forked from Timm877/BCI_Project
-
Notifications
You must be signed in to change notification settings - Fork 0
/
5_realtime_predictions.py
172 lines (140 loc) · 6.16 KB
/
5_realtime_predictions.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
# work with streaming data
# see if I can get it in a pandas DF
# I should collect for 3 sec, then do outlier detection, and design all features, and predict with RF model?
# then every 1 second a prediction should be made about the previous 3 seconds
# if data is 3 sec: do stuff and predict, then remove first second, then again add new data untill 3 sec are full again, and predict once more
# then later try to make it a video game or sumth'ng
# https://github.com/Sentdex/BCI/blob/master/testing_and_making_data.py
# step 1: receive data from streaming
from pythonosc import dispatcher
from pythonosc import osc_server
import random
import pandas as pd
import copy
import math
import matplotlib.pyplot as plt
from matplotlib.animation import FuncAnimation
import threading
import numpy as np
from Step_2_preprocess.OutlierDetection import DistributionBasedOutlierDetection
from Step_3_feature_engineering.Dim_reduction import PrincipalComponentAnalysis, IndependentComponentAnalysis
from Step_3_feature_engineering.TemporalAbstraction import NumericalAbstraction
from Step_3_feature_engineering.FrequencyAbstraction import FourierTransformation
from collections import defaultdict
import datetime as dt
import pickle
#ALL GLOBAL VARIABLES:
hsi = [4,4,4,4]
hsi_string = ""
#Note: The names of the cols from muse device do not align with names in earlier files (Fp1 instead of AF7 and Fp2 instead of AF8)
#Note2: Very important that this list is in the same order as for the earlier datasets as sklearn ML models do not change order!!
cols = ['Delta_TP9','Delta_Fp1','Delta_Fp2','Delta_TP10',
'Theta_TP9','Theta_Fp1','Theta_Fp2','Theta_TP10',
'Alpha_TP9','Alpha_Fp1','Alpha_Fp2','Alpha_TP10',
'Beta_TP9','Beta_Fp1','Beta_Fp2','Beta_TP10',
'Gamma_TP9','Gamma_Fp1','Gamma_Fp2','Gamma_TP10']
Vals = defaultdict(list, { k:[] for k in cols})
#call class instances:
OutlierDistr = DistributionBasedOutlierDetection()
PCA = PrincipalComponentAnalysis()
ICA = IndependentComponentAnalysis()
NumAbs = NumericalAbstraction()
FreqAbs = FourierTransformation()
datapoints = 0
place = 0
rf_model = pickle.load(open("final_random_forest_model_BCI.sav", 'rb'))
def hsi_handler(address: str,*args):
global hsi, hsi_string
hsi = args
if ((args[0]+args[1]+args[2]+args[3])==4):
hsi_string_new = "Muse Fit Good"
else:
hsi_string_new = "Muse Fit Bad on: "
if args[0]!=1:
hsi_string_new += "Left Ear. "
if args[1]!=1:
hsi_string_new += "Left Forehead. "
if args[2]!=1:
hsi_string_new += "Right Forehead. "
if args[3]!=1:
hsi_string_new += "Right Ear."
if hsi_string!=hsi_string_new:
hsi_string = hsi_string_new
print(hsi_string)
def wave_handler(address: str,*args):
global Vals, datapoints, rf_model, cols
wave = args[0][0]
# channel configuration = [TP9, Fp1, Fp2, TP10] as per
# https://web.archive.org/web/20181105231756/http://developer.choosemuse.com/tools/available-data#Absolute_Band_Powers
channels = ['TP9', 'Fp1', 'Fp2', 'TP10']
for i in [0,1,2,3]: # for each of the 4 sensors update the specific brain wave data (delta, theta etc)
key = wave + '_' + channels[i]
Vals[key].append(args[i+1]) #add values to dict
datapoints +=1
# we have 20 features, and we want to have 3 seconds of data, data comes in 10Hz, so we first have to add 20x30=600 datapoints before moving on
if datapoints == 600:
# step 1: create dataframe
# we add datetime to the df as this makes it compatible with our earlier code
df= pd.DataFrame.from_dict(Vals,orient='index').transpose()
df.index = pd.date_range("20180101", periods=df.shape[0], freq='100ms')
# step 2: outlier detection
for col in [c for c in df.columns]:
df = OutlierDistr.mixture_model(df, col, 3)
df.loc[df[f'{col}_mixture'] < 0.0005, col] = np.nan
del df[col + '_mixture']
df[col] = df[col].interpolate()
df[col] = df[col].fillna(method='bfill')
# step 3: pre-process
n_pcs = 4
df = PCA.apply_pca(copy.deepcopy(df), cols, n_pcs)
df = ICA.apply_ica(copy.deepcopy(df), cols)
window_sizes = [10,20,30]
fs = 100
for ws in window_sizes:
df = NumAbs.abstract_numerical(df, cols, ws,
['mean', 'std', 'max', 'min', 'median', 'slope'])
df = FreqAbs.abstract_frequency(df, cols, window_sizes[0], fs)
# now we have exactly 1 row which has no NaN, so choose that row
df.dropna(axis=0, inplace=True)
input = df.to_numpy()
# step 4: predict
pred = rf_model.predict(input)
proba = rf_model.predict_proba(input)
print(pred)
print(proba)
# step 5: update graph
plot_update(pred)
#step 6: now, we reinit datapoints and the Vals dict and start again
datapoints = 0
Vals = defaultdict(list, { k:[] for k in cols})
def init_plot():
ani = FuncAnimation(plt.gcf(), plot_update, interval=100) #update every 1 sec
plt.show()
def plot_update(prediction):
global place
plt.cla()
if prediction == 'label_left':
place -= 1
if prediction == 'label_right':
place += 1
plt.plot(place,0,'ro')
plt.xlim([-10,10])
plt.xticks(np.arange(-10,10,1))
plt.yticks([])
if __name__ == "__main__":
ip = "0.0.0.0"
port = 5000
thread = threading.Thread(target=init_plot)
thread.daemon = True
thread.start()
#Init Muse Listeners
dispatcher = dispatcher.Dispatcher()
dispatcher.map("/muse/elements/horseshoe", hsi_handler)
dispatcher.map("/muse/elements/delta_absolute", wave_handler,'Delta')
dispatcher.map("/muse/elements/theta_absolute", wave_handler,'Theta')
dispatcher.map("/muse/elements/alpha_absolute", wave_handler,'Alpha')
dispatcher.map("/muse/elements/beta_absolute", wave_handler,'Beta')
dispatcher.map("/muse/elements/gamma_absolute", wave_handler,'Gamma')
server = osc_server.ThreadingOSCUDPServer((ip, port), dispatcher)
print("Listening on UDP port "+str(port))
server.serve_forever()