python 手势检测和追踪_python对肌电信号进行简单的手势识别

获取平均值起始点，并将对应时间点作用于原始信号上，对四通道信号进行行动段提取，并将长度较小的部分过滤，视为噪音

for i in range(1,5):

names['period_%s'%i]=[]

names['sta_filt_%s'%i]=[]

names['end_filt_%s'%i]=[]

for j in range(len(names['sta_%s'%i])):

names['period_%s'%i].append(names['end_%s'%i][j]-names['sta_%s'%i][j])

for k in range(len(names['period_%s'%i])):

if names['period_%s'%i][k]>5000:

names['sta_filt_%s'%i].append(names['sta_%s'%i][k])

names['end_filt_%s'%i].append(names['end_%s'%i][k])

for i in range(1,len(sta_filt_1)+1):

names['data1_cut%s'%i]=data1[sta_filt_1[i-1]:end_filt_1[i-1]]

for i in range(1,len(sta_filt_2)+1):

names['data2_cut%s'%i]=data2[sta_filt_2[i-1]:end_filt_2[i-1]]

for i in range(1,len(sta_filt_3)+1):

names['data3_cut%s'%i]=data3[sta_filt_3[i-1]:end_filt_3[i-1]]

for i in range(1,len(sta_filt_4)+1):

names['data4_cut%s'%i]=data4[sta_filt_4[i-1]:end_filt_4[i-1]]

plt.figure(figsize=(50,3))

for i in range(1,21):

plt.subplot2grid((1,20),(0,i-1),colspan=1).plot(names['data1_cut%s'%i])

plt.ylim(0,10)

plt.title('fist')

plt.figure(figsize=(50,3))

for i in range(1,22):

plt.subplot2grid((1,21),(0,i-1),colspan=1).plot(names['data2_cut%s'%i])

plt.ylim(0,10)

plt.title('open')

plt.figure(figsize=(50,3))

for i in range(1,25):

plt.subplot2grid((1,24),(0,i-1),colspan=1).plot(names['data3_cut%s'%i])

plt.ylim(0,10)

plt.title('toright')

plt.figure(figsize=(50,3))

for i in range(1,21):

plt.subplot2grid((1,20),(0,i-1),colspan=1).plot(names['data4_cut%s'%i])

plt.ylim(0,10)

plt.title('toleft')

握拳

张手

内弯

外翻

对各通道行动段求区间的平均值MAV，可以看出对于不同的动作，MAV值区别明显，可以作为特征向量对信号进行特征提取

mav_fist=pd.DataFrame(columns=['ch1','ch2','ch3','ch4'],index=[np.arange(20)])

for i in range(1,21):

mav_fist.loc[i-1,'ch1']=names['data1_cut%s'%i].ch1.mean()

mav_fist.loc[i-1,'ch2']=names['data1_cut%s'%i].ch2.mean()

mav_fist.loc[i-1,'ch3']=names['data1_cut%s'%i].ch3.mean()

mav_fist.loc[i-1,'ch4']=names['data1_cut%s'%i].ch4.mean()

mav_open=pd.DataFrame(columns=['ch1','ch2','ch3','ch4'],index=[np.arange(21)])

for i in range(1,22):

mav_open.loc[i-1,'ch1']=names['data2_cut%s'%i].ch1.mean()

mav_open.loc[i-1,'ch2']=names['data2_cut%s'%i].ch2.mean()

mav_open.loc[i-1,'ch3']=names['data2_cut%s'%i].ch3.mean()

mav_open.loc[i-1,'ch4']=names['data2_cut%s'%i].ch4.mean()

mav_toright=pd.DataFrame(columns=['ch1','ch2','ch3','ch4'],index=[np.arange(24)])

for i in range(1,25):

mav_toright.loc[i-1,'ch1']=names['data3_cut%s'%i].ch1.mean()

mav_toright.loc[i-1,'ch2']=names['data3_cut%s'%i].ch2.mean()

