0.前言：
Hello，本篇文章不再探索EEG信号输入的形状问题，来讲一个比较好玩的机器学习。下面是我个人的一个研究。机器学习不必考虑数据输入的形状，最重要的一点在于特征的手动选择。在机器学习中，我们选择不同的特征作为分类器的输入，相当于深度学习中神经网络模型的调参。
最近继续看了一些论文，BCI信号作为输入，无论是原始数据作为输入还是提取了特征作为输入，都必须设计输入形状。有的论文说到，EEG信号输入必须设计特征作为输入，但其他实验证明，EEG原始信号作为输入仍然可以取得相当优秀的结果。
我发现，原始信号作为输入能够取得如此好的成绩，是因为人家设计的模型中各个超参数(kernel_size,filter_number,stride,padding,Activation Function... and etc)和原始信号的输入形状是息息相关，是非常有考究的。确定一个适合于特定输入形状的神经网络模型工作量是巨大的，在深度学习领域，这也是人们为何喜欢直接用前人建立好的模型，比如VGG/RestNet/EEGNet等；这也是在机器学习中人们设计完特征，直接在sklearn中直接调用各个模型的原因。当然了，机器学习中你可以自己写代码建立一个模型，比如写一个随机森林，我们可以对他的叶子节点数目进行修改，记得我之前教过一位研二学生，他也是搞脑电的，但硕导要求他用机器学习去做，他就是把特征设计好了，然后输入到sklearn调用的模型就完事了，我当时挺惊讶：硕士毕业这样就行了吗。他们实验团队都是这样搞得，甚至博士也是如此，只是做着重复的无意义的工作。
1、DEAP Datasets
先说一下本次研究用的数据：DEAP情感数据集。下面是数据集官网，需要你去邮件联系这家机构，说明来意，得到许可，才能下载。这里实验范式不多说，我给出数据地址：DEAP: A Dataset for Emotion Analysis using Physiological and Audiovisual Signals (qmul.ac.uk)
2、processing
预处理中，三次滤波提取了theta、alpha、beta的三个波形并做了加权平均。
3、Feature Extraction Methods
我使用了两种特征提取的方法：
小波变换(DWT)：提取EEG时间域特征，原理不多说，代码如下
%for graphs and all
numfiles = 32;
mydata1 = cell(1, numfiles);
spec=[];
hdr=lowpass_filter();
hda=alpha_band_filter();
hdb=beta_band_filter();
filter_data1 = cell(1,numfiles);
fs=128;
opfilename='testFile_01';
myfilename = sprintf('s0%d.mat', 1);
load(myfilename);
datastart=128*3;
datalength=8064-128*3;
for video=1:1
data1=zeros(1,datalength);
for ii =1:datalength
data1(1,ii)=data(video,4,datastart+ii);
end
% Plot for PSD of raw data in microVolts
psd_r=spectrum(data1,256);
hpsd_r = dspdata.psd(psd_r,'Fs',fs); % Create a PSD data object.
% plot(hpsd_r);
xlim([0 300]);
filter_data= filtfilt(hdr.Numerator,1,data1);
data2=filter_data;
avpow = norm(data2,2)^2/numel(data2);
% Plot for psd of filtered data
psd_f=spectrum(data2,256);
hpsd_f = dspdata.psd(psd_f,'Fs',fs); % Create a PSD data object.
