A real-time noninvasive and unconstrained method is proposed to determine the respiratory rhythm and pulse rate with an under-pillow sensor during sleep. The sensor is composed of two fluid-filled polyvinyl tubes set in parallel and sandwiched between two acrylic plates. One end of each tube is hermetically sealed, and the other end is connected to one of two pressure sensors. Inner pressure in each tube therefore changes in accordance with respiratory motion and cardiac beating. By employing the á trous algorithm of wavelet transformation (WT), the respiratory and cardiac cycle can be discriminated from the pressure waveforms. The respiratory rhythm was obtained from the WT 2 6 > scale approximation, and the pulse rate from the sum of WT 2 4 and 2 5 scale details without WT reconstruction after soft-threshold denoising. The algorithm's latency can be set to be minimal and the respiratory rhythm and pulse rate were estimated directly from the extracted respiration and pulse waveforms, respectively. This method has been tested with a total of about 25 h data collected from 13 subjects. By comparing the detection results with those of reference data, the average pulse rate detection sensitivity and positive predictivity were 99.17% and 98.53%, and the respiratory rhythm detection sensitivity and positive predictivity were 95.63% and 95.42%.