mav_toright.loc[i-1,'ch3']=names['data3_cut%s'%i].ch3.mean()

mav_toright.loc[i-1,'ch4']=names['data3_cut%s'%i].ch4.mean()

mav_toleft=pd.DataFrame(columns=['ch1','ch2','ch3','ch4'],index=[np.arange(20)])

for i in range(1,21):

mav_toleft.loc[i-1,'ch1']=names['data4_cut%s'%i].ch1.mean()

mav_toleft.loc[i-1,'ch2']=names['data4_cut%s'%i].ch2.mean()

mav_toleft.loc[i-1,'ch3']=names['data4_cut%s'%i].ch3.mean()

mav_toleft.loc[i-1,'ch4']=names['data4_cut%s'%i].ch4.mean()

plt.figure(figsize=(20,5))

mav_fist_ax=plt.subplot2grid((1,4),(0,0),colspan=1)

mav_fist_ax.scatter(x=np.arange(20),y=mav_fist.ch1,c='r')

mav_fist_ax.scatter(x=np.arange(20),y=mav_fist.ch2,c='g')

mav_fist_ax.scatter(x=np.arange(20),y=mav_fist.ch3,c='b')

mav_fist_ax.scatter(x=np.arange(20),y=mav_fist.ch4,c='y')

mav_open_ax=plt.subplot2grid((1,4),(0,1),colspan=1)

mav_open_ax.scatter(x=np.arange(21),y=mav_open.ch1,c='r')

mav_open_ax.scatter(x=np.arange(21),y=mav_open.ch2,c='g')

mav_open_ax.scatter(x=np.arange(21),y=mav_open.ch3,c='b')

mav_open_ax.scatter(x=np.arange(21),y=mav_open.ch4,c='y')

mav_toright_ax=plt.subplot2grid((1,4),(0,2),colspan=1)

mav_toright_ax.scatter(x=np.arange(24),y=mav_toright.ch1,c='r')

mav_toright_ax.scatter(x=np.arange(24),y=mav_toright.ch2,c='g')

mav_toright_ax.scatter(x=np.arange(24),y=mav_toright.ch3,c='b')

mav_toright_ax.scatter(x=np.arange(24),y=mav_toright.ch4,c='y')

mav_toleft_ax=plt.subplot2grid((1,4),(0,3),colspan=1)

mav_toleft_ax.scatter(x=np.arange(20),y=mav_toleft.ch1,c='r')

mav_toleft_ax.scatter(x=np.arange(20),y=mav_toleft.ch2,c='g')

mav_toleft_ax.scatter(x=np.arange(20),y=mav_toleft.ch3,c='b')

mav_toleft_ax.scatter(x=np.arange(20),y=mav_toleft.ch4,c='y')

mav_fist['action']=0

mav_open['action']=1

mav_toright['action']=2

mav_toleft['action']=3

sumup=mav_fist.append([mav_open,mav_toright,mav_toleft],ignore_index=True)

y=sumup.action

x=sumup.drop(['action'],axis=1)

from sklearn.model_selection import train_test_split

import xgboost as xgb

train_x,test_x,train_y,test_y=train_test_split(x.as_matrix(),y.as_matrix(),test_size=0.2)

xg_train=xgb.DMatrix(train_x,label=train_y)

xg_test=xgb.DMatrix(test_x,label=test_y)

param = {}

param['objective'] ='multi:softmax'

param['eta']=0.1

param['max_depth']=6

param['silent']=1

param['nthread']=4

param['num_class']=4

watchlist = [(xg_train, 'train'), (xg_test, 'test')]

num_round=5

bst = xgb.train(param, xg_train, num_round, watchlist)

pred = bst.predict(xg_test)

对四个不同的手势进行数字命名，通过xgboost进行训练分析，16个测试样的预测结果正确率为100%

更多Python视频、源码、资料加群683380553免费获取

转载至：https://zhuanlan.zhihu.com/p/41073513