audio_hdr.3

操作系统SunOS 4.1.3版本的源码

.\" @(#)audio_hdr.3 1.1 92/07/30 SMI
.ds Dc 7/30/92
.TH AUDIO_HDR 3 "\*(Dc" "" "Audio Library"
.SH NAME
audio_hdr
\- audio header structure
.SH DISCLAIMER
.LP
The manual pages, header files, and object library associated with
.B /usr/demo/SOUND/libaudio.a
are furnished on an AS IS basis as a preliminary interface to several
useful audio data-processing capabilities.  A future release of the
operating system may redefine the syntax or semantics of some of the
functions described herein.
.SH SYNOPSIS
.nf
.B #include <multimedia/audio_hdr.h>
.fi
.SH OVERVIEW
.LP
Digital audio data represents a quantized approximation of an analog
audio signal, which itself is directly proportional to the rapid changes
of air pressure that we perceive as sound.
Pulse Code Modulation (\s-1PCM\s0) refers to the sampling, at discrete
intervals, of an audio waveform to produce a (usually linear) quantization.
In order to achieve the best approximation of a signal,
the highest possible sampling frequency and precision should be used.
However, it is often impractical to process and store the huge amounts
of data that such a practice yields.
.LP
In lowering the overall data rate, a number of compromises must be
made.  To begin with, the sampling frequency is generally reduced to a level
commensurate with the quality of the audio signal.  For speech and telephony,
8, 12, and 16 kHz sampling rates are typically used.  For high-fidelity
audio, 44.1 and 48 kHz sampling rates are predominant.
.LP
Sampling precision may also be reduced to yield lower data rates.
Various encoding schemes further compress the signal.  Some of these,
such as \(*m-law and A-law companding, map the linear
.SM PCM
domain into a pseudo-logarithmic domain.  Like
.SM PCM
encoding, these schemes maintain a one-to-one relationship between
individual samples and waveform quantizations.
Other encoding schemes, such as Differential and Adaptive Differential
.SM PCM\s0,
convolve several samples into a single
.I sound unit
that must be treated as an atomic entity.
.LP
Multiple channels of audio data may be combined into a single audio stream
by interleaving individual sound units from each channel.  For each
quantized period of time, the set consisting of one sound unit from each
audio channel is known as a
.IR "sample frame" .
.SH AUDIO HEADER
.LP
For programming purposes, the encoding of audio data is described by an in-core
.B Audio_hdr
structure.  This audio header is defined in
.B <multimedia/audio_hdr.h>
as follows:
.RS
.nf
.ft B
typedef struct {
	unsigned	sample_rate;		/* samples per second */
	unsigned	samples_per_unit;	/* samples per unit */
	unsigned	bytes_per_unit;		/* bytes per sample unit */
	unsigned	channels;		/* # of interleaved channels */
	unsigned	encoding;		/* data encoding format */
	unsigned	data_size;		/* length of data (advisory) */
} Audio_hdr;
.ft P
.fi
.RE
.LP
The
.I sample_rate
field is an integer representing the sampling frequency, in samples per second,
of the audio data.
.LP
The
.I samples_per_unit
and
.I bytes_per_unit
fields describe an individual sound unit.  For
.SM PCM
and related encodings, the
.I samples_per_unit
field is 1, and
.I bytes_per_unit
gives the precision.  For encodings that group several samples together, 
.I bytes_per_unit
reflects the size of an atomic unit of audio data.
.LP
The
.I channels
field describes the number of interleaved audio channels.  The quantity
.RS
.ft B
channels \(** bytes_per_unit
.ft P
.RE
gives the size, in bytes, of each sample frame.
.LP
These four values are used to compute the byte offset given a
particular time, and vice versa.  Specifically:
.RS
.ft B
seconds = offset / C
.br
offset = seconds \(** C
.ft P
.RE
where:
.RS
.ft B
C = (channels \(** bytes_per_unit \(** sample_rate) / samples_per_unit
.ft P
.RE
.LP
The
.I encoding
field enumerates the specific data encoding.  The following values are defined:
.TP 35
.SB AUDIO_ENCODING_ULAW
ISDN \(*m-law companding
.TP
.SB AUDIO_ENCODING_ALAW
ISDN A-law companding
.TP
.SB AUDIO_ENCODING_LINEAR
linear PCM 2's-complement integer (0-center)
.TP
.SB AUDIO_ENCODING_FLOAT
linear IEEE floating-point (\(mi1. \(<= x \(<= \(pl1.)
.LP
Note that floating-point data is normalized to \(+-1.
Many signal-processing algorithms are simplified by such normalization
since the result of the multiplication of two signals is guaranteed to
lie within the acceptable range.  Thus, bounds-checking need only be performed
when converting to or from integer representations.
.LP
The
.I data_size
field contains the number of bytes in the data stream.  This value is
advisory only, as this quantity is not always known in advance.  In such
cases, the special value
.SB AUDIO_UNKNOWN_SIZE
should be used.
.SH SEE ALSO
.BR audio_convert (3),
.BR audio_filehdr (3),
.BR audio_misc (3),
.BR audio_ulaw2linear (3).