Fig.1

Dual-Mach-Zehnder schematic diagram"

Fig.2

Comparison curves of inertia weights and fitness values of different PSO algorithms"

Fig.3

Recognition steps of optical fiber vibration signal based on MPSO-SVM"

Table 1

Experimental instrument parameters"

Fig.4

Four kinds of optical fiber vibration signal patterns"

Fig.5

Singular value sequence distribution graph of jumping signal"

Fig.6

Comparison between original signal and denoising signal"

Table 2

SNR and RMSE when rank order is 5, 6, 7"

Table 3

Comparison of denoising effects of different methods"

Fig.7

Characteristic distribution of different optical fiber vibration signals"

Fig.8

Comparison of recognition accuracy between PSO-SVM and MPSO-SVM in different evolution times"

Fig.9

Comparison of recognition time between PSO-SVM and MPSO-SVM in different evolution times"

Table 4

Relative recognition rates between PSO-SVM and MPSO-SVM after 30 times run  %"

Table 5

Comparsion of run time between PSO-SVM and MPSO-SVM after 30 times run  s"

Table 6

Relative recognition rates after 100 times run when evolution time is 50, 40, 30, 20  %"

Table 7

C and g of different parameter optimization methods"

Table 8

Vibration signal recognition rates of different parameter optimization methods"

	无扰 动/%	跑跳/ %	攀爬/ %	敲击/ %	平均识 别率/%	平均识别 时间/s
网格搜索法	100	96.7	86.7	60	80	0.69
交叉验证法	96.7	86.7	93.3	80	89.2	47.26
PSO-SVM	100	86.7	93.3	90	92.5	2.4
MPSO-SVM	100	96.7	96.7	96.7	97.5	2.38

Fig.10

Effect comparison of different recognition methods"

Table 9

Comparison with traditional neural network recognition methods"

	无扰 动/%	跑跳/ %	攀爬/ %	敲击/ %	平均识 别率/%	平均识别 时间/s
ELM神经网络	83.3	90	86.7	70	82.5	0.54
BP神经网络	96.7	86.7	96.7	80	90	0.79
PNN	93.3	70	26.7	73.3	65.8	1.72
MPSO-SVM	100	96.7	96.7	96.7	97.5	2.38