On waveform-interpolation coding with asymptotically perfect reconstruction
Thomas Eriksson, Willem Bastiaan Kleijn · 2003
For coders which must produce high speech quality, it is beneficial to have a coding structure which gives zero distortion in the waveform when the quantizer error vanishes (asymptotically perfect reconstruction, APR). It is possible to introduce this property to waveform interpolation (WI) coders by using perfect reconstruction filter banks for analysis and synthesis. Unfortunately, the perfect-reconstruction filter banks are, in general, associated with disadvantages such as oversampling, a loss of physical meaning of the parameters, and increased delay. These disadvantages disappear for the filter bank based on the block DFT transform, but the latter method suffers from energy discontinuities. By using a pre-processor in combination with a block-DFT based WI coder, a coding structure is obtained which maintains the advantages of earlier WI coders and adds the APR property. This new structure is most useful for higher rate WI coders.