 |
Description  |
|
|
BACKGROUND OF THE INVENTION
The invention relates to high definition television (HDTV) and to spread
spectrum communication systems.
The United States is in the process of setting a standard for
high-definition television transmission in the same terrestrial
broadcasting (over-the-air) channels now used for transmitting standard
(known in the US and some other countries as NTSC) television. This plan
is quite different from the intentions in Europe and Japan, in which HDTV
will be delivered only by satellite transmission (DBS).
There are two important differences that affect the system design. DBS
provides a "clean" channel, essentially free of ghosts and interference,
and with a guaranteed minimum signal level. Terrestrial broadcasting is
typically impaired by ghosts, noise, interference, and frequency
distortion. These characteristics and impairments set the practical limit
to picture quality in typical homes, and they also affect the number of
channels that can be used in any one area (10 to 20) as compared with the
total number of channels allocated for television (about 68). In addition,
terrestrial spectrum is in very short supply. There are many more
claimants for spectrum assignments than can be accommodated by the
regulatory authorities. Consequently, these authorities are very much
concerned with the spectrum efficiency of proposed HDTV systems. For this
application, spectrum efficiency refers to the number cf different TV
programs of a given technical quality that can be provided to each viewer
within a given overall allotment of spectrum.
It is inherent in terrestrial broadcasting, due to the operation of the
inverse-square law, that the signal power level, and therefore the CNR
(carrier/noise ratio), drops rapidly as one goes further from the
transmitter. Since the theoretical capacity of the channel to deliver
information is proportional to the bandwidth times the CNR (in dB), it is
necessary to deliver more information, and hence a better picture, to
close-in receivers in order to achieve efficient utilization of channel
capacity. All current analog systems do this as a matter of course.
Systems that do not have a graceful degradation of image quality with
worsening CNR inherently waste bandwidth in the central cities, precisely
where a spectrum shortage is developing due to the rapid growth of mobile
services.
Recently, several all-digital HDTV systems have been proposed. All of them
deliver about 20 Mb/s to all viewers, and all have a very sharp threshold,
below which no reception at all is obtained. This sharp threshold is due
to the very rapid increase in bit error rate (BER) as a function of CNR.
Typically, the BER increases by a factor of 10 with a 1-dB drop in CNR. No
known scheme of error protection can overcome this effect and still
maintain good transmission efficiency.
An alternative is to use some form of progressive transmission in which the
transmitted signal is divided into a number of data streams that are
transmitted at different effective power levels, using frequency division
(FD) or time division (TD). The higher the CNR, the more data streams are
recovered, and the better the picture. While these methods are in the
right direction, they are not very efficient, since, at most levels of
CNR, one or more of the data streams is being delivered at excess CNR, and
therefore wastefully. In addition, since this class of schemes uses FD or
TD, there is a fixed upper limit of image quality, and there is no
convenient way to upgrade the system over time.
SUMMARY OF THE INVENTION
In general, in one aspect, the invention features a spread spectrum
transmission system for transmitting a television signal. The transmission
system includes means for dividing the television signal into a plurality
of frequency components; means for grouping at least some of the frequency
components into a plurality of blocks, the grouping based upon the
distance that each of the frequency components is from the origin in the
frequency plane; means for generating a plurality of spread spectrum
signals, each of the spread spectrum signals being generated from a
different one or more of the plurality of blocks; means for combining more
than one of the spread spectrum signals to form a combined signal; and a
transmitter which receives the combined signal and generates a
transmission signal therefrom.
Preferred embodiments include the following features. The dividing means,
which includes a quadraturemirror filter bank, employs subband
decomposition to divide the television signal into a plurality of spatial
frequency components. The combining means weights each of the spread
spectrum signals differently before combining them. The plurality of
frequency components includes lower frequency components and higher
frequency components and the frequency components that are grouped into
the plurality of blocks are selected only from among the higher frequency
components. The lower frequency components are transmitted using
quadrature amplitude modulation. The spread spectrum transmission system
also includes means for quantizing the lower frequency components so as to
generate quantized signals and it includes mean for superimposing the
combined signal onto the quantized signal prior to transmission.
