZones[k][1] for k in range(1, 8)]
ZoneCounts, ZoneBins = np.histogram(hr_sig, bins=h_zone_bounds)
# formatted print of histogram
SampleRate = 1.0
print >> OutStream, 'Heart-Rate Zone Histogram:'
for i in range(7):
dur = ZoneCounts[i] / SampleRate
pct = dur / sum(ZoneCounts / SampleRate) * 100
hh = dur // 3600
mm = (dur % 3600) // 60
ss = (dur % 3600) % 60
print >> OutStream, ' Zone %2s: %2i:%02i:%02i (%2i%%)' \
% ( h_zone_labels[i], hh, mm, ss, pct)
dur = sum(ZoneCounts) / SampleRate
hh = dur // 3600
mm = (dur % 3600) // 60
ss = (dur % 3600) % 60
print >> OutStream, ' total: %2i:%02i:%02i' % (hh, mm, ss)
########################################################################
### Power & TSS Estimation ###
########################################################################
from endurance_summary import BackwardMovingAverage
# see
# https://docs.scipy.org/doc/scipy/reference/signal.html
from scipy import signal
poles = 3
cutoff = 0.10 # Hz
Wn = cutoff / (SampleRate / 2)
'''
# construct and apply a differentiating, lowpass filter
NumB, DenB = signal.butter(poles, Wn, btype='lowpass',
output='ba', analog=True)
NumF = signal.convolve( NumB, [1,0]) # add differentiator
bDLP, aDLP = signal.bilinear( NumF,DenB, fs=SampleRate )
hr_dot = signal.lfilter(bDLP, aDLP, hr_sig)
'''
# apply a phaseless lowpass filter, then differentiate.
# for some reason, running the Butterworth analog filter
# through bilinear() gives a better result. Otherwise,
# set analog=False to get coefficients directly.
PadLen = int(SampleRate / cutoff) # one period of cutoff
NumB, DenB = signal.butter(poles,
Wn,
btype='lowpass',
output='ba',
analog=True)
bLPF, aLPF = signal.bilinear(NumB, DenB, fs=SampleRate)
hr_lpf = signal.filtfilt(bLPF, aLPF, hr_sig, padlen=PadLen)
hr_dot = np.gradient(hr_lpf, 1 / SampleRate)
FTP = ThresholdPower
FTHR = ThresholdHR
tau = HRTimeConstant
PwHRTable = np.array([
[0, 0.50 * FTHR], # Active resting HR
[0.55 * FTP, 0.70 * FTHR], # Recovery
[0.70 * FTP, 0.82 * FTHR], # Aerobic threshold
[1.00 * FTP, FTHR], # Functional threshold
[1.20 * FTP, 1.03 * FTHR], # Aerobic capacity
[1.50 * FTP, 1.06 * FTHR]
]) # Max HR
# loop through the time series building running power and TSS.
# Notice this is necessary because HRd[i] depends on TSS[i-1].
sPower = np.zeros(nPts)
TSS = np.zeros(nPts)
HRd = np.zeros(nPts) # fatigue drift
HRp = np.zeros(nPts) # power target
p30 = np.zeros(nPts) # 30-sec boxcar average
w = int(30 * SampleRate) # window for boxcar
NPower = np.zeros(nPts) # normalized power
for i in range(1, nPts):
HRd[i] = HRDriftRate * TSS[i - 1]
HRp[i] = hr_sig[i] + tau * hr_dot[i] - HRd[i]
sPower[i] = np.interp(HRp[i], PwHRTable[:, 1], PwHRTable[:, 0])
if i < w:
p30[i] = np.average(sPower[:i + 1]) # include i
else:
p30[i] = np.average(sPower[i - w:i + 1]) # include i
NPower[i] = np.average(p30[:i]**4)**(0.25)
TSS[i] = t_sig[i] / 36 * (NPower[i] / FTP)**2
print >> OutStream, 'estimated NP = %6i W' % NPower[-1]
print >> OutStream, 'estimated work = %6i kJ' % \
( np.cumsum(sPower)[-1] / 1e3 / SampleRate )
print >> OutStream, 'estimated TSS = %6i TSS' % TSS[-1]
if 'power' in signals.keys():
mPower = signals['power']
print >> OutStream, 'measured work = %6i kJ' % \
( np.cumsum( mPower )[-1] / 1e3 / SampleRate )
mP30 = BackwardMovingAverage(mPower)
mNPower = np.average(mP30**4)**(0.25)
mTSS = t_sig[-1] / 36 * (mNPower / FTP)**2
print >> OutStream, 'measured NP = %6i W' % mNPower
print >> OutStream, 'measured TSS = %6i TSS' % mTSS
###########################################################
### plotting ###
###########################################################
