Synthesizing Multichannel Recordings Based on Adaptive Spectrogram Morphing

In this paper we propose two graph-based morphing approaches that can synthesize microphone signals with acoustical characteristics which match those of a desired microphone. The overall system involves the short-time Fourier transform (STFT) and cosine similarity measure (CSM) for formulating and matching signals in the time frequency domain, respectively. In addition, the spectrogram is adjusted by morphing algorithms according to the trend extracted from the pairwise reference spectrograms. We also present results of the synthesis of microphone signals that simulate what a spot microphone would pick up near a target instrument.