Complex human-computer interfaces are more and more making use of high-level concepts extracted from sensory data for detecting aspects related to emotional states like fatigue, surprise, boredom, etc. Repetitive sensory patterns, for example, almost always will mean that the robot or agent will switch to a "bored" state, or that it will turn its attention to other entity. Novel structures in sensory data will normally cause surprise, increase of attention or even defensive reactions. The aim of this work is to introduce a simple mechanism for detecting such repetitive patterns in sensory data. Basically, sensory data can present two types of monotonous patterns: constant frequency (be it zero or greater than zero, be it a unique frequency or a wide spectrum) and repetitive frequency spectrum changes. Both types are considered by the proposed method in a conceptually and computationally simple framework. Experiments carried out using sensory data extracted both from the visual and auditory domains show the validity of the approach.