Abstract Personal factors are frequently involved in various kinds of accidents, and an important one is not being fully awake. In order to prevent such accidents, the development of a system is required that detects the point just before the dozing off state is reached. Results of our experiments about changes of electroencephalograms (EEGs) when carrying out monotonous tasks show that special kinds of EEGs called grouped α waves appear just before this point. In this paper, we define this point indicated by the appearance of the grouped α waves as 'inclined sleep' and propose a sensor which detects the inclined sleep stage by the appearance of the grouped α waves on EEGs. Then, we evaluate the sensor from the viewpoints of stability and reliability, considering the algorithm used for the sensor.