Decreasing a stabilizing yaw moment by a large side-slip angle of a vehicle causes an unstable motion in a nonlinear region. Recently, in order to stabilize the vehicle motion in the region, Direct Yaw Moment Control (DYC) by using differences of driving and braking forces on right and left wheels has been developed and focused. Especially, performances of the handling and the stability are improved by DYC became DYC the side-slip angle together with a yaw rate control. For such DYC, the side-slip angle is crucial information. However, it has been difficult to utilize a control system by using DYC with the side-slip angle, because a special devise for the measurement is necessary. Moreover, time-integration is not suitable for sensor data such as the lateral acceleration, the vehicle velocity, and the yaw rate because of an accumulation of noise and measurement errors. In this paper, a new estimation method for the side-slip angle is introduced and developed by applying a layered neural network. The yaw rate and the lateral acceleration acquired by ordinary sensors are used as inputs. The training for the neural network is done by using sampled signals such as a yaw rate, a lateral acceleration and a side-slip angle obtained by non-contact ground speed sensors. Results show good agreements between estimated and experimental side-slip angles through outside experiments on a concrete road.