The interdigital transducer (IDT) of the traditional Mach-Zehnder (MZ) acousto-optic modulator on a silicon-on-insulator (SOI) platform is located outside its two arms. The crest and trough of the surface acoustic wave (SAW) are used to modulate the two arms of the MZ interferometer so as to achieve a high modulation efficiency. Therefore, the distance between the two arms must be odd multiples of half acoustic wavelength. However, since the substrate is usually not uniform, the wavelength of the SAW changes as it transmits through the surface of the device. As a result, the exact distance between the two arms is difficult to choose. On the other hand, the SAW losses a portion of energy after passing through the first arm of the MZ interferometer, so the modulation of the second arm becomes much weaker. To solve these problems, we propose a new structure where its IDT is situated in the middle of the two arms of the MZ interferometer. With this scheme, the two arms of the MZ interferometer are located exactly at the crest and the trough of the SAW, while they are modulated with equal strength. In this paper, we first use the finite element method to simulate the acoustic frequency and the surface displacement of the excited SAW. Then we deduce the refractive index variations of all layers according to their acousto-optic effects. After that, we analyze the influences of different factors on the acousto-optic modulation efficiency, including the type and size of waveguide, the thickness of zinc oxide (ZnO) layer, and the area it covers, the number of electrodes, etc. These parameters are accordingly optimized to enhance the modulation efficiency. Modeling result based on COMSOL Multiphysics indicates that when the width of the strip waveguide is 6 m, the ZnO layer covers only the area under the IDT and has a thickness of 2.2 m, and the number of the electrodes is 50, the effective refractive index variation of the waveguide reaches 4.0810-4 provided that the amplitude of the driving voltage is 1 V. This value is 12% higher than that of the traditional structure.