The example of the invention discloses a process for establishing the audio frequency conference connection, which comprises distributing at least two media servers MS according to the number of terminals which join in the audio frequency, choosing one MS as a root node MS and the rest MS are used as leaf nodes MS, distributing the terminal on the MS, respectively establishing the connection between the terminal and the MS on which the terminal is located, and establishing the connection between all the leaf nodes MS and the root nodes MS. Additionally, the invention discloses a device for establishing the audio frequency conference connection. The invention can cascade several media servers when a single device can not fulfill the demand of the conference, and realizes the conference without limiting the terminal number, thereby solving the problem that the conference can not be held which is caused by the lack of the media server resource under the operation condition.