Analysis of microbit motion sensor data

This Jupyter notebook shows how to analyse BBC micro:bit motion sensor data saved into a csv text file.

I generated this data by fastening one microbit to my kettle as the water heated up then boiled. That microbit ran my logging program microbit-logger.py in 'tx' mode to transmit csv data by radio. The data was received by a second microbit, which was connected to my computer via USB. This microbit ran microbit-logger.py in 'relay' mode. Finally my computer ran usb-receiver.py redirecting output to the text file data-kettle.csv:

$ python usb-receiver.py > data-kettle.csv

The text file can be loaded most easily using pandas.



In [2]:

    
%matplotlib inline
import matplotlib
import matplotlib.pyplot as plt
import pandas as pd



In [7]:

    
df = pd.read_csv('data-kettle.csv')
df['Time'] /= 1000.0    # Convert from ms (int) to sec (float)
df.set_index('Time', inplace=True)
df.info()









    



<class 'pandas.core.frame.DataFrame'>
Float64Index: 5860 entries, 12.966 to 246.132
Data columns (total 8 columns):
AccX       5860 non-null int64
AccY       5860 non-null int64
AccZ       5860 non-null int64
MagX       5860 non-null int64
MagY       5860 non-null int64
MagZ       5860 non-null int64
Heading    5860 non-null int64
Gesture    3760 non-null object
dtypes: int64(7), object(1)
memory usage: 412.0+ KB



In [8]:

    
accXYZ = df[['AccX','AccY','AccZ']]
accXYZ.plot()









    Out[8]:





<matplotlib.axes._subplots.AxesSubplot at 0x7ff00c7cf320>



In [9]:

    
accXYZ = df[['AccX','AccY','AccZ']]
g = accXYZ[50.0:200.0].mean()
g









    Out[9]:





AccX    128.733753
AccY    802.477987
AccZ   -636.926625
dtype: float64



In [10]:

    
s = (accXYZ - g).pow(2).sum(axis=1)



In [11]:

    
s.plot()









    Out[11]:





<matplotlib.axes._subplots.AxesSubplot at 0x7ff00c24f1d0>



In [12]:

    
plt.specgram(s, NFFT=512, Fs=100, noverlap=32)[3]









    Out[12]:





<matplotlib.image.AxesImage at 0x7ff00c15fc88>