Abstract : Accelerometer arrays are used in biomechanics and other fields to estimate the rigid-body acceleration field and, thence, all kinematic variables describing the rigid-body displacements. However, the progress of this technology has been limited by that of micromachined gyroscopes, which turn out to be more accurate than accelerometer arrays in most applications. The work reported in this thesis aims at improving the accuracy of the angular velocity estimates provided by accelerometer arrays. The approach is twofold: A class of accelerometer mechanical architectures is proposed, with the goal of reducing the accelerometer cross-axis sensitivity, and robust algorithms are proposed to estimate the angular velocity from some or all components of the rigid-body acceleration field. The novel class of accelerometers is inspired from parallel-kinematics-machine (PKM) architectures, taking the accelerometer proof mass to be the PKM moving platform, and the accelerometer frame the PKM base. A common characteristic of the proposed PKM architectures is that their moving platforms are connected to their bases by n + 1 legs each, n = 1, 2, 3 being their respective number of sensitive directions. For this reason, the resulting class of accelerometers is referred to as "Simplicial Multiaxial Accelerometers" (SMA). A micro-scale version of the Simplicial Biaxial Accelerometer (SBA) was devised, designed, fabricated, and tested. Furthermore, the theory behind the estimation of the angular velocity from acceleration measurements is revisited. Hence, four algorithms are proposed, which allow for the estimation of the rigid-body angular velocity from centripetal acceleration measurements alone. Based on Kalman filtering, another method is proposed in order to obtain a rigid-body angular-velocity estimate from both the centripetal and tangential components of the rigid-body acceleration field.