Also in preferred embodiments, the spread spectrum transmission system
includes means for adaptively selecting samples of the plurality of
frequency components. The adaptive selection means selects samples of the
plurality of frequency components according to their importance to picture
quality. The importance to picture quality is determined based on the
amplitude of the samples. Of the samples of the plurality of frequency
components, only the selected samples are transmitted and the system
includes means for transmitting adaptive selection information that
enables a receiver to assign the selected samples to appropriate locations
in frequency, time, and space. Alternatively, the spread spectrum
transmission system includes means for setting the non-selected samples of
the plurality of spatial frequency components to zero prior to their being
combined to form the combined signal, and then both the selected and the
zero value non-selected samples are transmitted.
Also in other preferred embodiments, the spread spectrum transmission
system includes a scrambler for scrambling the order in which the samples
of the frequency components are transmitted. The system further includes
means for adaptively modulating the amplitude of the samples of the
frequency components. The adaptive modulation means increases the
amplitude of the smaller of the samples of the frequency components
relative to the largest samples of the frequency components. Moreover, the
adaptive modulation means may increase the amplitude all of the selected
samples of the frequency components that are to be transmitted to above a
predetermined minimum value.
Preferred embodiments may also include the following features. The
transmitter uses a single carrier to transmit the combined signal. The
television signal is processed so as to reduce its information content
before being input to the dividing means. The information content
reduction processing of the television signal employs prediction. The
generating means generates each of the plurality of spread spectrum
signals by multiplying the corresponding one or more of the plurality of
blocks by a different member of a set of pseudorandom sequences. The set
of pseudorandom sequences is a set of orthogonal pseudorandom sequences.
In general, in another aspect, the invention features a television receiver
for receiving a television transmission signal that was derived from a
television signal. The receiver includes means for multiplying a received
signal by each member of a set of pseudorandom sequences to generate a
plurality of signal components; means for calculating whether a selected
member of the plurality of signal components will improve picture quality
if used in reconstructing the video signal; means for selecting at least
some of the plurality of signal components based upon the calculations of
the determining means; and means for generating a reconstructed television
signal from the selected signal components.
In preferred embodiments, the calculating means includes means for
measuring the CNR of each of the signal components or it may include means
for measuring the amplitude of each samples of the plurality of signal
components and threshold means for accepting only those samples exceeding
a predetermined threshold associated therewith. The predetermined
threshold associated with a given signal component is derived from an
observed peak value for that frequency component. The generating means
includes a quadrature mirror synthesis filter bank. In the case that the
television transmission signal was generated by employing an adaptive
selection technique, the receiver further includes an adaptive selection
decoder for decoding adaptive selection information. In the case that the
television transmission signal was generated by employing a adaptive
modulation technique, the receiver further includes an adaptive modulation
decoder for decoding adaptive modulation information. In the case that the
television signal was processed so as to reduce its information content
before being transmitted, the generating means processes the selected
signal components in a manner complementary to the processing which was
done on the television signal to reduce its information content. When the
information content reduction processing of the television signal employs
prediction, the generating means processes the selected signal components
in a manner complementary to the prediction processing which was done on
the television signal.
One advantage of the invention is that it enables one to design a
high-definition TV system of maximum spectrum efficiency, i.e., which
delivers close to the maximum number of programs of a given technical
quality to each viewer within a minimum overall spectrum allocation. In
addition, a system that uses the invention delivers close to the maximum
possible quality to each viewer, taking account of the local CNR as well
as the assigned bandwidth, and it transmits a video image that degrades
gracefully as transmission conditions deteriorate.
The invention also enables one to build a TV broadcasting system that can
be improved over time in a nondisruptive manner, i.e., without requiring
replacement of existing HDTV receivers or of existing HDTV studio
equipment and transmitters. A TV system that incorporates the invention
can be tailored to the local interference situation, so that close to the
maximum possible quality images are received taking into account
interference from and to other stations.
The invention facilitates the design of receivers for an HDTV system in
which lower-resolution images can be produced at lower cost by recovering
less than all of the transmitted signal data. The invention also
facilitates the design of low-cost converters to convert the HDTV signal
into one that can be viewed on a standard receiver, such as intended for
NTSC.
