Based on wavelength modulation spectroscopy and off-axis integrated cavity output spectroscopy (WM-OAICOS), a highly sensitive dual-range real-time formaldehyde (CH2O) detection trace sensor was developed to detect CH2O in human exhaled breath, which can realize the early screening of lung cancer. Also, a 5.85-mu m quantum cascade laser (QCL) and an off-axis integrated cavity were adopted. Complete ensemble empirical mode decomposition with adaptive noise (CEEMDAN) is applied to improve the signal-to-noise ratio (SNR), and the sensor is embedded with a high-precision concentration inversion algorithm. The reliability of the sensor was verified by experiments with interfering gases in exhaled breath, and the results showed that the sensor has an excellent detection accuracy. The limit of detection (LoD) is similar to 140 parts per trillion (ppt) with an 88-s integration time by the Allan-Werle deviation analysis. The experimental results of exhaled gas showed that the concentration of CH2O exhaled from different subjects was significantly different. Therefore, the sensor provides a promising way for real-time detection of CH2O concentration in human exhaled breath.