Introduction

For many consumers, Internet radio was their first exposure to streaming media. The other driver for interest in compressed audio has been the universal acceptance of the MP3 format for low data rate audio files. Although not a streaming format (it is for download and play), many popular streaming architectures can be used to wrap MP3 files for streaming. Although MP3 gradually is being replaced with the more efficient AAC codecs, it still retains a very wide user base. There are several audio-only architectures that come from the Internet radio arena: Nullsoft's Shoutcast and Winamp, and Audion for the Mac platform. The primary multimedia architectures MPEG-4, Windows Media, QuickTime, and Real all support audio-only streaming as well as video/audio content.

To stream audio you have to go through three processes:

• Digitization

• Encoding (compression)

• Packetization

The professional digital audio format, AES-3, is often the basis for digitization. The AES standard has a data rate of over 3 Mbit/s for stereo audio sampled at 48 kHz. To transmit this over a 28-kbit/s dial-up circuit, the data would have to be reduced by a factor of 100:1, and that excludes the overheads of packetization and error resilience.

The compression schemes split between the speech (voice) coding developed by the telcos and the military, and the general audio or waveform encoders developed for multimedia applications. The speech coding algorithms generally use a model of the vocal tract and transmit voicing parameters for reconstruction by the decoder into intelligible speech. Much of...