Multisource DOA Estimation in Impulsive Noise Environments Using Convolutional Neural Networks

Table 6

DOA3 subnetwork architecture.

Layer and activation		Output shape	Parameter
L0:	Input	8×8	0
L1:	Convolution + Relu	80 × 8 × 8	800
L2:	Convolution + Relu	80 × 8 × 8	57680
L3:	Convolution + Relu	48 × 8 × 8	34608
L4:	Convolution + Relu	48 × 8 × 8	20784
L5:	Convolution + Relu	16 × 8 × 8	6928
L6:	Flatten	1024	0
L7:	Dense + Relu	12	12300
L8:	Dense + Relu	3	39