Statistics for DPCM with Forward Gain-Adaptive Quantizer and Simple Switched Predictor for High Quality Speech Signals