|
|
|
| United States Patent | 5375189 |
| Link to this page | http://www.wikipatents.com/5375189.html |
| Inventor(s) | Tsutsui; Kyoya (Kanagawa, JP) |
| Abstract | A method for compressing a digital audio input signal to provide a
recording signal in which the input signal is divided into frames
comprising plural samples. Each frame is transformed into a block of
spectral coefficients, which are divided into plural bands. For each band,
block floating is applied to the spectral coefficients therein; a block
floating coefficient is generated; the spectral coefficients are quantized
with an adaptive number of bits to provide quantized spectral
coefficients; and a word length is generated. Finally, there is added to
the recording signal a block of data derived from the block of spectral
coefficients and consisting of the quantized spectral coefficients, a main
word length and a main block floating coefficient for each band, and a
reserve word length at least for each lower frequency band. In a
variation, zero bits are allocated to quantize the spectral coefficients
in bands higher in frequency than a highest usable band. The block of data
consists of the quantized spectral coefficients, at least one word length,
and at least one block floating coefficient, for each band up to the
highest usable band; and data indicating the number of bands up to the
highest usable band. |
|
|
 |
Title Information  |
|
|
|
|
|
|
| Publication Date |
December 20, 1994 |
|
|
|
|
|
| Filing Date |
September 29, 1992 |
|
|
|
|
|
|
|
|
|
|
|
|
|
| Priority Data |
Sep 30, 1991[JP]3-276168
Sep 30, 1991[JP]3-276169 |
|
|
|
|
|
|
|
|
 |
|
 |
Title Information |
|
|
|
References |
|
|
| *references marked with an asterisk below are user-added references |
 |
U.S. References |
|
|
|
 |
|
|
|
U.S. References |
|
|
|
Foreign References |
|
|
|
|
|
|
|
|
Foreign References |
|
|
|
Other References |
|
|
|
|
|
|
|
|
Other References |
|
|
|
|
|
|
|
References |
|
|
|
|
|
|
| Market Size |
|
Estimate the gross annual revenues of the relevant market
sector:
|
| | |
| |
|
|
| Market Share |
|
Estimate the percentage of the relevant market sector this invention will capture:
|
| | |
| |
|
|
| Reasonable Royalty |
|
What percentage of gross sales should the inventor or assignee be paid?
|
| | |
| |
|
|
|
Public's "Guesstimation" of Royalty Value
|
| Market Size | N/A | [No votes] | | x | Market Share | N/A | [No votes] | | x | Reasonable Royalty | N/A | [No votes] |
| | N/A | |
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
Market Review |
|
|
|
Technical Review |
|
|
|
Claims |
|
|
I claim:
1. A method for compressing a digital audio input signal to provide a
recording signal, the method comprising the steps of:
dividing the input signal into frames comprising plural samples;
transforming each frame of plural samples into a block of spectral
coefficients and dividing the block of spectral coefficients into plural
bands, the plural bands including lower frequency bands, and a lowest
frequency band;
applying block floating to the spectral coefficients in each band and
generating a block floating coefficient for each band;
quantizing the spectral coefficients in each band with an adaptive number
of bits to provide quantized spectral coefficients in each band, and
generating a word length for each band;
adding a block of data derived from the block of spectral coefficients to
the recording signal, the block of data derived from the block of spectral
coefficients consisting of:
the quantized spectral coefficients,
a main word length for each band,
a main block floating coefficient for each band, and
a reserve word length at least for each of the lower frequency bands.
2. The method for compressing a digital audio input signal of claim 1,
wherein, the step of quantizing the spectral coefficients in each band
with an adaptive number of bits includes quantizing the spectral
coefficients in each band using an additional number of bits, the
additional number of bits being a number of bits equivalent to a sum of a
first difference and a second difference,
the first difference being a difference between a number of bits required
to provide a reserve word length for each band and a number of bits
required to provide a reserve word length at least for each of the lower
frequency bands, and
the second difference is a difference between a number of bits required to
provide a reserve block floating coefficient for each band and a number of
bits required to provide no reserve block floating coefficients.
3. The method for compressing a digital audio input signal of claim 1,
wherein, in the step of adding a block of data derived from the block of
spectral coefficients to the recording signal, the block of data derived
from the block of spectral coefficients additionally consists of a reserve
block floating coefficient for each of the lower frequency bands.
