Multiobjects Association and Abnormal Behavior Detection for Massive Data Analysis in Multisensor Monitoring Network

Table 1

Detailed structure parameters of each layer in LSTM.

Layer	Module	Size of convolution kernel	Number of convolution kernels	Activation function	Step	Parameter	Output
0	Input	—	—	—	—		300 × 1
1	1d-Conv	5 × 1	16	ReLU	1	96	300 × 16
2	BN	—	—	—	—	128	300 × 16
3	1d-Conv	5 × 1	16	ReLU	1	1424	300× 16
4	BN	—	—	—	—	1456	300 × 16
5	Max pool	2	16	—	2	32	150 × 16
6	1d-Conv	3 × 1	32	ReLU	1	3024	150 × 32
7	BN	—	—	—	—	3088	150 × 32
8	1d-Conv	3 × 1	32	ReLU	1	6192	150 × 32
9	BN	—	—	—	—	6256	150 × 32
10	Max pool	2	32	—	2	64	75 × 32
11	1d-Conv	5 × 1	64	ReLU	1	16560	75 × 64
12	BN	—	—	—	—	16688	75 × 64
13	1d-Conv	5 × 1	1	ReLU	1	28800	75 × 1
14	BN	—	—	—	—	28928	75 × 1
15	Max pool	2	1	—	2	2	38 × 1