In this study, we suggest a novel approach for the retrieval of regional moment tensors for earthquakes with small to moderate magnitudes. The first modification is the combined inversion of broadband and short‐period waveform data. The broadband waveforms are inverted in a frequency range suitable for surface waves, whereas for the short‐period data a frequency range suitable for body waves is applied. The second modification is the use of first‐motion body‐wave polarities to select the most probable solution out of all solutions from inversion. To combine three different criteria for selecting the most probable solution (i.e., residual from inversion, double‐couple content of solution, number of nonmatching first‐motion body‐wave polarities), the L2 norm is applied to the normalized parameters. We chose five earthquakes within the Alborz mountains, Iran, as a case study (3.1≤Mw≤4.1). In this area, several factors exacerbate the difficulty of performing inversion for moment tensors, for example, a heterogeneous station network and large azimuthal gaps. We have demonstrated that our approach supplies reliable moment tensors when inversion from broadband data alone fails. In one case, we successfully retrieved a stable solution from short‐period waveform data alone. Thus, our approach enables successful determination of seismic moment tensors wherever a sparse network of broadband stations has thus far prevented it.