4. The method for compressing a digital audio input signal of claim 3,
wherein
the step of quantizing the spectral coefficients in each band with an
adaptive number of bits includes quantizing the spectral coefficients in
each band using an additional number of bits, the additional number of
bits being a number of bits equivalent to a sum of a first difference and
a second difference,
the first difference being a difference between a number of bits required
to provide a reserve word length for each band and a number of bits
required to provide a reserve word length for each of the lower frequency
bands, and
the second difference is a difference between a number of bits required to
provide a reserve block floating coefficient for each band and a number of
bits required to provide a reserve block floating coefficient for each of
the lower frequency bands.
5. The method for compressing a digital audio input signal of claim 1,
wherein the step of adding a block of data derived from the block of
spectral coefficients to the recording signal includes the step of
arranging the quantized spectral coefficients sequentially in the block of
data derived from the block of spectral coefficients, beginning with the
quantized spectral coefficients in the lowest frequency band.
6. The method for compressing a digital audio input signal of claim 5,
wherein, the step of quantizing the spectral coefficients in each band
with an adaptive number of bits includes quantizing the spectral
coefficients in each band using an additional number of bits, the
additional number of bits being a number of bits equivalent to a sum of a
first difference and a second difference,
the first difference being a difference between a number of bits required
to provide a reserve word length for each band and a number of bits
required to provide a reserve word length for each of the lower frequency
bands, and
the second difference is a difference between a number of bits required to
provide a reserve block floating coefficient for each band and a number of
bits required to provide no reserve block floating coefficients.
7. The method for compressing a digital audio input signal of claim 5,
wherein, in the step of adding a block of data derived from the spectral
coefficients to the recording signal, the block of data derived from the
block of spectral coefficients consists of:
the quantized spectral coefficients,
a main word length for each band,
a main block floating coefficient for each band,
a reserve word length for each of a first number of the lower frequency
bands, and
a reserve block floating coefficient for each of a second number of lower
frequency bands, the second number of lower frequency bands being less
than the first number of lower frequency bands.
8. The method for compressing a digital audio input signal of claim 7,
wherein,
the step of quantizing the spectral coefficients in each band with an
adaptive number of bits includes quantizing the spectral coefficients in
each band using an additional number of bits, the additional number of
bits being a number of bits equivalent to a sum of a first difference and
a second difference,
the first difference being a difference between a number of bits required
to provide a reserve word length for each band and a number of bits
required to provide a reserve word length for each of the first number of
lower frequency bands, and
the second difference is a difference between a number of bits required to
provide a reserve block floating coefficient for each band and a number of
bits required to provide a reserve block floating coefficient for each of
the second number of lower frequency bands.
9. The method for compressing a digital audio input signal of claim 1,
wherein, in the step of adding a block of data derived from the block of
spectral coefficients to the recording signal, the block of data derived
from the block of spectral coefficients consists of:
the quantized spectral coefficients,
a main word length for each band,
a main block floating coefficient for each band,
a reserve word length for each band, and
a reserve block floating coefficient for each of the lower frequency bands.
10. The method for compressing a digital audio input signal of claim 9,
wherein
the step of quantizing the spectral coefficients in each band with an
adaptive number of bits includes quantizing the spectral coefficients in
each band using an additional number of bits, the additional number of
bits being a number of bits equivalent to a difference between a number of
bits required to provide a reserve block floating coefficient for each
band and a number of bits required to provide a reserve block floating
coefficient for each of the lower frequency bands.
11. A method for compressing a digital audio input signal to provide a
recording signal, the method comprising the steps of:
dividing the input signal into frames comprising plural samples;
transforming each frame of plural samples into a block of spectral
coefficients and dividing the block of spectral coefficients into plural
bands, the plural bands including a lowest frequency band, and a highest
frequency band;
generating block floating parameters;
applying block floating to the spectral coefficients in each band in
response to a block floating parameter;
quantizing the spectral coefficients in each band with an adaptive number
of bits to provide quantized spectral coefficients in response to a block
floating parameter, zero bits being allocated to the spectral coefficients
in bands higher in frequency than a highest usable band;
adding a block of data derived from the block of spectral coefficients to
the recording signal, the block of data derived from the block of spectral
coefficients consisting of:
the quantized spectral coefficients for each band up to the highest usable
band, there being a number of bands up to the highest useable band,
block floating parameters for each band up to the highest useable band, and
data indicating the number of bands up to the highest usable band.
12. The method for compressing a digital audio input signal of claim 11,
wherein, the step of quantizing the spectral coefficients in each band
with an adaptive number of bits includes quantizing the spectral
coefficients in each band using an additional number of bits, the
additional number of bits being a number of bits equivalent to a
difference between a number of bits required to provide block floating
parameters for each band and a number of bits required to provide block
floating parameters for each band up to the highest useable band.
