A speaker system is disclosed which is capable of supplying reproduced sounds vibrating in substantially the same manner as in the respiratory sphere to human's sense of hearing by using conventional unidirectional speaker units in combination in a contrived arrangement, and by applying real time digital signal processing by means of a digital signal processor to the speaker units to cancel a peak and a dip in frequency response and in phase response through inverse correction which cannot be canceled only by improving the arrangement of the speaker units, thereby forming a sound emitter capable of providing ideal reproduced sounds. The speaker system comprises an enclosure EC having a basic structure of a hollow 32-hedron composed of 12 pentagonal flat surfaces 1 and 20 hexagonal flat surfaces 2, speaker units 7, 8 or 78 mounted in all or 25-31 of the 32 surfaces, and a real time digital signal processing system inserted in a input line of each of the speaker units 7, 8 or 78. The digital signal processing system inverse-characteristically filtering driving signals of the speaker units to evenly correcting a peak 4 and a dip 5 caused in frequency response and in phase response of each of the speaker units 7, 8 or 78.