Defining respiratory epochs using an adaptive AR filter

As part of the analysis of respiratory signals, it is usually necessary to define timing events such as the end (beginning) of inspiration and/or the end (beginning) of expiration. Traditional methods using derivatives often require scaling and smoothing to give acceptable results. The work described in this paper has employed an adaptive, lattice structure, Autoregressive (AR) filter to replace the derivatives in locating the end of inspiration and of expiration. The time averaging property of the filter algorithm, together with a variable forgetting factor serves to reduce sensitivity to noise which plagues derivative techniques. In addition, the use of filter coefficients to locate respiratory timing information removes the need for separate scaling. Finally, employing the forgetting factor to define the length of time history observed by the filter permits utilization over a wide range of respiratory rates under program control. The technique is general and versatile enough to be implemented in microprocessor based instrumentation.