In the nocturnal stable boundary layer, turbulent heat fluxes are not continuous but rather intermittent. Intermittent heat fluxes are characterized by brief episodes of turbulent heat flux with intervening periods of relatively weak noise or immeasurably small fluctuations. Their characteristics should be studied by extracting them from experimental data with weak noise and other small fluctuations. In this paper, we propose a new method of extracting intermittent heat fluxes. In this method, the probability density function of intermittent turbulent fluxes is analyzed and compared with the stable distributions. A threshold is defined by the position at which the probability density function begins to deviate from the stable distribution. We find that intermittent turbulent heat fluxes can be effectively extracted from experimental data. On the basis of our results, we compare the power spectra of turbulent heat fluxes before and after extraction. We found that after extraction, the proportion of variance of the low-frequency signal decreases, but that of the high-frequency signal increases slightly. In addition, the power spectra follow a ?7/6 power law at higher frequencies.