Source separation out of a mix of signals has been under development for many years with different approaches. We use timefrequency representations of two microphone signals to estimate the mixing parameters of the source signals. In order to evaluate the robustness of the algorithm under real-world conditions we built a real-time implementation, which is suitable to detect the sources, their mixing parameters and performs the source separation based on the mixing parameters. Our implementation only needs a few parameters and then works as a stand-alone solution with the opportunity to apply further post-processing or digital audio effects to the source signals.

Download

This paper gives an example of an auto-regressive (AR) spectral analysis on transient musical sounds. The attack part of many musical sounds is mostly too short to be analysed by a short-time Fourier analysis, whereas this short period of time is long enough for several AR-methods. The AR-spectra obtained from short segments of signals with attack transients have a sufficiently high frequency resolution. These spectra contain more information about the evolution of a sound than a fast Fourier transform made over a small amount of samples.

Download