Digital Audio Compression

The most common system for nonlinear quantization is called p-law (mu law), which uses 8 nonlinear bits to achieve the same approximate SNR as 12-bit linear quantization. Nonlinear quantization is used in telephone systems, where 8-bit plaw at 8 kHz is used to transmit speech over a standard US digital phone line. 8-bit plaw decoding is easily accomplished with a 

3 MPEG and Other Frequency Domain (Transform) Compressors 

Transform coders endeavor to use information about human perception to selectively quantize regions of the frequency spectrum in order to hide the quantization noise in places that the human ear will not detect it. The frequency spectrum will be defined and discussed in Chapter 5. These coders strive for what is often called perceptual losslessness, meaning that human listeners cannot tell the difference between compressed and uncompressed data. 

Time masking is also computed, recognizing that weaker auditory events can linger significantly in time (100 ms or so) after a loud event, and can also even precede a loud event by a short time (10 ms or so) without being heard. 

Ken Steiglitz. A Digital Signal Processing Primer With Applications to Digital Audio and Computer Music. Menlo Park Addison Wesley, 1996. 

---

ATSC, 1995] ATSC (1995). Digital audio compression (AC-3) standard. Doc. A/52/10, U.S. Advanced Television Systems Committee. 

The application encoders/decoders as shown in Figure 16.49, refer to the bit rate reduction methods (also known as data compression) appropriate for application to the video, audio, and ancillary digital data streams. The purpose of compression is to minimize the number of bits needed to represent the audio and video information. The DTV system employs the MPEG-2 video stream syntax for the coding of video and the ATSC standard digital audio compression (AC-3) for the coding of audio [2]. 

The following service types are defined in the digital audio compression (AC-3) standard [2] and in the ATSC digital television standard [ 1 ] ... 

ATSC Standard A/52A. 2001. Digital Audio Compression (AC-3). Advanced Television Systems Committee, Washington, D.C. 

Data compression can be used to increase the capacity of numerous audio storage products such as hard and magneto-optical (MO) disk-based digital audio recorders and work stations. A modest compression ratio of 4 1 enables a 1-h stereo CD to be stored on 160 megabyte (MB) of disk space and to store some 30 s of full bandwidth stereo on a 1.44 MB, 31/2-in floppy diskette. Cinema audiences are now being offered virtual audio splendor with audio surround systems incorporating data compression. 

Smyth, S. 1992. Digital audio data compression. Broadcast Engineering (Feb.) 52-60. 

Bruno, R. Digital Audio and Video Compression, Present and Future, presented to the Delphi Club, Tokyo, Japan, July 1992. 

Smyth, S. Digital audio data compression. Broadcast Engineering Magazine, pp. 52-60, Feb. 1992. Springer, K.D. Interference between EM and Digital M-PSK Signals in the EM Band, National Association of Broadcasters, 1992. 

MP3 files contain audio that is digitally encoded using an algorithm that compresses the data by a factor of about eleven but yields a reasonably faithful reproduction. The quahty of sound reproduced depends on the data sampling rate, the quality of the encoder, and the complexity of the signal. 

Compression Removing portions of a digital signal to store audio data in less space. 

This section presents an overview of the both theory and practice of data compression. The discussion is quite general, as it applies to any kind of digital information regardless of what the digits represent (text, audio, video, etc.). Thus, although such additional knowledge about a data source can be quite useful toward its compression (for example, some voice compression techniques are based on the physical structure of the human vocal tract), no use will be made here of any special character of the source. 

So, we get 2 1 compression by lowering the sample rate (44 kHz > 22 kHz) and another 11 1 compression from the MP3 codec. Where does this major 11 1 compression come from This is the heart of any compression format and is where the differences between the various formats will emerge. It is also where audible artifacts creep into the signal. Claiming that the audio is equivalent to FM radio is a bit disingenuous, since a broadcast FM radio signal is not digitally compressed. This is not to pick on the MP3 format—all compressed formats claim compression that is equivalent to CD-quality or FM-quality and so on, but these references refer only to the sample rate, bit depth, and the arithmetic—not to the actual what-you-hear quality. 

---