PURPOSE: To attain the decoding of voice without occurring synchronizing step- out at the reception side by dividing the transmission frame of a voice signal into plural frames, giving identification information ID to each divided frame and designating silence to the identification information ID to a silence division frame.

CONSTITUTION: A frame conversion section 3 divides a voice frame into N sets of frames, 10msec long each and an identification ID code is given to each frame. A code designating whether the voice of the frame is a sound frame or a silence frame is set to the identification code ID based on a silence detection signal (b) from a silence detection section 1. A converted frame (d) is inputted to a packet data insertion section 4, where a packet data (f) is inserted to a frame in which the identification code ID designates the silence. Thus, a voice data forming one frame by N×10msec and a packet data forming one frame by 10msec are transmitted while being multiplexed at the silence and sure frame synchronization for voice decoding is taken at the reception side.