Estimation of the seismic source signature is an important problem in reflection seismology. Existing methods of source signature estimation (statistical methods and well-log-based methods) suffer from several drawbacks. For example, assumptions of whiteness of the earth response, stationarity of the data, and the phase characteristics of the wavelet have no real theoretical justification and the extracted wavelets may not be reliable. Here, I introduce a method of extracting the source signature based on the theory of seismic interferometry, also known as the virtual source method. Interferometry can be used to extract the scaled impulse response between two receivers. This is the Green's function scaled by the power spectrum of the source wavelet. If a source location coincides with one of the receiver locations (not necessarily a zero-offset receiver), the recording at the other receiver would be the Green's function convolved with the source signature. The scaled impulse response, thus differs from the real recording by having an extra source term convolved with it. Deconvolving the real recording with the scaled impulse response gives the source signature, and so this method is named as "Virtual Real Source". Through modeling examples, I show that the Virtual Real Source method produces accurate source signatures even for complicated subsurface and source signatures. The quality of the extracted wavelet can be improved by using particular time windows and stacking wavelets extracted from different time windows. Source variability within a seismic survey does not pose any problems because interferometry averages the source signatures and the individual source signatures can be extracted reliably using this method. Source signature of each shot can be extracted reliably if they all have similar amplitude spectra even though their phase spectra might be completely different. This method of source signature estimation not only gives accurate traveltimes and amplitudes of reflection events, but also has the potential to solve other issues, such as finding source radiation patterns, measuring intrinsic attenuation, and estimating statics.