subplot(2,1,2);
plot(hpsd_f);
xlim([0 100]);
hold all;
avpow = norm(data1,2)^2/numel(data1);
% % delta band
ts=(length(data1)/128);
Wp = [1 4]/(fs/2); Ws = [0.5 4.5]/(fs/2);
Rp = 3; Rs = 40;
[n,Wn] = buttord(Wp,Ws,Rp,Rs);
[z, p, k] = butter(n,Wn,'bandpass');
[sos,g] = zp2sos(z,p,k);
filt = dfilt.df2sos(sos,g);
fd1 = filter(filt,data1);
avpow1 = norm(fd1,2)^2/numel(fd1);
% % theta band
Wp = [4 8]/(fs/2); Ws = [3.5 8.5]/(fs/2);
[n,Wn] = buttord(Wp,Ws,Rp,Rs);
[z, p, k] = butter(n,Wn,'bandpass');
[sos,g] = zp2sos(z,p,k);
filt = dfilt.df2sos(sos,g);
fd2 = filter(filt,data1);
avpow2 = norm(fd2,2)^2/numel(fd2);
% % alpha band
Wp = [8 13]/(fs/2); Ws = [7.5 13.5]/(fs/2);
[n,Wn] = buttord(Wp,Ws,Rp,Rs);
[z, p, k] = butter(n,Wn,'bandpass');
[sos,g] = zp2sos(z,p,k);
filt = dfilt.df2sos(sos,g);
fd3 = filter(filt,data1);
fd3= filtfilt(hda.Numerator,1,data1);
avpow3 = norm(fd3,2)^2/numel(fd3);
% % beta band
Wp = [13 30]/(fs/2); Ws = [12.5 30.5]/(fs/2);
[n,Wn] = buttord(Wp,Ws,Rp,Rs);
[z, p, k] = butter(n,Wn,'bandpass');
[sos,g] = zp2sos(z,p,k);
filt = dfilt.df2sos(sos,g);
fd4 = filter(filt,data1);
fd4= filtfilt(hdb.Numerator,1,data1);
avpow4 = norm(fd4,2)^2/numel(fd4);
sumpow=avpow1+avpow2+avpow3+avpow4;
d_bpr=log10(avpow1/sumpow);
t_bpr=log10(avpow2/sumpow);
a_bpr=log10(avpow3/sumpow);
b_bpr=log10(avpow4/sumpow);
valence=labels(video,1);
arousal=labels(video,2);
dominance=labels(video,3);
liking=labels(video,4);
if(valence<5 && arousal<5)
op=1;
end
if(valence<5 && arousal>5)
op=2;
end
if(valence>5 && arousal<5)
op=3;
end
if(valence>5 && arousal>5)
op=4;
end
% fprintf(fid,'%d,%d,%.3f,%.3f,%.3f,%.3f,%d',1,video,d_bpr,t_bpr,a_bpr,b_bpr,op);
fprintf(fid,'\n');
end %the tetcase loop
fclose(myfile);功率谱密度(PSD)：提取EEG的时频特征，代码如下：
numfiles = 32;
mydata1 = cell(1, numfiles);
spec=[];
hdr=lowpass_filter();
hda=alpha_band_filter();
hdb=beta_band_filter();
filter_data1 = cell(1,numfiles);
fs=128;
for channel=1:32
fprintf('\ncreating testfile number %d:\n',channel);
if(channel<10)
opfilename ='testFile_0';
else
opfilename='testFile_';
end
filename=[opfilename int2str(channel) '.txt'];filename;
fid = fopen( filename, 'wt' );
fprintf(fid,'user,video,delta,theta,alpha,beta,op\n');
for i = 1:numfiles
fprintf('\nworking on file number %d:', i);
if(i<10)
myfilename = sprintf('s0%d.mat', i);
else
myfilename = sprintf('s%d.mat', i);
end
load(myfilename);
for video=1:40
data1=zeros(1,8064);
for ii =1:8064
data1(1,ii)=data(video,channel,ii);
end
avpow = norm(data1,2)^2/numel(data1);
% % delta band
ts=(length(data1)/128);
Wp = [1 4]/(fs/2); Ws = [0.5 4.5]/(fs/2);
Rp = 3; Rs = 40;
[n,Wn] = buttord(Wp,Ws,Rp,Rs);
