MyoTac: Real-Time Recognition of Tactical Sign Language Based on Lightweight Deep Neural Network

Table 2

Accuracy of different signal inputs.

Tag	Gestures	EMG	IMU	Fused signals
0	Male	98.0%	8.0%	99.8%
1	Female	50.0%	98.0%	99.8%
2	Commander	34.0%	50.0%	84.0%
3	Hostage	22.0%	60.0%	90.0%
4	Suspect	12.0%	20.0%	94.0%
5	You	99.8%	98.0%	99.8%
6	Me	60.0%	54.0%	96.0%
7	Come on	56.0%	12.0%	36.0%
8	Hear	4.0%	8.0%	42.0%
9	See	98.0%	98.0%	98.0%
10	Advance	96.0%	86.0%	96.0%
11	Message received	0.0%	6.0%	98.0%
12	Hurry up	58.0%	88.0%	90.0%
13	Stop	96.0%	0.0%	99.8%
14	Cover me	12.0%	99.8%	99.8%
15	Not understand	86.0%	99.8%	99.8%
16	Understand	0.0%	58.0%	96.0%
17	Squat down	20.0%	1.0%	99.8%
18	Ignore	8.0%	98.0%	99.8%
19	Pistol	24.0%	92.0%	82.0%
20	Rifle	94.0%	99.8%	99.8%
21	Automatic weapon	4.0%	98.0%	88.0%
22	Shotgun	0.0%	58.0%	90.0%
23	Car	26.0%	96.0%	99.8%
24	Doorway	52.0%	99.8%	98.0%
25	Corner	99.8%	99.8%	99.8%
26	Assemble	98.0%	98.0%	99.8%
27	Single column	1.0%	90.0%	99.8%
28	Two-way column	98.0%	98.0%	99.8%
29	One-way line	40.0%	99.8%	98.0%