Deep neural networks for Multi-Room Voice Activity Detection: Advancements and comparative evaluation