This thesis investigates the feasibility and stability of using atomic emission spectroscopy as a means of determining the composition of liquid steel in an electric arc furnace (EAF). In the first step of the data processing, the observed spectra from the spectroscope are smoothed (Savitzky-Golay polynomial smoothing) and the baseline is removed, with methods based on “minimizing non-quadratic cost functions” or “penalized least-squares”. Next, the corresponding chemical elements are assigned to the peaks in the spectra with the help of the online database “NIST Atomic Spectra Database”. In the final step, calibration curves are established for each element, in order to determine, whether there is correlation between the observations and the actual content of the corresponding element in the steel sample. The experimental setup, conceived and implemented in the course of this work, covers a small-scale laboratory EAF and a spectroscope. In order to conduct the required experiments, low alloy steel samples of known chemical composition are utilized and the arc that is established between the molten steel and the graphite electrode is observed by the spectroscope. The results of this thesis lead to the assumption, that the setup, that was available for the experimental work may not be sufficiently suitable to be conclusive with respect to an industrial oven and equipment. Another major finding of this work is, that the uncontrolled nature of the arc in an EAF is not suitable as an excitation source for atomic spectroscopy, due to the instability and the sensitivity to impurities in the sample.