A Prony method for high-quality audio coding

A technique for high-quality audio signal coding is proposed. The technique follows Prony modeling of the audio signal. A new method for dynamic estimation of the model order is proposed, based on an analysis/synthesis process with psychoacoustic criteria. According to this approach the model order is iteratively estimated, until the distortion introduced by inaccurate order selection becomes inaudible. An average coding rate gain of 6:1 can be achieved in this way without loss in signal quality.<>