Abstract:
Our project introduces WIFITUNED, a system designed for monitoring engagement in on- line participation, such as meetings, classes, or seminars. It operates through two non-intrusive sensing modalities: WiFi CSI and audio. Utilizing WiFi CSI, WIFITUNED tracks participants’ head movements, while it discerns the speaker’s intent through audio cues. By correlating these modalities, WIFITUNED accurately detects engagement levels. Our evaluation, conducted with 22 participants, demonstrates WIFITUNED’s ability to achieve an average accuracy exceeding 96 percent using the TCN model and 94 percent using Bi-LSTMA in engagement detection.
