Model-Informed Speech Enhancement Using Virtual Room Acoustics and Acoustic Descriptor Optimization | Synapse