[z, p, k] = butter(n,Wn,'bandpass');
[sos,g] = zp2sos(z,p,k);
filt = dfilt.df2sos(sos,g);
fd1 = filter(filt,data1);
avpow1 = norm(fd1,2)^2/numel(fd1);
% % theta band
Wp = [4 8]/(fs/2); Ws = [3.5 8.5]/(fs/2);
[n,Wn] = buttord(Wp,Ws,Rp,Rs);
[z, p, k] = butter(n,Wn,'bandpass');
[sos,g] = zp2sos(z,p,k);
filt = dfilt.df2sos(sos,g);
fd2 = filter(filt,data1);
avpow2 = norm(fd2,2)^2/numel(fd2);
% % alpha band
Wp = [8 13]/(fs/2); Ws = [7.5 13.5]/(fs/2);
[n,Wn] = buttord(Wp,Ws,Rp,Rs);
[z, p, k] = butter(n,Wn,'bandpass');
[sos,g] = zp2sos(z,p,k);
filt = dfilt.df2sos(sos,g);
fd3 = filter(filt,data1);
fd3= filtfilt(hda.Numerator,1,data1);
avpow3 = norm(fd3,2)^2/numel(fd3);
% % beta band
Wp = [13 30]/(fs/2); Ws = [12.5 30.5]/(fs/2);
[n,Wn] = buttord(Wp,Ws,Rp,Rs);
[z, p, k] = butter(n,Wn,'bandpass');
[sos,g] = zp2sos(z,p,k);
filt = dfilt.df2sos(sos,g);
fd4 = filter(filt,data1);
fd4= filtfilt(hdb.Numerator,1,data1);
avpow4 = norm(fd4,2)^2/numel(fd4);
% %gamma band
% Wp = [30 50]/(fs/2); Ws = [29.5 50.5]/(fs/2);
% [n,Wn] = buttord(Wp,Ws,Rp,Rs);
% [z, p, k] = butter(n,Wn,'bandpass');
% [sos,g] = zp2sos(z,p,k);
% filt = dfilt.df2sos(sos,g);
% fd5 = filter(filt,data1);
% avpow5 = norm(fd5,2)^2/numel(fd5);
sumpow=avpow1+avpow2+avpow3+avpow4;
d_bpr=log10(avpow1/sumpow);
t_bpr=log10(avpow2/sumpow);
a_bpr=log10(avpow3/sumpow);
b_bpr=log10(avpow4/sumpow);
valence=labels(video,1);
arousal=labels(video,2);
dominance=labels(video,3);
liking=labels(video,4);
if(valence<5 && arousal<5)
op=1;
end
if(valence<5 && arousal>5)
op=2;
end
if(valence>5 && arousal<5)
op=3;
end
if(valence>5 && arousal>5)
op=4;
end
fprintf('|');
fprintf(fid,'%d,%d,%.3f,%.3f,%.3f,%.3f,%d',i,video,d_bpr,t_bpr,a_bpr,b_bpr,op);
fprintf(fid,'\n');
end %the tetcase loop
end %the file loop
fclose(fid);
end4、Machine Learning Methods
直接在sklearn库中调用八种机器学习模型进行分类，8个模型：KMeans、SVC(linear)、SVC(rbf)、GaussianNB、RandomForestClassifier、DecisionTresClassifier、LogisticRegression、KNN，并做10折交叉验证。
DWT分类结果：
随机选取第二被试的数据
import numpy as np
from sklearn import neighbors
import pandas as pd
from sklearn import model_selection
from sklearn.model_selection import train_test_split
from sklearn import tree
from sklearn.cluster import KMeans
from sklearn.naive_bayes import GaussianNB
from sklearn import linear_model
from sklearn.linear_model import LogisticRegression
from sklearn import svm
from sklearn.svm import SVC
from sklearn.ensemble import RandomForestClassifier
from sklearn.model_selection import KFold
from sklearn.model_selection import cross_val_score
import warnings
warnings.filterwarnings("ignore")
df=pd.read_csv('testFile_02.txt')
df.drop(['user','video'],axis=1,inplace=True)
