| 摘要 |
A method and system for encoding a series of input digital signals in a higher order digital format, such as 16 bit, to a series of lower order digital signals such as 8 bit, and then recovering the original input signals with a greater accuracy than previously attainable is described. A distinct reduction in quantization noise is produced, making the system compatible with very high resolution audio equipment such as compact disks. For each input signal, a particular pair of complementary encoding and decoding transformations are selected from a set of numerous possible transformations. The transformations are nonlinear, with high resolutions near the origin and much lower resolutions further away. The high resolution range increases from table to table, while the absolute resolution within said ranges decreases. The value of each input signal is predicted from previous signals, and a differential quantity representing the prediction error is used in selecting the table for each signal. Overall quantization noise is significantly reduced by updating the table selections so that the differential quantity remains within the high resolution range. Identical decoding circuitry is provided in both the encoder and decoder to assure tracking between the two. An analog-to-digital conversion is provided at the input, with a reverse conversion at the output. |
