ABSTRACT
Event-based signal acquisition differs from traditional data acquisition in the features that one tries to capture from a continuous-time signal. But as a common point, one obtains a discrete description of the input signal in the form of a sequence of real values. The sampling- reconstruction problem is then brought to the familiar ground of linear algebra in finite dimension. The different type of data acquisition however escapes from the setting of Shannon sampling theorem, and the theoretical question of signal reconstruction from the discrete events needs to be mathematically revisited. Event-driven operations such as in continuous-time quantization are by construction time invariant but not linear. The appealing point of the nonuniform sampling approach is the preservation of linearity, a property that is at least commonly understood in finite-dimensional vector spaces where linear operations are described by matrices.