X=np.array(df.drop(['combined'],axis=1))
y=np.array(df['combined'])
kfold = KFold(n_splits=10)
xyz=[]
accuracy=[]
std=[]
classifiers=['Linear Svm','Radial Svm','Logistic Regression','KNN','Decision Tree','Naive Bayes','Random Forest','KMeans']
models=[svm.SVC(kernel='linear'),svm.SVC(kernel='rbf'),
LogisticRegression(),neighbors.KNeighborsClassifier(n_neighbors=9),
tree.DecisionTreeClassifier(),GaussianNB(),RandomForestClassifier(n_estimators=100),KMeans()
]
for i in models:
model = i
cv_result = cross_val_score(model,X,y, cv = kfold,scoring = "accuracy")
xyz.append(cv_result.mean())
std.append(cv_result.std())
accuracy.append(cv_result)
print(accuracy)[array([0.9765625, 1. , 0.9765625, 0.9921875, 0.984375 , 0.9765625,
0.9609375, 0.953125 , 0.9765625, 0.953125 ])]
[array([0.9765625, 1. , 0.9765625, 0.9921875, 0.984375 , 0.9765625,
0.9609375, 0.953125 , 0.9765625, 0.953125 ]), array([0.934375 , 0.97421875, 0.94140625, 0.9671875 , 0.97421875,
0.94921875, 0.9390625 , 0.9234375 , 0.96484375, 0.93203125])]
[array([0.9765625, 1. , 0.9765625, 0.9921875, 0.984375 , 0.9765625,
0.9609375, 0.953125 , 0.9765625, 0.953125 ]), array([0.934375 , 0.97421875, 0.94140625, 0.9671875 , 0.97421875,
0.94921875, 0.9390625 , 0.9234375 , 0.96484375, 0.93203125]), array([0.9765625, 1. , 0.9765625, 0.9921875, 0.984375 , 0.9765625,
0.9609375, 0.953125 , 0.9765625, 0.953125 ])]
[array([0.9765625, 1. , 0.9765625, 0.9921875, 0.984375 , 0.9765625,
0.9609375, 0.953125 , 0.9765625, 0.953125 ]), array([0.934375 , 0.97421875, 0.94140625, 0.9671875 , 0.97421875,
0.94921875, 0.9390625 , 0.9234375 , 0.96484375, 0.93203125]), array([0.9765625, 1. , 0.9765625, 0.9921875, 0.984375 , 0.9765625,
0.9609375, 0.953125 , 0.9765625, 0.953125 ]), array([0.87421875, 0.91171875, 0.87890625, 0.92578125, 0.9265625 ,
0.88671875, 0.9015625 , 0.8546875 , 0.91328125, 0.88125 ])]
[array([0.9765625, 1. , 0.9765625, 0.9921875, 0.984375 , 0.9765625,
0.9609375, 0.953125 , 0.9765625, 0.953125 ]), array([0.934375 , 0.97421875, 0.94140625, 0.9671875 , 0.97421875,
0.94921875, 0.9390625 , 0.9234375 , 0.96484375, 0.93203125]), array([0.9765625, 1. , 0.9765625, 0.9921875, 0.984375 , 0.9765625,
0.9609375, 0.953125 , 0.9765625, 0.953125 ]), array([0.87421875, 0.91171875, 0.87890625, 0.92578125, 0.9265625 ,
0.88671875, 0.9015625 , 0.8546875 , 0.91328125, 0.88125 ]), array([0.94921875, 0.971875 , 0.9484375 , 0.96015625, 0.96171875,
0.940625 , 0.9421875 , 0.9359375 , 0.95 , 0.9296875 ])]
[array([0.9765625, 1. , 0.9765625, 0.9921875, 0.984375 , 0.9765625,
0.9609375, 0.953125 , 0.9765625, 0.953125 ]), array([0.934375 , 0.97421875, 0.94140625, 0.9671875 , 0.97421875,
