In this work, a simple and low-cost air flow sensor, based on a novel fiber-optic sensing technique has been developed for monitoring air flows rates supplied by a neonatal ventilator to support infants in intensive care units. The device is based on a fiber optic sensing technique allowing (a) the immunity to light intensity variations independent by measurand and (b) the reduction of typical shortcomings affecting all biomedical fields (electromagnetic interference and patient electrical safety). The sensing principle is based on the measurement of transversal displacement of an emitting fiber-optic cantilever due to action of air flow acting on it; the fiber tip displacement is measured by means of a photodiode linear array, placed in front of the entrance face of the emitting optical fiber in order to detect its light intensity profile. As the measurement system is based on a detection of the illumination pattern, and not on an intensity modulation technique, it results less sensitive to light intensity fluctuation independent by measurand than intensity-based sensors. The considered technique is here adopted in order to develop two different configurations for an air flow sensor suitable for the measurement of air flow rates typically occurring during mechanical ventilation of newborns: a mono-directional and a bi-directional transducer have been proposed. A mathematical model for the air flow sensor is here proposed and a static calibration of two different arrangements has been performed: a measurement range up to 3.00 × 10(-4) m(3)∕s (18.0 l∕min) for the mono-directional sensor and a measurement range of ±3.00 × 10(-4) m(3)∕s (±18.0 l∕min) for the bi-directional sensor are experimentally evaluated, according to the air flow rates normally encountered during tidal breathing of infants with a mass lower than 10 kg. Experimental data of static calibration result in accordance with the proposed theoretical model: for the mono-directional configuration, the coefficient of determination r(2) is equal to 0.997; for the bi-directional configuration, the coefficient of determination r(2) is equal to 0.990 for positive flows (inspiration) and 0.988 for negative flows (expiration). Measurement uncertainty δQ of air flow rate has been evaluated by means of the propagation of distributions and the percentage error in the arrangement of bi-directional sensor ranges from a minimum of about 0.5% at -18.0 l∕min to a maximum of about 9% at -12.0 l∕min.