IEEE Consumer Electronics Magazine - March 2018 - 70

0
−20
−40
−60

175

Z

2,000

165
160
150
145
155
40
35
30 50
45

280

285

290 190 295
200
205
195

305
300
310
315

1,800
−50

320
55

−100

60

X
(a)

−100

315
75
85
310 80
90
305
95
295
300
290

70

65

−150
−200
Y
(a)

−3
20
−3
00
−2
80
−2
60
−2
40
−2
20
−2
00
−1
80
−1
60
−1
40
−1
20

−200

175

1,900
275

215
235
225
230
220
340
345
350
335
330

210

205

170

185
70
255
260
265
270
75
85
80

200

325

65

−140

185195
190
180

2,100

180

−120

−180

2,200

60

−100

−160

160
165
170

155

150

55

−80
Y

145

135 140
40
45 130
50

−250

−150
X
−200

0
160
155
165
170
270175
380
275
385
390
280
180395
285
290 400

−100
Y

40
45
55
60 65 70
50

−150

305

325
330
320
335

95

420
425
430

200
210
230
215
225
220

−250
−400 −350 −300 −250 −200 −150 −100
X
(b)
−80

2535
20
30 40

−100

50

45

2,100

55

60

2,000
1,900
1,800

435
455
480
460
490
450
495
475
465
470
485
440
445

205

315

400
395
390
95
385
90
100
85
380
375
370
80
365
360 75
155
150
160 165 260
355
170
265
175
345
350
270
210
205
215
200
180
275
185 190 195
70
280
65
285
60
55
30
35
45
50
40
290

2,200

110
100
105

415

195

310

−200

190

300

2,300

90
405
85
410

185
75 295 80

Z

−50

−150

295310
300
305

1,700
0

−50

−100

−50

−100

−150
Y
(b)

−200

−250

−200

X

65 70
75
80

−120
−140

2,300
130
145
150
140
135

155
160

−180

165

2,100

−200
−220
−240
−260
−350

240 265
245
250
260
270
255

−300

275

−250

280

−200
X
(c)

285

120
115
125
110
105

2,200

175185
190
180
170
195

Z

Y

−160

−100

70
75
85
80

190
185
170
180
175

−50

1,800
−50

−100

305
300
295

−150
Y
(c)

−200

−100

290
285
265280
270275

1,900

300
320
330
315
340
305
335
325
310
290 295 345

−150

2,000

100
95
90

325
320
315
310

210
205
200
195

−250

−150
−200
−250

X

FIGURE 11. (a)-(c) The recording trajectory of writing characters
directly facing the stereo camera, corresponding to S1.

FIGURE 12. (a)-(c) The recording trajectory of writing characters
facing the left of the stereo camera, corresponding to S2.

We apply this character input method in a VR environment to free users of the limitations of various devices. As
it can realize natural 3-D hand-gesture recognition, it will
be more convenient and accurate because of our feature
selection and network training. For the character recording, stereo calculation makes it more robust than traditional 2-D methods.

Accurate trajectory recording and high recognition accuracy indicate high communication capacity of this gesture
interaction method with computers. When controlling the targeted VR devices or texting messages in a VR system, 3-D
natural-gesture recognition and character recording can offer
more information than traditional methods. However, the 3-D
information used in the algorithms and the training process

70 IEEE Consumer Electronics Magazine

^

MARCH 2018
Table of Contents for the Digital Edition of IEEE Consumer Electronics Magazine - March 2018
