Abstract: An acoustic system was developed to detect the sugar content of watermelon nondestructively, which consists of a steel ball wrapped with rubber, 6 piezoelectric accelerate sensors, an electricity amplifier, an optical trigger, a data acquisition card and a computer. Acoustic transmissivity can be calculated from the amplitude spectrum. And 6 characteristic frequencies, 752、869、1001、4556、322、3950 Hz, were chosen with the function of SMLR (Stepwise Multiple Linear Regression) using TQ software. The acoustic transmissivity at frequencies were used to setup a multi regression models for the detection of watermelon sugar content. 47 watermelons were tested, and the results showed that the best detection model can be obtained by tapping the equator part and receiving the signals from the opposite side with watermelon being put in its growing position, the R value was 0.80753, RMSEC was 0.646, RMSEP was 0.655. The research proved the possibility of the system in detection of sugar content of watermelon and can provide a reference for further research in this area.