PURPOSE: To provide a codec which increases a cell channel capacity twice and can be put to practical use for electric communication by using two pitch estimate values and two sets of quantization filter coefficients to classify sound frames into two modes.

CONSTITUTION: In an audible sound data compression system, audible sound data is received by an audible sound frequency preprocessor 31 and is divided into audible sound frames, and in a next block 32, two audible sound windows which have centers placed in the center and the end of an audible sound frame are used to perform linear predictive analysis and quantization. In a block 33, the output of the block 32 is used to generate two pitch estimate value with two similar audible sound windows, and an A mode where voiced sound is dominant or a B mode where unvoiced sound is dominant is determined as a mode responding to pitch estimate value in a block 34. The determined mode determines excitation modeling in a block 35, and thereafter, compressed sounds are packed by a processor 36.