# power zones from "Cyclist's Training Bible", 5th ed., by Joe Friel, p51
pZones = {
1: [0, 0.55 * FTP],
2: [0.55 * FTP, 0.75 * FTP],
3: [0.75 * FTP, 0.90 * FTP],
4: [0.90 * FTP, 1.05 * FTP],
5: [1.05 * FTP, 1.20 * FTP],
6: [1.20 * FTP, 1.50 * FTP],
7: [1.50 * FTP, 2.50 * FTP]
}
# heart-rate zones from "Cyclist's Training Bible" 5th ed. by Joe Friel, p50
hZones = {
1: [0, 0.82 * FTHR], # 1
2: [0.82 * FTHR, 0.89 * FTHR], # 2
3: [0.89 * FTHR, 0.94 * FTHR], # 3
4: [0.94 * FTHR, 1.00 * FTHR], # 4
5: [1.00 * FTHR, 1.03 * FTHR], # 5a
6: [1.03 * FTHR, 1.06 * FTHR], # 5b
7: [1.07 * FTHR, 1.15 * FTHR]
} # 5c
# get zone bounds for plotting
p_zone_bounds = [
pZones[1][0], pZones[2][0], pZones[3][0], pZones[4][0], pZones[5][0],
pZones[6][0], pZones[7][0], pZones[7][1]
]
h_zone_bounds = [
0.4 * FTHR, # better plotting
hZones[2][0],
hZones[3][0],
hZones[4][0],
hZones[5][0],
hZones[6][0],
hZones[7][0],
hZones[7][1]
]
# power simulation plot
import matplotlib.pyplot as plt
import matplotlib.dates as md
from matplotlib.dates import date2num, DateFormatter
import datetime as dt
base = dt.datetime(2014, 1, 1, 0, 0, 0)
x = [base + dt.timedelta(seconds=t) for t in t_sig.astype('float')]
x = date2num(x) # Convert to matplotlib format
fig1, (ax0, ax1) = plt.subplots(nrows=2, sharex=True)
ax0.plot_date(x, hr_sig, 'r-', linewidth=2)
ax0.plot_date(x, tau * hr_dot, 'b-', linewidth=2)
ax0.plot_date(x, HRd, 'g-', linewidth=2)
ax0.plot_date(x, HRp, 'k-', linewidth=2)
ax0.set_yticks(h_zone_bounds, minor=False)
ax0.grid(True)
ax0.legend(['HR', 'tau*HRdot', 'HRd', 'HRp'], loc='upper left')
ax0.set_title('heart rate, BPM')
if 'power' in signals.keys():
mPower = signals['power']
ax1.plot_date(x, mPower, 'k-', linewidth=1)
ax1.plot_date(x, sPower, 'b-', linewidth=2)
ax1.legend(['measured power', 'simulated power'], loc='upper left')
else:
ax1.plot_date(x, sPower, 'b-', linewidth=2)
ax1.legend(['simulated power'], loc='upper left')
ax1.xaxis.set_major_formatter(DateFormatter('%H:%M:%S'))
ax1.set_yticks(p_zone_bounds, minor=False)
ax1.grid(True)
ax1.set_title('power, watts')
fig1.autofmt_xdate()
fig1.suptitle('Pw:HR Transfer Function', fontsize=20)
fig1.tight_layout()
fig1.canvas.set_window_title(FitFilePath)
plt.show()
# plot heart rate and calories
import matplotlib.dates as md
from matplotlib.dates import date2num, DateFormatter
import datetime as dt
base = dt.datetime(2014, 1, 27, 0, 0, 0)
x = [base + dt.timedelta(seconds=t) for t in t_sig]
x = date2num(x) # Convert to matplotlib format
fig1, ax0 = plt.subplots()
ax0.plot_date(x, hr_sig, 'r-', linewidth=3)
ax0.set_yticks(h_zone_bounds, minor=False)
ax0.grid(True)
ax0.set_title('heart rate, BPM')
ax0.set_title('Heart Rate Analysis')
fig1.autofmt_xdate()
fig1.tight_layout()
fig1.canvas.set_window_title(FitFilePath)
plt.show()
# plt.plot(t_sig/60, hr_sig, 'r.-')
# plt.title('Heart Rate and Calories')
# plt.ylabel('BPM')
# plt.subplot(2, 1, 2)
# plt.plot(t_sig/60, running_calories, 'b.-')
# plt.xlabel('time (min)')
# plt.ylabel('calories')
# heart rate histogram plot
fig2, ax2 = plt.subplots()
bar_width = 0.80 # 0.35
opacity = 0.4
#error_config = {'ecolor': '0.3'}
zone_ints = np.arange(7) + 1
LogY = False
rects1 = ax2.bar(zone_ints + bar_width / 2,
ZoneCounts / SampleRate / 60,
bar_width,
alpha=opacity,
color='r',
log=LogY,
label='heart rate')
ax2.set_xlabel('Zone')
ax2.set_ylabel('minutes')
ax2.set_title('Heart Rate Zone Histogram')
ax2.set_xticks(zone_ints + bar_width / 2)
ax2.set_xticklabels(h_zone_labels)
ax2.legend()
fig2.tight_layout()
fig2.canvas.set_window_title(FitFilePath)
plt.show()
def ClosePlots():
plt.close('all')
