In [1]:

    

from scipy import signal, stats
import numpy as np
import pandas as pd
from matplotlib.pyplot import *

%matplotlib inline


    

In [2]:

    

# the sampling frequency of ganglion is 200 Hz
fs = 200


    

In [3]:

    

d = pd.read_csv('data/smell_1491953550.csv')
d.head()


    

    
Out[3]:






  

    

      

      
timestamps
      
C1
      
C2
      
C3
      
C4
      
tag
    
  
  

    

      
0
      
1.491954e+09
      
11.126202
      
16.109617
      
11.887271
      
13.461769
      
baseline
    
    

      
1
      
1.491954e+09
      
13.285994
      
11.640438
      
13.228025
      
12.272481
      
baseline
    
    

      
2
      
1.491954e+09
      
10.552127
      
13.127048
      
16.128317
      
12.289310
      
baseline
    
    

      
3
      
1.491954e+09
      
12.311749
      
16.928656
      
14.437882
      
13.244855
      
baseline
    
    

      
4
      
1.491954e+09
      
12.068656
      
13.413151
      
15.056836
      
15.520584
      
baseline
    
  

In [4]:

    

## check the sampling rate (should be ~200Hz)
diffs = np.diff(d.timestamps)
diffs = diffs[diffs < 0.1]
1/np.mean(diffs)


    

    
Out[4]:





200.02977421197346

In [5]:

    

X = np.array(d.ix[:, 1:5])
tag = np.array(d.tag)


    

In [6]:

    

freq, y_baseline = signal.welch(X[tag == 'baseline'].T, fs=fs)
freq, y_smell = signal.welch(X[tag == 'smell'].T, fs=fs)

y_baseline = y_baseline.T
y_smell = y_smell.T


    

In [13]:

    

figure(figsize=(14,4))
for i in range(y_baseline.shape[1]):
    _ = plot(freq, np.log(y_baseline[:, i]), label='Channel {}'.format(i+1))
legend()
title('Baseline')
xlim(0,50)
_ = xlabel('Frequency (Hz)')


    

In [14]:

    

figure(figsize=(14,4))
for i in range(y_smell.shape[1]):
    _ = plot(freq, np.log(y_smell[:, i]), label='Channel {}'.format(i+1))
legend()
title('During smell')
xlim(0,50)
_ = xlabel('Frequency (Hz)')


    

In [15]:

    

def moving_average(sig, n=5):
    filt = np.ones(n) / n
    return np.convolve(sig, filt, 'same')


    

In [17]:

    

figure(figsize=(14,4))
diffs = np.log(y_smell) - np.log(y_baseline)
for i in range(diffs.shape[1]):
    avg = moving_average(diffs[:, i], n=3)
    _ = plot(freq, avg, label='Channel {}'.format(i+1))
title('Difference of\nDuring smell vs Baseline')
xlim(0,50)
legend()
_ = xlabel('Frequency (Hz)')


    

In [ ]:

    




    

In [ ]:

    




    

In [ ]: