Table II. Feature-Based Techniques for Extraction of Respiratory Signals.

Extract BW as the mean amplitude of peaks and preceding troughs [18] or the mean signal value between consecutive troughs [180]. Extract AM as the difference between the amplitudes of peaks and preceding troughs [8]. Extract FM as the time interval between consecutive peaks [8]. Extract peak amplitudes [8]. Extract trough amplitudes [180]. PCA of heartbeats [181], [227]. Extract the kurtosis between adjacent peaks [80]. Extract the morphological scale variation of part of the signal (e.g., QRS complexes) by comparing the current beat to a template beat [45]. Extract QRS durations [179]. Extract QRS areas [196]. Extract maximum Q–R or R–S slopes (using either a straight line fit [113] or fourth-order central moments [185]) or QRS-wave angles from the difference between them [113]. Extract PPG pulse widths [111]. Extract the difference between the durations of the upslope and downslope of the PPG [207]. Extract the direction or magnitude of the mean QRS vector axis using the arctangent of the ratio of QRS complex areas from two simultaneous ECG leads [195], [224]. Extract rotation angles of vectorcardiogram loops using multiple ECG leads [53]. Extract the main direction of the electric heart vector at a specific phase in the cardiac cycle (e.g., T-wave) [165]. Extract the pulse transit time using the R-wave of the ECG and the subsequent PPG pulse onset [44], [97], peak [213], or upslope midpoint [110]. Extract the maximum upslope during diastole of a venous signal extracted from dual wavelength PPG signals [219].