return ClosePlots
def channel_inspect_anls(FitFilePath, ConfigFile=None, OutStream=sys.stdout,
ParentWin=None):
(FilePath, FitFileName) = os.path.split(FitFilePath)
if ConfigFile is None:
ConfigFile = FindConfigFile('', FilePath)
if (ConfigFile is None) or (not os.path.exists(ConfigFile)):
raise IOError('Configuration file not specified or found')
#
# Parse the configuration file
#
from ConfigParser import ConfigParser
config = ConfigParser()
config.read(ConfigFile)
print >> OutStream, 'reading config file ' + ConfigFile
if config.has_section('units'):
# convert list of tuples to dictionary
UserUnits = dict( config.items('units') )
else:
UserUnits = None
# get the signals
from datetime import datetime
from fitparse import Activity
from activity_tools import extract_activity_signals, UnitHandler
activity = Activity(FitFilePath)
signals = extract_activity_signals(activity, resample='existing')
# convert units
if UserUnits is not None:
unithandler = UnitHandler(UserUnits)
signals = unithandler.ConvertSignalUnits(signals)
ChannelList = signals.keys()
ChannelList.remove('time')
ChannelList.remove('metadata')
print >> OutStream, 'Signals contained:'
for s in ChannelList:
print >> OutStream, ' ' + s
if ParentWin is None:
app = wx.App()
dlg = wx.MultiChoiceDialog( ParentWin,
"Pick channels\nto plot",
"channel inspector: " + FitFileName,
ChannelList)
if (dlg.ShowModal() == wx.ID_OK):
selections = dlg.GetSelections()
ChannelNames = [ ChannelList[x] for x in selections ]
print >> OutStream, "Plotting: %s" % (ChannelNames)
dlg.Destroy()
#
# extract lap results
#
from activity_tools import extract_activity_laps
activity = Activity(FitFilePath)
laps = extract_activity_laps(activity)
lap_start_time = laps['start_time'] # datetime object
lap_timestamp = laps['timestamp' ]
nLaps = len(lap_start_time)
t0 = signals['metadata']['timestamp']
lap_start_sec = np.zeros(nLaps) # lap start times in seconds
for i in range(nLaps):
tBeg = (lap_start_time[i] - t0).total_seconds()
tEnd = (lap_timestamp[i] - t0).total_seconds()
lap_start_sec[i] = tBeg
#
# time plot
#
nPlots = len(ChannelNames)
PlotColors = { 'power' : 'm' ,
'heart_rate' : 'r' ,
'cadence' : 'g' ,
'speed' : 'b' ,
'temperature' : 'brown' }
import matplotlib.pyplot as plt
import matplotlib.dates as md
from matplotlib.dates import date2num, DateFormatter
import datetime as dt
base = dt.datetime(2014, 1, 1, 0, 0, 0)
x = [base + dt.timedelta(seconds=t) for t in signals['time'].astype('float')]
x = date2num(x) # Convert to matplotlib format
#fig1, (ax0, ax1) = plt.subplots(nrows=2, sharex=True)
x_laps = [ base + dt.timedelta(seconds=t) \
for t in lap_start_sec.astype('float') ]
x_laps = date2num(x_laps)
axislist = []
fig = plt.figure()
for i, channel in zip( range(nPlots), ChannelNames ):
if PlotColors.has_key(channel):
pcolor = PlotColors[channel]
else:
pcolor = 'k-'
if i > 0:
ax = plt.subplot( nPlots, 1, i+1, sharex=axislist[0] )
else:
ax = plt.subplot( nPlots, 1, i+1 )
ax.plot_date( x, signals[channel], pcolor, linewidth=1, linestyle='-' );
for j in range(nLaps):
ax.axvline( x_laps[j], label=str(j+1) )
ax.grid(True)
if signals['metadata']['units'].has_key(channel):
YLabel = channel + '\n' \
+ signals['metadata']['units'][channel]
else:
YLabel = channel + '\n' + 'none'
ax.set_ylabel(YLabel)
ax.xaxis.set_major_formatter(DateFormatter('%H:%M:%S'))
ax.grid(True)
axislist.append(ax)
fig.autofmt_xdate()
fig.suptitle(FitFileName, fontsize=20)
fig.tight_layout()
fig.canvas.set_window_title(FitFilePath)
fig.subplots_adjust(hspace=0) # Remove horizontal space between axes
plt.show()
def ClosePlots():
plt.close('all')
return ClosePlots