An Improved Hybrid Companding Delta Modulator

We present two new forms of "hybrid companding" delta modulation for speech encoding. These algorithms differ from the earlier hybrid companding delta modulation (HCDM) scheme in that they employ Song voice adaptation as the basis of the instantaneous companding, rather than constant factor adaptation. We show that although the two new algorithms produce SQNR curves which are similar to that of HCDM, a subjective evaluation of the three systems significantly favors the new ones. In addition, the new algorithms offer simpler implementation in digital hardware.