Audio Signal Source Separation

Kurzfassung

A third level institute in Dublin has developed a method, algorithm and software capable of separating individual sound sources from a mixture of sound sources in real-time. Audio source separation technology has a multitude of uses in areas such as telecommunications, professional audio production, surround sound up mixing, music education and automatic metadata generation for music to name but a few. A license agreement is sought.

Details

The innovative technology uses a cross-channel time frequency transform in order to calculate the mixing coefficients of each source after which individual sources can be re-synthesised. Various parameters allow users to specify which source to separate and how much separation is to be achieved. The real-time software includes an interface which gives users real-time audio and visual feedback indicating source activity. As well as separating sources, the algorithm can also estimate the direction of arrival of sources in certain cases; this may have applications in security systems. In the context of audio post production, the algorithm has uses in; mastering, noise reduction, speech enhancement for field reporting, and surround up-mixing.

The audio research group at the third level institute is a group of 11 researchers working in the domain of signal processing for audio applications. The group has had over 50 peer reviewed publications in the last 3 years and are partners in several national and internationally funded research projects.

Innovative Aspects:
This is the only technology of its kind commercially available and is also protected by a patent. Other approaches to the problem struggle to operate in real-time whilst our approach is capable of operating several times faster than real-time.