We propose two methods for coding and transmitting three-dimensional multichannel sound signals: scalable coding and transmission, and modeling of the quantization error. The first method converts N-channel sound signals into M-channel basic and (N-M)-channel additional signals using a matrix operation. The matrix is trained by simulated annealing to reduce its condition number and the energy of additional signals. The unmasking artifact may occur when the N-channel signals are restored from the decoded signals using the inverse matrix. The second method estimates the quantization error signals by the polynomial approximation of the decoded signals. Experimental results showed that the combination of both methods could realize a 1.2 Mbps scalable transmission of 22-channel sounds without a notable sound degradation.
This paper costs $33 for non-members and is free for AES members and E-Library subscribers.