This is a part of my PhD thesis! The first paper already been published ( Paper Link).
Here is the abstract of the paper!
This paper describes an automatic mixing system for improving audio quality in teleconferencing applications.
The work was focused on applying audio effects such as multitrack level balancing, spatialization, and equalization in order to reduce speech masking, thus allowing simultaneous speakers to be heard in a teleconference.
The system used the ITU-R BS.1770 loudness measurement method and cross-adaptive audio effects to achieve average level balancing. A novel Force-directed model was implemented to automatically set the virtual position of each source. The equalization method was based on spectral decomposition techniques and a target of equal average perceptual loudness in each frequency band.
Subjective evaluation was performed in the form of a multi-stimulus listening test, which indicated that the proposed automatic mixing system could compete with a manual mix by an experienced sound engineer.
Xiaojing's website 