0.94921875, 0.9390625 , 0.9234375 , 0.96484375, 0.93203125]), array([0.9765625, 1. , 0.9765625, 0.9921875, 0.984375 , 0.9765625,
0.9609375, 0.953125 , 0.9765625, 0.953125 ]), array([0.87421875, 0.91171875, 0.87890625, 0.92578125, 0.9265625 ,
0.88671875, 0.9015625 , 0.8546875 , 0.91328125, 0.88125 ]), array([0.94921875, 0.971875 , 0.9484375 , 0.96015625, 0.96171875,
0.940625 , 0.9421875 , 0.9359375 , 0.95 , 0.9296875 ]), array([0.9765625, 1. , 0.9765625, 0.9921875, 0.984375 , 0.9765625,
0.9609375, 0.953125 , 0.9765625, 0.953125 ])]
[array([0.9765625, 1. , 0.9765625, 0.9921875, 0.984375 , 0.9765625,
0.9609375, 0.953125 , 0.9765625, 0.953125 ]), array([0.934375 , 0.97421875, 0.94140625, 0.9671875 , 0.97421875,
0.94921875, 0.9390625 , 0.9234375 , 0.96484375, 0.93203125]), array([0.9765625, 1. , 0.9765625, 0.9921875, 0.984375 , 0.9765625,
0.9609375, 0.953125 , 0.9765625, 0.953125 ]), array([0.87421875, 0.91171875, 0.87890625, 0.92578125, 0.9265625 ,
0.88671875, 0.9015625 , 0.8546875 , 0.91328125, 0.88125 ]), array([0.94921875, 0.971875 , 0.9484375 , 0.96015625, 0.96171875,
0.940625 , 0.9421875 , 0.9359375 , 0.95 , 0.9296875 ]), array([0.9765625, 1. , 0.9765625, 0.9921875, 0.984375 , 0.9765625,
0.9609375, 0.953125 , 0.9765625, 0.953125 ]), array([0.9765625 , 0.99921875, 0.97578125, 0.9921875 , 0.984375 ,
0.97578125, 0.9609375 , 0.953125 , 0.97578125, 0.9515625 ])]
[array([0.9765625, 1. , 0.9765625, 0.9921875, 0.984375 , 0.9765625,
0.9609375, 0.953125 , 0.9765625, 0.953125 ]), array([0.934375 , 0.97421875, 0.94140625, 0.9671875 , 0.97421875,
0.94921875, 0.9390625 , 0.9234375 , 0.96484375, 0.93203125]), array([0.9765625, 1. , 0.9765625, 0.9921875, 0.984375 , 0.9765625,
0.9609375, 0.953125 , 0.9765625, 0.953125 ]), array([0.87421875, 0.91171875, 0.87890625, 0.92578125, 0.9265625 ,
0.88671875, 0.9015625 , 0.8546875 , 0.91328125, 0.88125 ]), array([0.94921875, 0.971875 , 0.9484375 , 0.96015625, 0.96171875,
0.940625 , 0.9421875 , 0.9359375 , 0.95 , 0.9296875 ]), array([0.9765625, 1. , 0.9765625, 0.9921875, 0.984375 , 0.9765625,
0.9609375, 0.953125 , 0.9765625, 0.953125 ]), array([0.9765625 , 0.99921875, 0.97578125, 0.9921875 , 0.984375 ,
0.97578125, 0.9609375 , 0.953125 , 0.97578125, 0.9515625 ]), array([0.07734375, 0.128125 , 0.125 , 0.11484375, 0.10703125,
0.07890625, 0.1203125 , 0.14609375, 0.1109375 , 0.0609375 ])]
PSD分类结果：
随机选取第三被试数据
import numpy as np
from sklearn import neighbors
import pandas as pd
from sklearn import model_selection
from sklearn.model_selection import train_test_split
from sklearn import tree
from sklearn.cluster import KMeans
