DUET using automatic peak detection and histogram thresholding

The degenerate unmixing estimation technique is a blind source separation algorithm used to separate audio source signals from the stereo audio mixtures without earlier learning. In this paper, a new blind source separation algorithm is developed utilizing automatic peak detection and histogram thresholding. First, a power weighted 2D histogram is created from the ratio of the time-frequency domain of the microphone signals. The 2D histogram detects a peak point for each original signal automatically with maximum position corresponds to the relative attenuation and delay mixture parameters. Further, the time-frequency masking is employed to separate one mixture into the original sources. The proposed methodology just requires a couple of parameters, and it acts as a solitary answer with a chance to implement extra post processing features or digital audio effects to the original audio sources. The test results on audio mixtures prove the advantages and superiority of the proposed approach.