13. The method for compressing a digital audio input signal of claim 11,
wherein
the block floating parameters include a word length and a block floating
coefficient,
the step of applying block floating includes the step of applying block
floating in response to the block floating coefficient,
the step of quantizing the spectral coefficients includes the step of
quantizing the spectral coefficients in response to the word length, and
in the step of adding a block of data derived from the block of spectral
coefficients to the recording signal, the block floating parameters in
block of data derived from the block of spectral coefficients consist of:
a main word length for each band up to the highest useable band,
a main block floating coefficient for each band up to the highest useable
band, and
a reserve word length for each of the lower frequency bands.
14. The method for compressing a digital audio input signal of claim 13,
wherein in the step of adding a block of data derived from the block of
spectral coefficients to the recording signal, the block floating
parameters in block of data derived from the block of spectral
coefficients additionally consist of a reserve block floating coefficient
for each of the lower frequency bands.
15. The method for compressing a digital audio input signal of claim 13,
wherein the step of adding a block of data derived from the block of
spectral coefficients to the recording signal includes the step of
arranging the quantized spectral coefficients sequentially in the block of
data derived from the block of spectral coefficients, beginning with the
quantized spectral coefficients in the lowest frequency band.
16. The method for compressing a digital audio input signal of claim 15,
wherein, in the step of adding a block of data derived from the block of
spectral coefficients to the recording signal, the block floating
parameters in block of data derived from the block of spectral
coefficients consist of:
a main word length for each band up to the highest useable band,
a main block floating coefficient for each band up to the highest useable
band,
a reserve word length for each of a first number of the lower frequency
bands, and
a reserve block floating coefficient for each of second number of lower
frequency bands, the second number of lower frequency bands being less
than the first number of lower frequency bands.
17. An apparatus for compressing a digital audio input signal to provide a
recording signal, the apparatus comprising:
a means for dividing the input signal into frames comprising plural
samples;
a means for transforming each frame of plural samples into a block of
spectral coefficients and for dividing the block of spectral coefficients
into plural bands, the plural bands including lower frequency bands;
a means for applying block floating to the spectral coefficients in each
band and for generating a block floating coefficient for each band;
a quantizing means for quantizing the spectral coefficients in each band
with an adaptive number of bits to provide quantized spectral coefficients
in each band, and for generating a word length for each band;
a means for adding a block of data derived from the block of spectral
coefficients to the recording signal, the block of data derived from the
block of spectral coefficients consisting of:
the quantized spectral coefficients,
a main word length for each band,
a main block floating coefficient for each band, and
a reserve word length at least for each of the lower frequency bands.
18. The apparatus for compressing a digital audio input signal of claim 17,
wherein, the quantizing means is for quantizing the spectral coefficients
in each band with an adaptive number of bits using an additional number of
bits, the additional number of bits being a number of bits equivalent to a
sum of a first difference and a second difference,
the first difference being a difference between a number of bits required
to provide a reserve word length for each band and a number of bits
required to provide a reserve word length for each of the lower frequency
bands, and
the second difference is a difference between a number of bits required to
provide a reserve block floating coefficient for each band and a number of
bits required to provide no reserve block floating coefficients.
19. The apparatus for compressing a digital audio input signal of claim 17,
wherein, the block of data derived from the block of spectral coefficients
additionally consists of a reserve block floating coefficient for each of
the lower frequency bands.
20. The apparatus for compressing a digital audio input signal of claim 19,
wherein, the quantizing means is for quantizing the spectral coefficients
in each band with an adaptive number of bits using an additional number of
bits, the additional number of bits being a number of bits equivalent to a
sum of a first difference and a second difference,
the first difference being a difference between a number of bits required
to provide a reserve word length for each band and a number of bits
required to provide a reserve word length for each of the lower frequency
bands, and
the second difference is a difference between a number of bits required to
provide a reserve block floating coefficient for each band and a number of
bits required to provide a reserve block floating coefficient for each of
the lower frequency bands.
21. The apparatus for compressing a digital audio input signal of claim 17,
wherein the adding means includes a means for arranging the quantized
spectral coefficients sequentially in the block of data derived from the
block of spectral coefficients, beginning with the quantized spectral
coefficients in the lowest frequency band.
22. The apparatus for compressing a digital audio input signal of claim 21,
wherein,
the quantizing means is for quantizing the spectral coefficients in each
band with an adaptive number of bits using an additional number of bits,
the additional number of bits being a number of bits equivalent to a sum
of a first difference and a second difference,
the first difference being a difference between a number of bits required
to provide a reserve word length for each band and a number of bits
required to provide a reserve word length for each of the lower frequency
bands, and
the second difference is a difference between a number of bits required to
provide a reserve block floating coefficient for each band and a number of
bits required to provide no reserve block floating coefficients.
