Adjust to the Jitter

Audio applications that can adapt to Internet congestion will deliver much higher sound quality.

Rainer Mauth

The architecture of the Internet is not a good environment for real-time audio and video transmissions. The delay between the arrival of subsequent packets, known as jitter, depends on the traffic conditions on the Net. Packet-loss rates often vary between 15 percent and 40 percent. Delays between packets can be as much as 1.5 seconds, seriously compromising the quality of conversation over the Net.

New protocols, such as the Resource Reservation Protocol (RSVP) and the Real Time Transport Protocol (RTP), might eventually improve the quality of service on the Net. However, they are not yet widely used.

Another approach is to adapt applications to the jitter present on the network. A group of researchers at the French National Research Institute, INRIA (Sophia-Antipolis), has developed an extended, adaptive loss-recovery and rate-control system that gives reasonable quality to audio transmissions with loss rates as high as 50 percent.

The INRIA audio tool adjusts the audio packet send rate to the current network conditions, adds redundant information to packets when the loss rate surpasses a certain level, and establishes a feedback channel to control the send rate and redundant information. Simply put, the scheme minimizes the impact of packet loss and delay between subsequent packets on perceived audio quality.

In a process called forward error correction (FEC) , the system adds to each packet a highly compressed version of the previous packet. When the network load and packet-loss rates are high, the process increases the amount of redundancy carried in each packet. It does this by adding to each packet compressed versions of the previous three to four packets.

The complete process is controlled by a feedback loop that gradually increases the send rate if the loss rate is above a certain threshold. In 5-second intervals, the receiver returns quality-of-service reports to the sender to adapt the amount of redundant information being sent.

"Adaptive internet telephony might eventually provide better voice quality than the standard 8-kHz sampling on today's telephone networks," says Jean-Chrysostome Bolot, a project manager at INRIA.

---

Where to Find

INRIA
Sophia-Antipolis, France
Phone:    +33 493 65 7747
Internet: http://www.sophia.inria.fr

---

Better Sound over the Internet

illustration_link (30 Kbytes)