import sklearn

import pandas as pd

import numpy as np

import sys

import os

import shutil

import matplotlib.pyplot as plt

import scipy

from scipy import signal

import sklearn.preprocessing

from sklearn.model_selection import train_test_split

from scipy.ndimage.measurements import variance

from bokeh.layouts import column

from sklearn.decomposition import PCA

from sklearn.neighbors import KNeighborsClassifier

from sklearn.model_selection import train_test_split

from sklearn import metrics

from sklearn.model_selection import cross_val_score, KFold

from sklearn.ensemble import RandomForestClassifier

from sklearn.metrics.classification import confusion_matrix, f1_score

from sklearn.metrics import accuracy_score, precision_score, recall_score, f1_score

from statsmodels.sandbox.regression.kernridgeregress_class import plt_closeall

from astropy.visualization.hist import hist

from sqlalchemy.sql.expression import false

from sklearn.metrics import confusion_matrix

from sklearn.ensemble import RandomForestClassifier

ROOT_DIR = "C:/Users/Administrator/Desktop/elec811/project/"

PROCESSED_DIR = ROOT_DIR + "Processed/"

FIG_DIR = ROOT_DIR + "Figures/"

HEADER = ["BB-EMG", "TB-EMG", "BR-EMG", "AD-EMG", "LES-EMG", "TES-EMG", "Hand-switch", "Box-switch", "Motion-sensor1", "Motion-sensor2", "Motion-sensor3", "Motion-sensor4", "Motion-sensor5", "Motion-sensor6"]

DO_PLOT = False

SUBJECT_TO_PLOT = 3

TRIAL_TO_PLOT = 3

LOAD_TO_PLOT = 10

SAMPLING_FREQUENCY = 1000

FEATURE_WINDOW_SIZE = 200

OVERLAP = 0.5 # 50% overlapping sliding window

BUTTERWORTH_ORDER = 4

BANDPASS_LOW_CUTTING_FREQUENCY = 20

BANDPASS_HIGH_CUTTING_FREQUENCY = 490

MIN_LENGTH_TO_LABEL = 1000 # only label the consecutive box switch signal longer than 500ms, remove the shaking (usually at start and end)

# extract features from sliding window here

def calculate_features_for_each_column(column_data):

return [rms]

def calculate_waveform_length(column_data):

return sum

def calculate_zero_crossing(column_data):

return zero_crossing

def get_new_dir_for_filtered_data(dir):

return PROCESSED_DIR + "filtered/" + dir

def get_new_dir_for_labeled_data(dir):

return PROCESSED_DIR + "labeled/" + dir

def get_new_dir_for_features(dir):

return PROCESSED_DIR + "features/" + dir

def clean():

if os.path.exists(PROCESSED_DIR):

shutil.rmtree(PROCESSED_DIR)

def init_dir():

os.mkdir(get_new_dir_for_features(dir))

#*****************************************#

# 1.filter all the raw data file #

#*****************************************#

def get_dir_name(subject):

# return ROOT_DIR + "/S0" + str(subject) + " Raw Data 30 Files/"

return "S0" + str(subject) + " Raw Data 30 Files/"

def get_file_name(subject, trial, load):

# dir = get_dir_name(subject)

return "LiftCycle_" + load + "kg" + trial + ".txt"

def filter_data(subject):

data.to_csv(get_new_dir_for_filtered_data(dir) + file, sep="\t", index=False)

def butterworth_filter(data_column):

return filtered

#*****************************************#

# 2.label all file #

#*****************************************#

def get_load_from_file_name(file):

raise ValueError('label error')

def get_trial_from_file_name(file):

second_part = file.split("kg")[1]

return int(second_part[0])

def label_data(subject):

data.to_csv(get_new_dir_for_labeled_data(dir) + file, sep="\t", index=False)

# plt.plot(data.iloc[:,6], color="green")

# plt.plot(data.loc[:,"Load"] + 7, color="blue")

# # plt.hlines(base-0.25, 0, 15000, color="green")

# plt.hlines(median, 0, 15000, color="black")

# plt.yticks(np.arange(-2, 29, 1))

# plt.xticks(np.arange(0, 15000, 1000))

# plt.grid()

# plt.title("label")

# figure = plt.gcf()

# figure.set_size_inches(15, 8)

# plt.savefig(FIG_DIR + "label/" + str(subject) + "_" + str(get_load_from_file_name(file)) + "_" + str(get_trial_from_file_name(file)) +".png", dpi=100)

#*****************************************#

# 3.feature extraction #

#*****************************************#

def window_and_extract_features(data, subject, file):

features.to_csv(get_new_dir_for_features(dir) + file, sep="\t", index=False)

def feature_extraction(subject):

window_and_extract_features(data, subject, file)

#*****************************************#

# 4.catenate features #

#*****************************************#

def catenate_feature_from_same_subject(subject):

df.to_csv(feature_dir + "s" + str(subject) + "_all.features", sep="\t", index=False)

def catenate_feature_from_all_subject():

df.to_csv(PROCESSED_DIR + "final.features", sep="\t", index=False)

#*****************************************#

# 5.normalization #

#*****************************************#

def normalize_data(subject):

data.to_csv(feature_dir + "s" + str(subject) + "_all_normalized.features", sep="\t", index=False)

def plot_data(data, title, columns, subject, trial, load):

plt.show()

# fig = plt.gcf()

# fig.set_size_inches(15, 8)

# fig.savefig("C:/Users/Administrator/Desktop/elec811/project/test/" + title + ".png")

def KNN(X_train, X_test, y_train, y_test):

print(confusion_matrix(y_test, test_predict))

pd.set_option('display.max_rows', 30)

pd.set_option('display.max_columns', 30)

pd.set_option('display.width', 1000)

print("########################## start ##########################")

"""

check the folder "Processed" under ROOT_DIR for intermediate files.

"""

init_dir()

"""

bandpass between BANDPASS_LOW_CUTTING_FREQUENCY and BANDPASS_HIGH_CUTTING_FREQUENCY.

"""

print("\n ******************** filtering ********************\n")

for subject in range(1, 6):

filter_data(subject)

"""

by the shape of box-switch, as shown by the professor's pdf. then the "on platform" part and the unstable part near the contraction are removed.

only signals that last longer than MIN_LENGTH_TO_LABEL are retained.

"""

print("\n ******************** labeling ********************\n")

for subject in range(1, 6):

label_data(subject)

"""

a sliding window of size FEATURE_WINDOW_SIZE moves along a EMG column, and extract features from each window.

modify the function calculate_features_for_each_column for feature extraction.

"""

print("\n ******************** extracting features ********************\n")

for subject in range(1, 6):

feature_extraction(subject)

# catenate data from the same subject

for subject in range(1, 6):

catenate_feature_from_same_subject(subject)

"""

simple min-max normalization, on all the data from the same subject.

"""

print("\n ******************** normalizing ********************\n")

for subject in range(1, 6):

normalize_data(subject)

# catenate all

catenate_feature_from_all_subject()

print("\n ******************** see " + PROCESSED_DIR + "final.features ********************\n")

labels = None

without_labels = None

data = pd.read_csv(PROCESSED_DIR + "final.features", sep="\t")

data.dropna(inplace=True)

lab_enc = sklearn.preprocessing.LabelEncoder()

without_labels = data.iloc[:, 3:]

labels = lab_enc.fit_transform(data.iloc[:, 2])

print(pd.unique(data.iloc[:, 2]))

X_train, X_test, y_train, y_test = train_test_split(without_labels, labels, test_size=0.2)

KNN(X_train, X_test, y_train, y_test)

print("########################## done ##########################")