23. The apparatus for compressing a digital audio input signal of claim 21,
wherein the block of data derived from the block of spectral coefficients
consists of:
the quantized spectral coefficients,
a main word length for each band,
a main block floating coefficient for each band,
a reserve word length for each of a first number of the lower frequency
bands, and
a reserve block floating coefficient for each of a second number of lower
frequency bands, the second number of lower frequency bands being less
than the first number of lower frequency bands.
24. The apparatus for compressing a digital audio input signal of claim 23,
wherein, the quantizing means is for quantizing the spectral coefficients
in each band with an adaptive number of bits using an additional number of
bits, the additional number of bits being a number of bits equivalent to a
sum of a first difference and a second difference,
the first difference being a difference between a number of bits required
to provide a reserve word length for each band and a number of bits
required to provide a reserve word length for each of the first number of
lower frequency bands, and
the second difference is a difference between a number of bits required to
provide a reserve block floating coefficient for each band and a number of
bits required to provide a reserve block floating coefficient for each of
the second number of lower frequency bands.
25. The apparatus for compressing a digital audio input signal of claim 17,
wherein the block of data derived from the block of spectral coefficients
consists of:
the quantized spectral coefficients,
a main word length for each band,
a main block floating coefficient for each band,
a reserve word length for each band, and
a reserve block floating coefficient for each of the lower frequency bands.
26. The apparatus for compressing a digital audio input signal of claim 25,
wherein the quantizing means quantizes the spectral coefficients in each
band with an adaptive number of bits using an additional number of bits,
the additional number of bits being a number of bits equivalent to a
difference between a number of bits required to provide a reserve block
floating coefficient for each band and a number of bits required to
provide a reserve block floating coefficient for each of the lower
frequency bands.
27. An apparatus for compressing a digital audio input signal to provide a
recording signal, the apparatus comprising:
a means for dividing the input signal into frames comprising plural
samples;
a means for transforming each frame of plural samples into a block of
spectral coefficients and for dividing the block of spectral coefficients
into plural bands, the plural bands including lower frequency bands, and a
lowest frequency band;
a means for generating block floating parameters;
a block floating means for applying block floating to the spectral
coefficients in each band in response to a block floating parameter;
a quantizing means for quantizing the spectral coefficients in each band
with an adaptive number of bits to provide quantized spectral coefficients
in response to a block floating parameter, the quantizing means allocating
zero bits to the spectral coefficients in bands higher in frequency than a
highest usable band;
a means for adding a block of data derived from the block of spectral
coefficients to the recording signal, the block of data derived from the
block of spectral coefficients consisting of:
the quantized spectral coefficients for each band up to the highest usable
band, there being a number of bands up to the highest useable band,
block floating parameters for each band up to the highest useable band, and
data indicating the number of bands up to the highest usable band.
28. The apparatus for compressing a digital audio input signal of claim 27,
wherein the quantizing means is for quantizing the spectral coefficients
in each band with an adaptive number of bits using an additional number of
bits, the additional number of bits being a number of bits equivalent a
difference between a number of bits required to provide block floating
parameters for each band and a number of bits required to provide block
floating parameters for each band up to the highest useable band.
29. The apparatus for compressing a digital audio input signal of claim 27,
wherein
the block floating parameters include a word length and a block floating
coefficient,
the block floating means applies block floating in response to the block
floating coefficient,
the quantizing means quantizes the spectral coefficients in response to the
word length, and
the block floating parameters in block of data derived from the block of
spectral coefficients consist of:
a main word length for each band up to the highest useable band,
a main block floating coefficient for each band up to the highest useable
band, and
a reserve word length for each of the lower frequency bands.
30. The apparatus for compressing a digital audio input signal of claim 29,
wherein the block floating parameters in block of data derived from the
block of spectral coefficients additionally consist of a reserve block
floating coefficient for each of the lower frequency bands.
31. The apparatus for compressing a digital audio input signal of claim 29,
wherein the adding means includes a means for arranging the quantized
spectral coefficients sequentially in the block of data derived from the
spectral coefficients, beginning with the quantized spectral coefficients
in the lowest frequency band.
32. The apparatus of claim 31, wherein the block floating parameters in
block of data derived from the block of spectral coefficients consist of:
a main word length for each band up to the highest useable band,
a main block floating coefficient for each band up to the highest useable
band,
a reserve word length for each of a first number of lower frequency bands
a reserve block floating coefficient for each of a second number of the
lower frequency bands, the second number of the lower frequency bands
being less than the first number of the lower frequency bands. |
|
|
|
|
|
|
Claims |
|
|
|
Description |
|
|
FIELD OF THE INVENTION
This invention relates to a method for processing audio signals by block
floating and compression.
