Context. An increasing number of potentially habitable terrestrial planets and planet candidates are found by ongoing planet search programs. The search for atmospheric signatures to establish planetary habitability and the presence of life might be possible in the future. Aims: We want to quantify the accuracy of retrieved atmospheric parameters (composition, temperature, pressure) that might be obtained from infrared emission spectroscopy. Methods: We use synthetic observations of the atmospheres of hypothetical potentially habitable planets. These were constructed with a parametrized atmosphere model, a high-resolution radiative transfer model and a simplified noise model. The simulated observations were used to fit the model parameters. Furthermore, classic statistical tools such as χ2 statistics and least-square fits were used to analyze the simulated observations. Results: When adopting the design of currently planned or proposed exoplanet characterization missions, we find that emission spectroscopy could provide weak limits on the surface conditions of terrestrial planets, hence their potential habitability. However, these mission designs are unlikely to allow the composition of the atmosphere of a habitable planet to be characterized, even though CO2 is detected. Upon increasing the signal-to-noise ratios by about a factor of 2-5 (depending on spectral resolution) compared to current mission designs, the CO2 content could be characterized to within two orders of magnitude. The detection of the O3 biosignature remains marginal. The atmospheric temperature structure could not be constrained. Therefore, a full atmospheric characterization seems to be beyond the capabilities of such missions when using only emission spectroscopy during secondary eclipse or target visits. Other methods such as transmission spectroscopy or orbital photometry are probably needed in order to give additional constraints and break degeneracies.< Réduire