Air flow estimation via a Wiener filter approach

In this work, a new methodology to estimate the respiratory air flow rate from pulmonary sounds is proposed. Previously in the literature, it has been suggested to use the auto-regressive model to estimate the flow rate from pulmonary sounds, moreover, it has been presented that sounds are more correlated with the flow rate in certain frequency bands. In this work, on the other hand, the optimal auto-regressive model order for the estimation is determined, and the estimates from the optimal auto-regressive model and from the frequency bands are combined through a Wiener filter to increase the success of the estimation. The correlation coefficient is increased to 0.71 with the proposed method, which was 0.66 for the optimal-order auto-regressive model and 0.62 for the most-correlated frequency band.