In this paper, an abstract definition and formal specification is presented for the task of adaptive-threshold OSAHS events detection and severity characterization. Specifically, a low-level pseudocode is designed for the algorithm of raw oximetry signal pre-processing, calculation of the 'drop' and 'rise' frames in the related time series, detection of valid apnea/hypopnea events via SpO2 saturation level tracking, as well as calculation of corresponding event rates for OSAHS severity characterization. The designed algorithm can be used as the first module in a machine learning application where these data can be used as inputs or encoded into higher-level statistics (features) for pattern classifiers, in the context of computer-aided or fully automated diagnosis of OSAHS and related pathologies.