To avoid back pain and related diseases, an appropriate sitting posture should be maintained. Inertial measurement units (IMUs) or marker-less motion cameras, such as Kinect (c), has recently been used to achieve simpler posture measurements than optical motion capture camera systems. However, multiple IMUs can affect the natural posture of users. The space requirement to guarantee reliable camera data is also somewhat excessive (>1 m) for some personal space setups. Therefore, we propose an unobtrusive method for estimating sitting posture on the basis of ground reaction force measurement, which can be achieved without the use of markers or additional space for measurement. To eliminate additional measurement information other than the ground reaction force underneath the chair and desk, we modeled the posture as a multi-segment rigid body. Several assumptions were proposed and verified to simplify the model and data processing without deteriorating the posture information. Furthermore, to examine whether the combined GRF information provides the appropriateness of the posture, we performed sitting tests for various postures. Results showed that the combinations of GRF measurement could reasonably estimate the sitting posture by the simplified rigid body model and could reliably differentiate the inappropriate forward bent posture. The results showed that the proposed method could serve as a sensing mechanism of posture monitoring systems.