BACKGROUND OF THE INVENTION
As a high efficiency coding technique for compressing a digital audio
signal, it is known to divide digital audio input signal in time into
plural frames of a predetermined samples, to transform each frame into
spectral coefficients in the frequency domain, and to divide the block of
spectral coefficients resulting from transforming a frame into plural
frequency bands. The spectral coefficients in each band are processed by
block floating and are quantized by adaptive bit allocation.
Block floating is a normalization process applied to a block of data
comprising plural words, such as a band of spectral coefficients. Block
floating is applied by multiplying each word in the data block by a common
value for the data block to improve quantization efficiency. In a typical
block floating process, the maximum absolute value of the words in the
data block is found and is used as a block floating coefficient common to
all the words in the data block. Using the maximum absolute value in the
band as the block floating coefficient prevents data overflow because the
absolute value of no other word in the data block can be greater than the
maximum absolute value. A simplified form of block floating determines the
block floating coefficient using a shift quantity, which provides block
floating in 6 dB steps.
The data compressor employing block floating generates, for each band,
various block floating parameters BF that are transmitted or recorded on a
recording medium together with the quantized spectral coefficients, or
main information. The block floating parameters include a block floating
coefficient SF and a word length WL, which provides information concerning
the adaptive bit allocation, indicating the difference between the value
of the block floating coefficient SF and the allowable noise level which
is determined for each band, taking account of masking.
In the following description, reference will be made to recording on or
reproducing from a recording medium. When such references are made, they
are to be understood as additionally encompassing transmitting to and
receiving from a transmission medium.
Masking is a psychoacoustic phenomenon in which a sound is rendered
inaudible, or "masked," by other sounds occurring simultaneously with, or
slightly earlier than, or later than, the sound. Masking effects may be
classed into time domain masking effects, that is, masking by sounds
occurring earlier or later than the masked sound, and concurrent masking
effects, which is masking is by simultaneously-occurring sounds having a
frequency different from the frequency of the masked sound.
Masking enables a sound to render inaudible any noise within its time or
frequency masking range. This means that in the presence of a signal that,
when reproduced, produces a sound, a digital encoding system that produces
quantizing noise may have quantizing noise levels that are high compared
with the noise level that is allowable in the absence of the signal,
provided that the quantizing noise lies within the masking range of the
sound produced by the signal. Since relatively high levels of quantizing
noise are allowable if masked by the sound resulting from the signal, the
number of bits required to quantize the signal representing the sound, or
parts of the signal, may be significantly reduced.
A critical band is a frequency band that takes advantage of the masking
characteristics of the human sense of hearing. A critical band is the band
of noise that can be masked by a pure sound that has the same intensity as
the noise and has a frequency in the vicinity of the frequency of the
noise. The width of the critical band increases with increasing frequency
of the pure sound. The entire audio frequency range of 0 Hz to 20 kHz can
be divided into, for example, 25 critical bands.
If, for some reason, data is destroyed or lost between the output of the
data compressor and the input of a complementary data expander, it is
possible to reduce the audible effects of the missing data in the data
expander by reducing the signal component in the frequency band
corresponding to the missing data to zero.
However, since the block floating parameters BF is related to the spectral
coefficients in each band, the effect on sound quality of losing a block
floating parameter BF is more noticeable than loss of the main information
(i.e., the quantized spectrum signals).
To mitigate the effects of possibly losing a block floating parameter BF,
it is known to include the block floating parameters BF in the compressed
signal provided by the data compressor twice, so that they are recorded
twice on the recording medium. This provides a redundant set of block
floating parameters in case a block floating parameter is lost or
erroneous.
In the following description, reference will be made to lost data, such as
block floating parameters and quantized spectral coefficients on the
understanding that this term also covers erroneous or corrupted data, such
as block floating parameters and quantized spectral coefficients.
As shown in FIG. 15, the quantized spectral coefficients (main information)
are recorded together with the block floating coefficients SF and word
length WL as the above-mentioned block floating parameters BF, and are
recorded as the block floating coefficients SF1 and the word lengths WL1,
respectively, and are recorded a second time as the block floating
coefficients SF2 and the word lengths WL2, respectively.
With the above method, since all the block floating parameters BF need to
be recorded twice to deal with normal occurrences of data loss, the number
of bits allocated to the main information must be reduced to accommodate
the additional block floating parameters. Consequently, in systems having
a high compression ratio or a low bit rate, a satisfactory sound quality
cannot be attained.
