Block diagram of the optimization process for the ART system with four loudspeakers. The upper outputs from the RAZR room simulation are used in the reproduction chain, which includes mapping the simulated virtual sources onto four loudspeakers, convolving the loudspeaker signals with their respective BRIRs in the playback room, and applying Gammatone filtering. The lower output from RAZR is the binaurally simulated room used as a reference signal (BRIR_ref). For simplicity, only the BRIR_ref corresponding to one ear is depicted. The direct and reverberant components of BRIR_ref are separated at time T_ref, and Gammatone filtering is applied in the same manner as for the upper outputs. As shown in the upper part of the figure, the direct signal is rendered using the two front loudspeakers, while the reverberant signal is rendered using all four. The Gammatone filterbank outputs of the rendered binaural room-in-room impulse responses (BRinR) and the BRIR_ref are then compared to optimize the energy of the direct sound using g_d⁽ⁱ⁾ gains, and the energy of the reverberant sound as well as the interaural coherence (IC) of the rendered BRinR using g_r⁽ⁱ⁾ gains and cross-mixing coefficients α⁽ⁱ⁾ and β⁽ⁱ⁾.