PURPOSE: To reduce stably and effectively the distortion of a whole sound reproduction system including a speaker, by extracting the distortion component of the whole sound reproduction system and feeding back only the microsignal corresponding to the distortion component.

CONSTITUTION: An adder 5 adds an input signal ei to the output of a differential amplifier 7 and supplies them to an amplifier 1 and a filter 8 respectively. The output of the amplifier 1 is supplied to an end of a voice coil 9 of a speaker 2, and the other end of the coil 9 is grounded. An electrode 12a formed with a cylindrical metal is provided to the outer circumference surface of a bobbin 11 of the speaker 2 at the area excepting the coil 9. At the same time, an electrode 12b which is coaxial to the electrode 12a and made of a cylindrical metal is provided on the side surface of a pole piece 14 of a bottom yoke 13. The vibration of the bobbin 11 is detected as a change of electric capacity between electrodes 12a and 12b. Then this output of detection is supplied to an inverse input terminal of a differential amplifier 7 via an operational amplifier which form a voltage follower 19. The output of the filter 8 having the transmission characteristics equivalent to the amplitude-frequency characteristics of the speaker 2 is supplied to the non-inverse input terminal of the amplifier 7.