The survival rate in cardiac arrest is associated to the quality of the chest compressions (CCs) and ventilations provided during cardiopulmonary resuscitation (CPR). Hyperventilation remains common whenever ventilation is manual during resuscitation from cardiac arrest. The capnogram is used to monitor respiration and ventilation rates. During CPR chest compressions induce artefacts in the capnogram signal that challenge the detection of ventilations. The evaluation of ventilation detectors during CCs has not been well characterized. In this study an algorithm for ventilation rate monitoring and hyperventilation detection was developed. The processing method consists of detecting transitions in the first difference of the signal, and applying feature based classification to identify every ventilation. The instantaneous rate and hyperventilation minutes were then computed. A set of 20 out-of-hospital episodes, totalling 50864 s (86.6% during CCs) and 6305 ventilations was used to define and evaluate the algorithm. The algorithm had a sensitivity/ positive predictive value of 96.9%/96.2% respectively for the ventilation detection (96.7%/95.8% during ongoing CCs), 98.7%/98.7% for the hyperventilation detection, and a mean error of 0.4 (0.8) min−1 for the instantaneous ventilation rate.