EURASIP Journal on Audio, Speech, and Music Processing

Table 2 Parameters for TCN and CRNN model architecture

From: A large TV dataset for speech and music activity detection

Model Arch.	Parameters	Values
TCN	Kernel size	{3, 5, 5}
	No. filters	{32, 16, 32}
	No. stacks	{9, 5, 2}
	No. dilations	{3, 7, 2}
	Use skip connections	{False, true, true}
CRNN	Kernel size	{3, 11, 11}
	No. filters	{64, 64, 16}
	No. GRU units	{80, 40}

Back to article page