In a speech communication device capable of bidirectional communication, a clock deviation between the speaker side (the receiving side) and the microphone side (the transmitting side) is detected, and based on the above deviation, either a first frequency signal (speaker signal (reference signal)) or a second frequency signal (microphone signal) is frequency-shifted in the frequency domain, and therefore, an influence caused by the clock deviation can simply be corrected without an increased processing amount, and thus, an echo can be suppressed effectively.