Dynamic respiratory modeling for non-contact live monitoring by particle filter approach

Although a Doppler radar detects the respiratory motion in the non-contact way, there is a problem of the performance decrement in proportion to the distance between a radar and a human body due to the reduction of the reflected rate. A model-based method is proposed to evaluate the respiratory presence in such cases. This model expresses the chest-wall displacement. The expression is composed of the periodic function which has five parameters. This model has a new property that the inspiration and the expiration are given by each of the two independent parameters. Therefore the problem is solved by tracking the radar outputs using the particle filter framework. The experiment was carried out to the six subjects at the distance 3.25m. The result showed that there was the statistically significant difference between the evaluation value of the unattended state and that of the attended one. In addition, the estimated model was favorably compared with the reference data which was measured by the high-precision displacement sensor. Consequently, the efficacy of the proposed method for the long distance (>3.00m) and that of the proposed respiratory model are established.