Using the technique of unknown input observer, this paper aims at estimatinginternal variables of people living with a complete spinal cord injury (SCI). The goal is to providea better understanding on the sitting control strategy of SCI people. The observer design is basedon a Head-Two-Arms-Trunk (H2AT) model, belonging to a class of nonlinear descriptor systems.For observer design, this model is represented in a specific Takagi-Sugeno (TS) fuzzy form withnonlinear consequents. In contrast to previous fuzzy estimation results based on conventionalTS fuzzy modeling, the new TS formulation allows separating all unmeasured premise variablesin the nonlinear consequent parts. This contributes to reduce the computational burden of theobserver design and the structural complexity of the designed fuzzy observer. In particular,the new formulation enables a more effective way to deal with unmeasured premise variables.Using Lyapunov stability theorem, sufficient conditions to design the unknown input observerare derived in the form of linear matrix inequalities, conveniently solved by convex optimizationtechniques. Simulation results demonstrate the effectiveness of the proposed observer design