A nonparametric statistical approach to breath segmentation

A nonparametric statistic is used to identify change-points along the respiratory waveform that correspond to shifts between inspiration and expiration, thereby segmenting the airflow waveform into discrete breaths. The nonparametric statistic used to segment the respiratory waveform is robust in the presence of wild-points, which makes it well suited for processing the often artifactual respiratory airflow waveform. This technique can be used to identify extremely shallow trends as long as the digitizing resolution is adequate to distinguish the sample magnitudes within the N-tuple. As a result, a wide range of breath magnitudes can be identified. Computationally, this technique requires only addition, subtraction, and a look-up table, so it can be implemented efficiently with a microprocessor.<>