The present invention is a speaker system for reproducing sound close to the actual sound of instruments or actual acoustic field. The instruments output sounds in various directions. For example, reverse phase sound is output and human voice is output in forward direction. Therefore, the speaker capable of reproducing stereophonic sound close to the original sound without limiting a listening point is required. In a speaker cabinet 5, left and right speakers 3, 4 are arranged back to back with each other to output normal phase sound and reverse phase sound from left and right. Front left and front right speakers 1, 2 are slightly facing to each other to output the sound like a voice by a point sound source. Sounds of the front speakers and left and light speakers are overlapped with each other and stereophonic sound can be reproduced.