The COVID-19 pandemic has posed a significant threat to the health of elderly individuals, particularly those with respiratory conditions. Therefore, daily monitoring of lung capacity is necessary to assess the pulmonary condition and initiate timely treatment measures. This paper aims to explore an effective non-contact method for fine-grained pulmonary function sensing. Specifically, we propose to utilize millimeter-wave radar to contactless sensing of the overall chest and abdominal motion. Then the LUNet (Lung Unet) is constructed to recover the Expiratory Volume (EV) curves using the entire motion information of the chest and abdomen. Finally, the pulmonary function indicators are extracted based on the predicted EV curve directly. The experimental results demonstrate that the average correlation of the predicted EV curve is 96.74% and the mean relative errors for FEV1, FVC, and FEV1/FVC are 7.73%, 8.15%, and 6.70%, respectively. These results suggest that our method has the potential for clinical monitoring and assessment of respiratory diseases.