Little hack scripts to compare heartrate measurements from Polar and Garmin.

heartrate_comparison.py

import numpy as np
import matplotlib.pyplot as plt
import matplotlib.dates as mdates
import pandas as pd
import os

# (1) Garmin Connect: export gpx
# (2) convert gpx with heartrate_comparison_dataprep.sh <file.gpx>
# (3) Polar Flow: export csv

# reads data from garmin
f1=os.path.join(os.sep, os.getenv('HOME'), 'Downloads','garmin_activity_xxx.gpx.csv')
ga = pd.read_csv(f1, header=None, names=['time','hr'], parse_dates=[0])

# reads data from polar
f2=os.path.join(os.sep, os.getenv('HOME'), 'Downloads','polar_username_xxx.csv')
po = pd.read_csv(f2,skiprows=2)

# timestamps from Polar csv are in relative time (starts at 00:00:00) so I add a delta time
# on the basis of the first time sample from the Garmin file
delta=ga.time[0]-po.Time[0]
po['Time_shift'] = po['Time'] + delta

f, ax = plt.subplots(figsize=(8,5))
plt.plot(ga['time'], ga['hr'], '-b', lw=4, alpha=0.5, label='Garmin FR935')
plt.plot(po['Time_shift'], po['HR (bpm)'], '-r', lw=4, alpha=0.5,  label='Polar M400')
plt.legend()
plt.grid(axis='y')
ax.xaxis.set_major_formatter(mdates.DateFormatter("%H:%M"))
_=plt.xticks(rotation=45)

heartrate_comparison_dataprep.sh

#!/bin/sh

GPX=$1
tempfoo=`basename $0`
TMPFILE1=`mktemp /tmp/${tempfoo}.XXXXXX` || exit 1
TMPFILE2=`mktemp /tmp/${tempfoo}.XXXXXX` || exit 1

echo "--> heartrate_comparison_dataprep.sh: input Garmin Connect gpx, output csv with time,hr"
echo "--> temporary files created (and deleted):"
echo "    $TMPFILE1, $TMPFILE2"
echo "--> output: $GPX.csv"

awk -F "[><]" '/time/{print $3}' $GPX  | sed '1d' > $TMPFILE1
awk -F "[><]" '/ns3:hr/{print $3}' $GPX > $TMPFILE2

paste -d ,  $TMPFILE1 $TMPFILE2 > $GPX.csv
rm -f  $TMPFILE1 $TMPFILE2

Little hacks to compare heartrate measurements from Polar M400 (with HR chest strap) and my new Garmin FR935 (with optical heart rate monitor).

What to do:

• run with the two watches
• export gpx from Garmin Connect
• convert this gpx file with the bash script heartrate_comparison_dataprep.sh <file.gpx>
• export csv from Polar Flow
• modify the names and folder of your two input files in the python script
• run the script

This is what you should get.

I know, the accuracy of the optical HR measurement is really surprising!