Algorithms that manipulate, remix, or transform stereo audio signals to suit the format of sound reproduction systems make assumptions about the way in which the stereo signal was originally mixed. However, these algorithms usually do not test if the assumptions are valid. In order to determine the type of mixing used in a stereo mix, this research compares two blind classification algorithms that divide individual time–frequency regions into six classes of mixing types. The results show that mixing strategies vary with the musical genre, and the classification algorithm can reduce the instability and center disintegration when upmixing a stereo signal to a multichannel format.
This paper costs $33 for non-members and is free for AES members and E-Library subscribers.