In a conventional data compressor, the number of block floating parameters
BF recorded per frame of the input signal is usually fixed. FIG. 16 shows
how the data corresponding to each frame of the input signal is arranged
in the recording signal produced by a conventional data compressor. In the
example shown, the values of the block floating parameters BF for the
bands to which no bits are allocated must still to be recorded, which
reduces the number of bits available for coding the spectral coefficients
in the main information. This makes it difficult to achieve a satisfactory
sound quality when the compressed signal from the compressor is subject to
complementary expansion, and is reproduced. This is especially so in
systems having a high compression ratio or a low bit rate.
The system shown in FIG. 17 is also known. In this, no block floating
coefficients SF are recorded for those bands to which no bits are actually
allocated, i.e., for bands having a word length WL=0. Correspondingly more
bits are available for allocation to coding the spectral coefficients. In
the example of FIG. 17, the number of recorded block floating coefficients
SF is reduced by four, which is the number of bands to which no bits are
allocated. In the arrangement shown in FIG. 17, it is still necessary to
record the word length WL for all bands, and to determine whether the word
length WL of each band is not zero when reading the block floating
coefficients SF in the expander.
It is also necessary for the data compressor to calculate the number of
bits necessary for quantizing the spectral coefficients in each band by a
process that determines masking. The number of bits thus calculated is
compared with the total number of bits allocated to the frame, after
which, the bit allocation to each band may be adjusted. However, if the
change in bit allocation changes whether the block floating coefficient SF
of a block is recorded or not, the total number of bits that is allocated
to the main information is also changed, which complicates the process of
adjusting the bit allocation.
OBJECTS AND SUMMARY OF THE INVENTION
In view of the above-described state of the art, it is an object of the
present invention to provide a method for processing an audio signal in
which more bits can be allocated to quantized spectral coefficients and
which is resistant to data loss.
In view of the above-described state of the art, it is another object of
the present invention to provide a method for processing an audio signal
wherein the bit allocation may be easily adjusted and wherein the sound
quality is not impaired by adjusting the bit allocation.
For accomplishing the above object, the present invention provides a method
for compressing a digital audio input signal to provide a recording
signal. According to the method, the input signal is divided into frames
comprising plural samples. Each frame of plural samples is transformed
into a block of spectral coefficients. The block of spectral coefficients
is divided into plural bands that include lower frequency bands, and a
lowest frequency band. Block floating is applied to the spectral
coefficients in each band and a block floating coefficient is generated
for each band. The spectral coefficients in each band are quantized with
an adaptive number of bits to provide quantized spectral coefficients in
each band, and a word length is generated for each band. Finally, a block
of data derived from the block of spectral coefficients is added to the
recording signal. The block of data derived from the block of spectral
coefficients consists of the quantized spectral coefficients, a main word
length for each band, a main block floating coefficient for each band, and
a reserve word length at least for each of the lower frequency bands.
In a first variation, there is a reserve word length for each band in the
block of data derived from the block of spectral coefficients.
In a second variation, the block of data derived from the spectral
coefficients that is added to the output signal additionally consists of a
reserve block floating coefficient for each of the lower frequency bands.
In a third variation, the method additionally comprises arranging the
quantized spectral coefficients sequentially in the block of data derived
from the block of spectral coefficients, beginning with the quantized
spectral coefficients in the lowest frequency band.
With the arrangement just described, when the block of data derived from
the spectral coefficients is added to the recording signal, the block of
data derived from the block of spectral coefficients consists of the
quantized spectral coefficients, a main word length for each band, a main
block floating coefficient for each band, a reserve word length for each
of a first number of the lower frequency bands, and a reserve block
floating coefficient for each of a second number of lower frequency bands.
The second number of lower frequency bands is less than the first number
of lower frequency bands.
With the method for processing audio signals according to the present
invention, the block floating coefficients for higher frequency bands,
which are less critical to the human sense of hearing, are not included in
the recording signal twice, which increases the number of bits available
for quantizing the spectral coefficients.
Also, only the word lengths for lower frequency bands are recorded twice,
so that bits that otherwise would be allocated to the floating
coefficients may be allocated to quantizing the spectral coefficients.
Further, by arranging the quantized spectral coefficients in the recording
signal beginning with lower frequency spectral coefficients, the sound
quality impairment resulting from the loss of higher frequency spectral
coefficients is reduced because of masking by the lower frequency spectral
coefficients.
