A probability model describing the output of a wideband stereo audio coder

We investigate the problem of variable-bit rate coding of wideband stereo audio signals for transmission over asynchronous transfer mode networks. We develop a layered source coding algorithm which first decomposes the stereo signal into subband signals. The subband signals are then analyzed with respect to their relative energy and are assigned to high and low priority data streams. The bit rate at the output of audio coder has to be characterized and the statistics of the bit rate should be provided at the time when the initial connection to the network is established. We derive a probability distribution model for the output data rate expressed in terms of the asymptotic distribution of the DFT coefficients. The derivation of the probability distribution of the output data rate assumes knowledge of the PSD function of the audio signal.