A frequency band compression method, using Time Compression, for use with a frequency division multiplex (FDM) signal carrying voice signal. On the transmission end, the multiplexed signal is divided into a plurality of time segments, each segment having a certain duration, then deleting a part of the each segment and time-expanding the remaining part of the segment by a sawtooth variable delay, thus lowering the frequency of the signal to a predetermined ratio. On the receive end, the signal segments are compressed by a corresponding variable delay, thus making the received frequency higher and reproducing the original multiplexed voice signals.