Lossy compression
Regardless of the advances in UDP and RTSP transmission protocols, streaming media would not be possible without the rapid innovation in encoding algorithms or codecs that compress and decompress audio and video data. Uncompressed audio files are huge. One minute of playback of a CD-quality stereo audio file requires 10 MB of data, approximately enough disk space to capture a small library of books or a 200-page web site.
Standard modem speed connections--including cable modems and xDSL systems--do not have the capacity to deliver pure, uncompressed CD-quality 16-bit, 44.1 kHz audio. In order to stream across the limited bandwidth of the Web, audio has to be compressed and optimized with codecs, which are compression-decompression encoding algorithms. In general, compression schemes can be classified as "lossy" and "lossless."
Lossy compression schemes reduce file size by discarding some amount of data during the encoding process before it is sent over the Internet. Once received on the client side, the codec attempts to reconstruct the information that was lost or discarded. The benefit to this sort of compression lies in the smaller file size that results from discarding the "lost" information. The JPEG image format uses lossy compression to sample an image and discard unnecessary color information. Similarly, lossy audio compression discards frequencies on the high and low end of the spectrum and attempts to locate and remove unnecessary audio data. The technique is often referred to as "perceptual encoding" since the user is unlikely to notice the absence of this information. Lossy compression offers file savings on the order of 10:1.
Since small file size is so important on the Internet, practically all of the formats we're interested in employ lossy compression. Here's how it works. First, the client player decompresses the audio file as it downloads to your computer. Then it fills in the missing information according to the instructions set by the codec. To illustrate why lossy compression is so crucial, consider the phrase, "Now is the time for all good men to come to the aid of their country". One way to compress this would simply be to remove all the vowels and spaces: "Nwsthtmfrllgdmntcmtthdfthrcntry".
That cuts the message from 71 characters to 31, a 56% file savings, but of course our compressed message is unintelligible. Imagine that our codec, however, has appropriate rules for decompressing this message with minimal distortion. The conversion likely wouldn't be perfect, but it would be good enough to understand the message, something like, "Now's tha ti'm for oll gudm en to com to the aad of their country".
This is exactly what happens with lossy audio compression. The compressed file is unintelligible to the listener; the decompressed file is intelligible but of a lower quality than the original.
For example, a RealAudio speech file encoded from a standard AIFF or WAV file is generally one-tenth the size of the original file after encoding. To reduce that file's size, first you preserve the integrity of the 1,000 Hz to 4,000 Hz frequency spectrum of the human voice and then discard the frequencies above and below those ranges. By eliminating the unnecessary low- and high-end frequencies, the encoder is able to reduce the file size while maintaining speech intelligibility. It should be noted that speech tends to have aural characteristics (sound) that extend into the 7,000 Hz range. When the area between 4,000 Hz and 7,000 Hz is reduced or removed entirely, encoded speech will sound intelligible, but it may lose clarity and sound unnatural. Furthermore, since some voices and sounds often reach into even higher frequency ranges, lossy compression and encoding can result in dull, muted, or abrasive sounds.
|
