Voice mail system

Using the IRT lab's RTSP multimedia storage server, design an Internet voice mail system that can replace a telephone answering machine.

The storage server receives PLAY, RECORD and SHUTDOWN commands from the SIP server to play the outgoing message and to record the voice mail.

The SIP proxy or redirect server redirects incoming calls to an answering machine server to be written as part of this project.

This separate SIP user agent acts as the "answering machine" and sends out RTSP requests to the server to play the outgoing message and record audio/video based on the session description contained in the call setup message. It also sends email to the user with a URL containing the audio file created, a description of the call (Date, Subject, Organization, and Priority headers from the SIP message) and possible a SIP URL to call back the original caller. It might be possible to construct this user agent from the SIP server by adding appropriate cgi-bin interfaces. This might also make it easy to set things up so that the outgoing message, for example, depends on the time of day or day of week.

The user plays back recorded messages using, for example, a RealAudio server or a home-grown RTSP client. It might be useful to integrate this client with the SIP Tcl client, so that the user has a single phone/answering machine.

The RTSP server needs to be extended with recording capabilities, including the ability to DESCRIBE the recording. Caller information such as time of call and subject (from SIP header) should be recorded in the session description file created. The recording should be started after the outgoing message has completed, by using the information returned in the RTSP PLAY response.

Initially, DTMF signals are carried in-band as RTP packets. There are existing routines to detect DTMF tones and translate them into digits.

The outgoing message should be recordable using any standard audio recording tool. The outgoing message should be switchable via a web interface. Consider using a text-to-speech system to render the user's vacation message if it is activated. Consider a CPL-like language for selecting the OGM based on caller identity, time-of-day and other inputs.

Longer range goals include control of playback and recording of the OGM through a regular telephone, using the 3Com gateway.

Azeem Khan


Last updated by Henning Schulzrinne