摘要 |
A method (400) for determining that an audio conference is in a first phase of a plurality of phases. The method comprises determining (401) a plurality of sequences (301) of events (305) for a plurality of terminals (120, 170); wherein an event from the sequences (301) of events (305) indicates that a talker activity at a terminal (120, 170) has been detected; determining (403) a sequence (350) of feature vectors (330) based on the plurality of sequences (301) of events (305); wherein a feature vector (330) of the sequence (350) of feature vectors (330) is indicative of the talker activity (332) of at least one of the plurality of terminals (120, 170) relative to the talker activity at least another one of the plurality of terminals (120, 170); and determining (406) that the audio conference is in the first phase based on the sequence (350) of feature vectors (330). DETAILS: audio-conference phases identified as: I) introductory phase (comprising a waiting phase, an uncoordinated activity or chatting phase, and a participant introductions/attendance check phase); II) cooperative phase (comprising productive discussions between participants); III) sign-off phase. In addition to participant join/leave/mute events, voice-activity detection algorithms are used for measuring a ratio of voice-activity of each participant (RTT, Relative Talking Time) and other features (CTT, Concurrent Talking Time, NT, number of Turns, NAE, Number of Active Endpoints) and for inferring the curren phase of the conference; conference logs or recordings may be used instead of real-time measurements; the time difference (delay) between scheduled time or first call and the reach of phase II is used as a measure of efficiency of the conference. |