For further accomplishing the above objects, the present invention
additionally provides a method for compressing a digital audio input
signal to provide a recording signal. According to the method, the input
signal is divided into frames comprising plural samples. Each frame of
plural samples is transformed into a block of spectral coefficients. The
block of spectral coefficients is divided into plural bands. The plural
bands include a lowest frequency band, and a highest frequency band. Block
floating parameters are generated. Block floating is applied to the
spectral coefficients in each band in response to a block floating
parameter. The spectral coefficients in each band are quantized with an
adaptive number of bits to provide quantized spectral coefficients in
response to a block floating parameter. Zero bits are allocated to the
spectral coefficients in bands higher in frequency than a highest usable
band. A block of data derived from the block of spectral coefficients is
added to the recording signal. The block of data derived from the block of
spectral coefficients consists of the quantized spectral coefficients for
each band up to the highest usable band, block floating parameters for
each band up to the highest useable band, and data indicating the number
of bands up to the highest usable band.
If the higher frequency spectral coefficients are not included in the
recording signal because these signals make no perceptible contribution to
the reproduced audio signal, the block floating parameters for the high
frequency bands of the frame, i.e., the block floating coefficient and the
word length, are not included in the recording signal. The bits thus saved
are allocated to the main information at lower frequencies, which is
crucial to the human sense of hearing.
When block floating parameters are omitted from the recording signal, data
is included in the recording signal indicating the number of block
floating parameters in the recording signal. The number of block floating
parameters corresponds to the number of bands up to the highest useable
band.
According to the audio data compression method of the present invention,
the block parameters for each band are included in the recording signal in
each frame for those bands that need such parameters, i.e., for the bands
up to the maximum useable band. Block floating parameters for the bands
for which the block floating parameters are unnecessary, i.e., for bands
above the maximum useable band are omitted from the recording signal, and
the bits thus saved are allocated for the coding the lower frequency
spectral coefficients in the main information.
The invention also encompasses an apparatus to which the methods of the
invention are applied.
BRIEF DESCRIPTION OF THE DRAWINGS
FIG. 1 is a diagrammatic view for illustrating the recording signal
produced by the method according to a first aspect of the present
invention.
FIG. 2 is a diagrammatic view for illustrating the recording signal
produced by the method according to second and third aspects of the
present invention.
FIG. 3 is a block circuit diagram for illustrating an arrangement of a data
compressor to which the methods according to the present invention are
applied.
FIG. 4 is a block circuit diagram showing a practical arrangement of the
orthogonal transform circuit of the data compressor.
FIG. 5 is a block circuit diagram showing an arrangement of a complementary
data expander.
FIG. 6 is a block circuit diagram showing a practical arrangement of the
orthogonal transform circuit of the expander.
FIG. 7 is a flow chart for illustrating the block floating coefficient
reading sequence in the block floating coefficient reading circuit in an
expander according to the first to sixth aspects of the invention.
FIG. 8 is a flow chart for illustrating the word length reading sequence in
the word length reading circuit in an expander according to the first
aspect of the invention.
FIG. 9 is a flow chart for illustrating the word length reading sequence in
the word length reading circuit in an expander according to the second and
sixth aspects of the invention.
FIG. 10 is a flow chart for illustrating the quantized spectral coefficient
reading sequence in the quantized spectral coefficient reading circuit in
an expander according to the first through sixth aspects of the invention.
FIG. 11 is a diagrammatic view for illustrating data recording according to
a fifth aspect of the present invention, in which higher frequency
spectral coefficients are not recorded.
FIG. 12 is a diagrammatic view for illustrating data recording according to
a sixth aspect of the present invention, in which higher frequency
spectral coefficients are recorded.
FIG. 13 is a flow chart for illustrating the processing by the data
compressor.
FIG. 14 is a flow chart for illustrating the processing by the data
expander.
FIG. 15 is a diagrammatic view for illustrating the recording signal
produced by a conventional data compressor.
FIG. 16 is a diagrammatic view for illustrating the recording signal
produced by a conventional system in which the number of block floating
parameters is constant.
FIG. 17 is a diagrammatic view for illustrating the recording signal
produced by a conventional system in which the number of block floating
coefficients is variable.
DETAILED DESCRIPTION OF THE INVENTION
The preferred embodiments of the present invention will now be described in
detail with reference to the drawings.
In a first aspect, the present invention provides a method for compressing
a digital audio input signal comprising transforming a frame of the input
audio signal TS in the time domain into plural spectral coefficients SP in
the frequency domain. The spectral coefficients are divided into plural
frequency bands, block floating is applied to each band, and the spectral
coefficients in each band are quantized by adaptive bit allocation. The
quantized spectral coefficients QSP, the block floating coefficient SF1,
and the word length WL1, for all the bands are recorded once. In addition,
the block floating coefficients SF2 for the lower frequency bands and the
word lengths WL2 for all the bands are recorded a second time.
