 Hello, I am Takumi Abe from the University of Tokyo, Japan. I would like to talk about transducer projection in ultrasonic phase array to form multiple forces on a non-planar surface. AUTD can remotely provide tactile sensation by driving a large number of ultrasonic transducer arrays in an array. It can form focal points at desired location or generate fine sound pressure distribution. Other tactile presentation devices have been proposed, but AUTD has some advantages such as temporary and specially high resolution, low latency, and don't require the user to wear a device. To form desired sound pressure distribution, we have to solve optimization problems to determine driving signal for transducer. Various methods have been proposed and they are effective enough in free space. Some methods can obtain driving signal at high speed. However, this method may not work when focal formation on non-planar surface such as human hand because of shielding by object itself. Conventional methods are inefficient because all transducer try to contribute to the focus formation to all target points, so we try to form the focus efficiently by projecting transducer to be used for each target point. When the object itself becomes an obstacle, the angle between the normal of the object surface at the target point and the direction from the target point to the transducer is considered to be large. Therefore, we introduce pre-selection metric whose value becomes lower as the angle becomes larger. The transducer that is above the threshold is pre-selected by each target point before the optimization calculation. For example, when pinching the partial object in the center from the side and the set threshold to zero, dividing role as the left side transducer showing red from the focal point on the index finger and the right side showing blue from the focal point on the sub. Since the pre-selection can be represented by a transfunction matrix, the program formation is the same as before. So conventional first method can be used for proposed method and we can derive the transducer efficiently with small computational cost. We perform American simulations. In the case of pinching the object as mentioned earlier, the pressure on the desired surface was weak by conventional method and except pressure was upright on the other side. On the other hand, the pressure distribution on each surface was improved by proposed method. Next, we change the angle difference between the two normal directions. When the angle difference was large, the proposed method is effective. On the other hand, when it was small, in other words, that normal directions are concentrated in one direction, effect of proposed method decreased. However, we would avoid negative effect by setting appropriate threshold. This is Conclusion. Thank you for listening.