This paper introduces a method for indoor self-localization of a monaural microphone, which is required for various location-based services. By generating two pairs of dipole sound fields, localization is performed on each device, irrespective of the number of devices, based on orthogonal detection of observed signals and some simple operations that are feasible with limited computational resources. A method using multiple source frequencies for enhancing robustness against the effects of reflection and scattering is also proposed. The effectiveness of this method was evaluated by numerical simulations and experiments in an anechoic chamber and indoor environment, and the average errors for the azimuth and zenith angles were 4.8 and 1.9 deg, respectively, in the anechoic chamber and 21 and 11 deg, respectively, in the indoor environment.