Resistive flex sensors were increasingly used in different areas for their interesting property to change their resistance when bent. In particular, they can be applied to human segment in biomedical devices to register static and dynamic postures. In spite of their interesting properties, such as robustness, low price and long life, they often demonstrate non-linear response and lower sensitivity at small bending angles. This paper provides investigation to improve flex sensors linearity and sensitivity to measure body joint angles with better accuracy. To this aim, an empirical model of the sheet (or surface) resistance of the active layer, to simulate its behavior against the layer shape and size as well as the bending angle, was provided, to investigate whether changes of the standard rectangular shape can improve sensitivity and linearity. In addition, to date commercial flex sensors have been characterized only against the bending angle with a radius of curvature smaller than the device length, so limiting the application to small joints such as finger or knee. In order to extend the flex sensor applications, for instance, to measure the trunk posture in back disease and rehabilitation monitoring, the sensor response against a radius of curvature greater than the sensor length was analyzed. Finally, a new modeling technique, based on the inverse model of the sensor characteristic, to enable fast measurements of the bending angle or the radius of curvature from sensor response also in real time, and fast calibration procedures, fitting the same model to measurements with different joint size and even device, were developed.