FIG. 1 shows how one block of data in the frequency domain resulting from
transforming one frame of the input signal is recorded. The main
information shown in FIG. 1 is all the quantized spectral coefficients in
the block.
Of the block floating parameters BF, the word lengths WL1 and WL2 for all
the bands in the block are recorded twice, the block floating coefficients
SF1 for all the bands in the block are recorded once, and the block
floating coefficients SF2 for only the lower frequency bands in the block
are recorded a second time.
The advantage of recording the block floating coefficients SF2 for only the
lower frequency bands a second time will now be explained. Of the block
floating parameters BF, the word length WL represents the difference
between the block floating coefficient SF and the allowable noise level
determined for each band, taking account of masking. The word length
indicates information concerning adaptive bit allocation for quantizing
the spectral coefficients SP, i.e., the number of bits used to quantize
the spectral coefficients in the recorded signal. As a result, if only one
of the word lengths WL for the block is lost, none of the quantized
spectral coefficients in the block following the quantized spectral
coefficients QSP corresponding to the lost word length WL can be read. On
the other hand, if one of the block floating coefficients SF for a band of
spectral coefficients is lost, only the spectral coefficients SP in the
band corresponding to the erroneous or lost block floating coefficient SF
cannot be restored. Consequently, the impairment of sound quality
resulting from a lost block floating coefficient is less than that
resulting from the loss of a word length.
Further, as far as a human listener is concerned, lower frequency audio
signals represented by the lower frequency spectral coefficients SP,
effectively mask higher frequency signals represented by higher frequency
spectral coefficients SP. Consequently, the loss of higher frequency
spectral coefficients SP has a minimal effect on sound quality.
Because of this, in the audio signal processing method according to the
first aspect of the present invention, impairment of the sound quality of
the expanded, decoded and reproduced signal due to lost data can be
minimized if, of the block floating parameters BF, the block floating
coefficients of only the lower frequency bands are recorded twice. An
overall improvement of sound quality can be achieved if the bits thus
saved are allocated for quantizing the spectral coefficients more
accurately.
In the audio signal processing method according to the second aspect of the
present invention, the quantized spectral coefficients QSP, are
sequentially recorded beginning with the quantized spectral coefficients
in the lowest frequency band, as shown by the arrows in FIG. 2. Of the
block floating parameters BF, the word length WL1 and block floating
coefficient SF1 for each band are recorded once. Additionally, the word
lengths WL2 for the lower frequency bands only are recorded a second time.
The block floating coefficients are not recorded a second time. Only the
block floating coefficient SF1 is recorded.
The third aspect of the present invention is a variation on the second
aspect. In the third aspect, the block floating coefficients SF2 for the
lower frequency bands are recorded a second time, as shown in FIG. 2. In
addition, the number of bands for which the word length WL is recorded
twice is set to be larger than the number of bands for which the block
floating coefficient SF is recorded twice.
The advantages of the methods according to the second and third aspects of
the invention will now be explained. As in the first aspect, as far as a
human listener is concerned, lower frequency audio signals represented by
the lower frequency spectral coefficients SP, effectively mask higher
frequency signals represented by higher frequency spectral coefficients
SP. Consequently if higher frequency spectral coefficients SP are lost,
the impairment of sound quality is small.
Of the block floating parameters BF, the word length WL represents the
difference between the block floating coefficient SF and the allowable
noise level determined for each band taking account of masking. The word
length indicates information concerning adaptive bit allocation for
quantizing the spectral coefficients SP, i.e., the number of bits used to
quantize the spectral coefficients in the recorded signal. As a result, if
only one of the word lengths WL for the block is lost, none of the
quantized spectral coefficients in the recorded signal following the
spectral coefficients SP corresponding to the lost word length WL can be
read. However, if the quantized spectral coefficients QSP are recorded
sequentially, beginning with the lowest frequency band, as in the second
aspect of the invention, the quantized spectral coefficients QSP can be
correctly read in the decoder up to the frequency band corresponding to
the lost word length WL. The correctly expanded lower frequency spectral
coefficients mask the defects in the expanded signal resulting from the
unexpanded higher frequency spectral coefficients.
If a block floating coefficient SF is destroyed, only the spectral
coefficients in the band corresponding to the lost block floating
coefficient can:not be restored, so that impairment of sound quality is
less than if the word length WL of the band is destroyed.
It is seen from above that, in the second embodiment, in which the
quantized spectral coefficients QSP are sequentially recorded, beginning
with the lowest frequency band, with reg | | |
