Capturing and storing samples only when needed is a way to ensure a drastic reduction of the data to be processed, which is a crucial issue in many applications, such as autonomous and communicating smart devices (Internet of Things). This leads to nonuniform data for which specific processing chains have to be designed. In this paper, we discuss the issues to be address to generalize Infinite Impulse Response filters to the nonuniform case. We illustrate the performance of the constructed filters on an electrocardiogram signal, for two ways to obtain the nonuniform samples, and select linear decimation and the bilinear scheme as a good combination for this application.