Abstract: Improved down mixing of audio channels of compressed digital audio signals by down mixing in the frequency domain. Fast virtual transform is applied to transform short DCT coefficients into long DCT coefficients, and down mixing is performed on the long DCT coefficients. Inverse discrete cosine transform is performed on the down mixed set of long DCT coefficients, generating signals in the windowing domain. The windowing domain signals are then overlapped and added to generate time domain signals suitable for further amplification. Down mixing in the frequency domain reduces the number of computations required.