Abstract : The dynamics of the human middle ear (ME) has been studied during the last years due to the interest in better understanding its different aspects and to observe the influence of different pathologies and their corrective surgeries, or the use of passive or active prostheses. In order to characterize the ME dynamics several approaches have been proposed, including experimental procedures and computational models. Among the computational models, the finite element (FE) method has been widely used because of the level of detail in the definition of geometries and materials. However, its application implies in a high computational cost and the observation of the relative movement between the components of the ossicular chain is not clear. In the other hand, other simpler models of lumped parameters (LP) have also been used, however limiting the analysis of ME vibration to 1- D. As an alternative, multi-body (MB) models combine the analysis of flexible structures with rigid body dynamics, resulting in a lower number of degrees of freedom (DOFs) than the FE model and a simpler description of the dynamics of the ME in 3-D. In the present work the development of several MB models of the human ME is proposed, through the simplification of a FE model. Initially, the geometry, mechanical properties and boundary conditions of the FE model are defined by comparing the model responses with experimental results obtained with human temporal bones (TBs) and with experimental results presented in the literature. Afterwards, the MB models are developed following a simplification sequence: i) applying the MB formulation for flexible bodies; ii) considering the ossicles as rigid bodies; iii) replacing ligaments and tendons individually by lumped springdamper elements; iv) replacing the joints individually by kinematic joints with different DOFs; and v) implementing all the simplifications together. The simplified representation of the flexible components allows to analyze the influence of the dynamic properties, especially the stiffness, on the ME dynamics. The responses of the different models obtained in each stage of simplification are compared in terms of the FRF of stapes velocity vs. sound pressure at the TM, the velocity in other points in the ossicular chain, and the vibration modes using correlation methods. Finally, a reduced MB model of the human ME is used to analyze the response of a piezoelectric MEMS accelerometer when implanted at different positions of the human OC. In addition, some restrictions of the implantation process are considered, such as the surgical access, methods of fixation and accuracy and precision in sensor placement.