from sklearn.naive_bayes import GaussianNB
from sklearn import linear_model
from sklearn.linear_model import LogisticRegression
from sklearn import svm
from sklearn.svm import SVC
from sklearn.ensemble import RandomForestClassifier
from sklearn.model_selection import KFold
from sklearn.model_selection import cross_val_score
import warnings
warnings.filterwarnings("ignore")
df=pd.read_csv('testFile_03.txt')
df.drop(['user','video'],axis=1,inplace=True)
X=np.array(df.drop(['op'],axis=1))
y=np.array(df['op'])
kfold = KFold(n_splits=10)
xyz=[]
accuracy=[]
std=[]
classifiers=['Linear Svm','Radial Svm','Logistic Regression','KNN','Decision Tree','Naive Bayes','Random Forest','KMeans']
models=[svm.SVC(kernel='linear'),svm.SVC(kernel='rbf'),
LogisticRegression(),neighbors.KNeighborsClassifier(n_neighbors=9),
tree.DecisionTreeClassifier(),GaussianNB(),RandomForestClassifier(n_estimators=100),KMeans()
]
for i in models:
model = i
cv_result = cross_val_score(model,X,y, cv = kfold,scoring = "accuracy")
xyz.append(cv_result.mean())
std.append(cv_result.std())
accuracy.append(cv_result)
print(accuracy)
[array([0.2578125, 0.34375 , 0.375 , 0.3515625, 0.234375 , 0.4296875,
0.390625 , 0.375 , 0.4375 , 0.3046875])]
[array([0.2578125, 0.34375 , 0.375 , 0.3515625, 0.234375 , 0.4296875,
0.390625 , 0.375 , 0.4375 , 0.3046875]), array([0.2578125, 0.34375 , 0.375 , 0.3515625, 0.234375 , 0.4296875,
0.390625 , 0.375 , 0.4375 , 0.3046875])]
[array([0.2578125, 0.34375 , 0.375 , 0.3515625, 0.234375 , 0.4296875,
0.390625 , 0.375 , 0.4375 , 0.3046875]), array([0.2578125, 0.34375 , 0.375 , 0.3515625, 0.234375 , 0.4296875,
0.390625 , 0.375 , 0.4375 , 0.3046875]), array([0.25 , 0.34375 , 0.3671875, 0.3515625, 0.25 , 0.40625 ,
0.3984375, 0.3359375, 0.4375 , 0.3125 ])]
[array([0.2578125, 0.34375 , 0.375 , 0.3515625, 0.234375 , 0.4296875,
0.390625 , 0.375 , 0.4375 , 0.3046875]), array([0.2578125, 0.34375 , 0.375 , 0.3515625, 0.234375 , 0.4296875,
0.390625 , 0.375 , 0.4375 , 0.3046875]), array([0.25 , 0.34375 , 0.3671875, 0.3515625, 0.25 , 0.40625 ,
0.3984375, 0.3359375, 0.4375 , 0.3125 ]), array([0.25 , 0.2578125, 0.234375 , 0.3515625, 0.2578125, 0.2421875,
0.3203125, 0.2734375, 0.3203125, 0.1875 ])]
[array([0.2578125, 0.34375 , 0.375 , 0.3515625, 0.234375 , 0.4296875,
0.390625 , 0.375 , 0.4375 , 0.3046875]), array([0.2578125, 0.34375 , 0.375 , 0.3515625, 0.234375 , 0.4296875,
0.390625 , 0.375 , 0.4375 , 0.3046875]), array([0.25 , 0.34375 , 0.3671875, 0.3515625, 0.25 , 0.40625 ,
0.3984375, 0.3359375, 0.4375 , 0.3125 ]), array([0.25 , 0.2578125, 0.234375 , 0.3515625, 0.2578125, 0.2421875,
