This paper proposes a new speech enhancement system named SPAD (Speech Processing system by use of Auto-Difference function). SPAD is a non-parametric speech enhancement system and a modified version of SPAC(Speech Processing system by use of Autocorrelation function). The algorithm of SPAD is very simple, but SPAD can reduce white noise and pink noise as large as 19 dB (at accumulation time is 20 ms). Results of computer simulation and hearing tests reveal that SPAD can reduce noise level as same as SPAC does.