Recognizing steps taken by people walking is a technique used in a method for estimating a person's indoor position called Pedestrian Dead Reckoning (PDR), and also plays a key role in health-care applications such as pedometers and activity trackers. Generally, accelerometers are used for step recognition, because the sensor signals exhibit clear periodic changes as the user walks. As an alternative, step recognition methods that use angular velocity or magnetic sensors to measure hip rotation have also been proposed. In this article, we propose another step recognition method that uses an air-pressure sensor. We discovered that air-pressure sensor values vary periodically as the user walks in cases such as when a smartphone is held in a hand and the user consciously swings their arms while walking, when it is placed in a sealed case attached to the waist, or when placed in a trouser pocket. We perform step recognition by applying a simple algorithm to these periodic signals, with digital filtering, extreme value detection, and thresholding. Experiments confirmed that steps were estimated with an error rate of approximately 7.5%. Performing sensing with alternate sensors in this way can help reduce the number of sensors needed in a device or system, contributing to reducing power consumption and device size. As an example, activity monitors sometimes calculate the number of steps using accelerometers, and the number of stairs ascended or descended using an air-pressure sensor. These research results could be used to implement an activity monitor that only uses an air-pressure sensor.