0.3203125, 0.2734375, 0.3203125, 0.1875 ]), array([0.203125 , 0.2734375, 0.359375 , 0.359375 , 0.25 , 0.2265625,
0.28125 , 0.2890625, 0.3125 , 0.28125 ])]
[array([0.2578125, 0.34375 , 0.375 , 0.3515625, 0.234375 , 0.4296875,
0.390625 , 0.375 , 0.4375 , 0.3046875]), array([0.2578125, 0.34375 , 0.375 , 0.3515625, 0.234375 , 0.4296875,
0.390625 , 0.375 , 0.4375 , 0.3046875]), array([0.25 , 0.34375 , 0.3671875, 0.3515625, 0.25 , 0.40625 ,
0.3984375, 0.3359375, 0.4375 , 0.3125 ]), array([0.25 , 0.2578125, 0.234375 , 0.3515625, 0.2578125, 0.2421875,
0.3203125, 0.2734375, 0.3203125, 0.1875 ]), array([0.203125 , 0.2734375, 0.359375 , 0.359375 , 0.25 , 0.2265625,
0.28125 , 0.2890625, 0.3125 , 0.28125 ]), array([0.2734375, 0.3203125, 0.3359375, 0.3203125, 0.25 , 0.2890625,
0.328125 , 0.3125 , 0.25 , 0.2890625])]
[array([0.2578125, 0.34375 , 0.375 , 0.3515625, 0.234375 , 0.4296875,
0.390625 , 0.375 , 0.4375 , 0.3046875]), array([0.2578125, 0.34375 , 0.375 , 0.3515625, 0.234375 , 0.4296875,
0.390625 , 0.375 , 0.4375 , 0.3046875]), array([0.25 , 0.34375 , 0.3671875, 0.3515625, 0.25 , 0.40625 ,
0.3984375, 0.3359375, 0.4375 , 0.3125 ]), array([0.25 , 0.2578125, 0.234375 , 0.3515625, 0.2578125, 0.2421875,
0.3203125, 0.2734375, 0.3203125, 0.1875 ]), array([0.203125 , 0.2734375, 0.359375 , 0.359375 , 0.25 , 0.2265625,
0.28125 , 0.2890625, 0.3125 , 0.28125 ]), array([0.2734375, 0.3203125, 0.3359375, 0.3203125, 0.25 , 0.2890625,
0.328125 , 0.3125 , 0.25 , 0.2890625]), array([0.1875 , 0.28125 , 0.3359375, 0.3203125, 0.28125 , 0.2890625,
0.234375 , 0.3046875, 0.296875 , 0.265625 ])]
[array([0.2578125, 0.34375 , 0.375 , 0.3515625, 0.234375 , 0.4296875,
0.390625 , 0.375 , 0.4375 , 0.3046875]), array([0.2578125, 0.34375 , 0.375 , 0.3515625, 0.234375 , 0.4296875,
0.390625 , 0.375 , 0.4375 , 0.3046875]), array([0.25 , 0.34375 , 0.3671875, 0.3515625, 0.25 , 0.40625 ,
0.3984375, 0.3359375, 0.4375 , 0.3125 ]), array([0.25 , 0.2578125, 0.234375 , 0.3515625, 0.2578125, 0.2421875,
0.3203125, 0.2734375, 0.3203125, 0.1875 ]), array([0.203125 , 0.2734375, 0.359375 , 0.359375 , 0.25 , 0.2265625,
0.28125 , 0.2890625, 0.3125 , 0.28125 ]), array([0.2734375, 0.3203125, 0.3359375, 0.3203125, 0.25 , 0.2890625,
0.328125 , 0.3125 , 0.25 , 0.2890625]), array([0.1875 , 0.28125 , 0.3359375, 0.3203125, 0.28125 , 0.2890625,
0.234375 , 0.3046875, 0.296875 , 0.265625 ]), array([0.09375 , 0.0546875, 0.1015625, 0.140625 , 0.125 , 0.1484375,
0.0625 , 0.0859375, 0.15625 , 0.0859375])]
5、结语：
通过结果对比得知，不同的特征是机器学习中影响分类准确率的重大因素。我们试想一下，为何DWT特征分类准确率每折在96%左右，而PSD分类只在30%-40%左右呢？是因为人类的情感有着强烈的时域信息吗？试着结合该数据集的原始信号波形图查看波峰，伴随着人类情感的变化，波峰是不是起伏较大呢？波峰起伏较大代表着什么呢？提取的DWT特征对应着峰值的哪些部分呢？