Sätt och anordning för audiokodning

摘要

Consecutive frames of an audio signal are transformed into the frequency domain using a modified discrete cosine transform (MDCT) so as to generate a stream of spectral data values for each spectral component. A set of linear prediction coefficients are calculated for each spectral value and are based on a predetermined number of previously determined reconstructed values. The coefficients are recalculated only after receiving several spectral values and thus the same coefficients are used for several consecutive predictions before being updated. A predicted spectral value is generated using the coefficients and the error between the predicted and actual spectral values is calculated. The rate at which the coefficients are updated may be variable. A decoder for receiving and decoding the error values is also disclosed. Application is to a mobile phone system.