The purpose of this study was to analytically exploit the capabilities of head-mounted systems instrumented with linear accelerometers (ACs) for field use in redundant configurations. We simulated different headsets equipped with uni-, bi- or triaxial sensors with a number of axes that lie in the range of 12–24; the ACs were located on a hemispherical surface by adopting a priori criterion while their orientation was randomized. In addition, for a comparative purpose the nine accelerometer scheme (one triaxial AC and three biaxial ACs addressed in the following as “3-2-2-2 configuration”) was also analyzed in the present paper. We simulated and statistically assessed the performances of hemispherical headsets in the test case of a healthy subject walking freely at normal pace over level ground. The numerical results indicated that a well designed instrumented headset can retrieve the angular acceleration and component with rms errors of about 2% and 0.5%, respectively, and angular velocity with a drift error of about 20% in a trial. On the contrary, the pose of the headset cannot be evaluated because of the drift induced by the integration process. In general, we can state that headsets with uni-, bi- or triaxial ACs have comparable performances. The main implications of the above-mentioned observations are (a) neither expensive triaxial ACs nor assembling procedure based on the use of orthogonal mounting blocks are needed; (b) redundant arrays of low-cost uni- or biaxial ACs can effectively be used to reach adequate performances in biomechanical studies where head acceleration and velocity are investigated; (c) while estimates of angular acceleration with accelerometers are accurate, estimations of angular velocities, linear velocities and pose are not.