Detection of obstructive sleep apnoea using dynamic filter-banked features

摘要: Highlights? This study presents a methodology for OSA detection based on HRV time series. ? The is characterized by dynamic banked features extracted from its spectrogram. spectral splitting scheme to search frequency boundaries introduced. Attained results show similar performance compared with another outcomes. An easier clinical interpretation of HRV-derived parameters obtained. There need developing simple signal processing algorithms less costly, reliable and noninvasive Obstructive Sleep Apnoea (OSA) diagnosing. One the promising directions provide analysis heart rate variability (HRV), which clearly shows non-stationary behavior. So, feature extraction approach, being capable capturing information suitable detection, remains an open issue. Grounded discriminating capability bands activity between normal patients, can be extracted. However, some spectrograms resemble like pathological ones, vice versa; so, prior extract set, energy spatial contribution contained in each sub?band should clarified. paper set short-time that are spectrogram introduces scheme, searches components alike stochastic behavior improving accuracy. Two different approaches considered (heuristic relevance-based); both them performing minute-by-minute classification comparable other outcomes reported literature, but avoiding more complex methods or computed features. For validation purposes, tested 1-min HRV-segments estimated 50 Physionet database recordings. Using parallel combining k-nn classifier, assessed reaches as much 80% value accuracy, splitting. oriented research focused finding alternative used costly diagnosing additional benefit parameters.