Currently, most plastic products are manufactured using injection molding machines. In the force control system of a typical injection molding machine, the force information from the machine's environment is obtained using a force sensor. However, these sensors have several disadvantages, which include signal noise, sensor cost, and a narrow bandwidth. Thus, sensorless force detection methods are desirable. We have previously proposed a new injection force estimation method based on a high-order reaction force observer (HORFO), which is not significantly influenced by a nonlinear friction phenomenon. In this paper, a dual HORFO is proposed to improve the estimation accuracy of the HORFO. However, this method needs both a long calculation time and a large CPU memory. To overcome this problem, an automatic parameter-switching reaction force observer (APS-RFO) is proposed to improve the estimation accuracy of the HORFO. Moreover, this study evaluates the possibility of a sensorless force control system using the proposed APS-RFO.