| 摘要 |
<p>PURPOSE:To obtain vector quantizer of high-efficiency by separating a DC component whose probability model is hard to set from an input signal and quantizing it by vectors. CONSTITUTION:An encoder arranges a sample sequence of an input signal into a specific number of blocks, which are inputted as input vectors VC to a register 2. A mean value operator 3 calculates and inputs the mean value of the blocks to a parallel adder 7. An output VC code table memory 5 is stored previously with divisions of a signal space for minimum distortion and an optimum output VC and address counter 4 reads the outputs VC out of the memory 5 successively and inputs them to an adder 7 through a register 6 to adds it to the mean value, calculating a DC correction output VC; and this correction output is subtracted from the input VC by a subtracter 8 and the difference is converted into an absolute value to obtain element distortion, which is outputted from a converter 9. The minimum value of the element distortion is detected by detectors 10 and 11 and latched in a latch 12 to output an index signal 14, which is mixed with the mean value 13 at a section 15 to send out an encoded output signal 16. A decoder separates the signal 14 and mean value 13 at a section 7 and processes them to output a DC correction VC18.</p> |