Other advantages and features will become apparent from the following
description of the preferred embodiment and from the claims.
DESCRIPTION OF THE PREFERRED EMBODIMENT
FIG. 1 is a high-level block diagram of a television encoder/transmitter;
FIG. 2 is a high-level block diagram of a television decoder/receiver;
FIG. 3 is a block diagram of a direct-sequence encoder;
FIG. 4 is a block diagram of a direct-sequence decoder;
FIG. 5 is a block diagram of a modulator circuit;
FIG. 6 is a block diagram of a demodulator circuit;
FIG. 7 is a block diagram of a spread spectrum version of the MIT-CC
system;
FIG. 8 illustrates an 8.times.8 decomposition of the television signal in
the spatial frequency plane;
FIG. 9 illustrates an average selection rate for each of the subbands in
the spatial frequency plane;
FIG. 10 illustrates a grouping of subbands based upon average selection
rates;
FIG. 11 shows the magnitude of the frequency of the components used to
generate the signal groupings for the direct-sequence encoder;
FIG. 12 illustrates a plan for generating the order of component groupings
for the direct-sequence encoder;
FIG. 13 is a more detailed block diagram of a portion of the transmitter
including the channel coder;
FIG. 14 is a more detailed block diagram of the receiver; and
FIG. 15 is the transfer function of the thresholding circuit.
STRUCTURE AND OPERATION
For the purposes of this application, spread spectrum (SS) means a method
of multiplexing N signals into a single transmission channel, sometimes
called code division multiple access (CDMA) as an alternative to frequency
division multiple access (FDMA) or time division multiple access (TDMA).
In the latter schemes, the sum of the bandwidths of the N signals is equal
to or less than the channel bandwidth. In TDMA, each signal is
time-compressed by a factor of N, raising its bandwidth to that of the
channel, and the signals are transmitted in sequence. The ensemble of
signals thus occupies all of the time and all of the bandwidth. In FDMA,
each signal is transmitted in a separate frequency channel, so that the
ensemble of signals has a total bandwidth equal to that of the channel. In
both cases, any one signal uses either 1/Nth of the time or 1/Nth of the
bandwidth. In spread spectrum, each component uses all of the bandwidth
and all of the time; the signals are separated by coding. Each signal is
exanded to the full channel bandwidth by multiplying each signal sample by
a pseudorandom sequence of length N whose clock rate is twice the channel
bandwidth in Hz. There are exactly N orthogonal sequences of length N. All
such products are added together and transmitted in the channel.
What has just been described is a baseband system. For RF transmission by
quadrature amplitude modulation, two sum-of-products signals are produced,
each comprising half of the samples and half the channel bandwidth. These
two signals are then modulated in quadrature on a single carrier. For
systems more advanced than quadrature modulation, such as trellis coding,
a more complicated scheme is used, but the general idea is the same.
Overview of Transmitter
FIG. 1 is a high-level block diagram of a basic embodiment of the system.
The description does not include required synchronization and storage
elements and it does not specify whether individual signals are in digital
or analog form. Such details would be obvious to those skilled in the art.
An original video signal 10 is input to a source coder 12, which produces
two classes of output signals, namely a coded "lows +" signal 14 and coded
"highs" signals 16. Coded "lows +" signal 14, which includes the low
frequency components of television signal plus certain data required for
decoding, is transmitted essentially without error. Coded highs signals
16, which represent the high frequency components of the video signal and
which typically have zero average value, may suffer some error in
transmission and still be useable.
Coded highs signals 16 are transmitted by spread spectrum. That is, each
such sample of such components is multiplied by a pseudorandom (PN)
sequence of length N, where N is such as to expand the bandwidth of each
stream of samples to be the full channel bandwidth (or half the channel
bandwidth if a quadrature amplitude modulation (QAM) system is used).
Coded highs signals 16 pass to a channel coder 18, which includes a
direct-sequence encoder (DSE) 20 and a pseudorandom sequence generator
(PNG) 22. PNG 22 generates the group of pseudorandom sequences (PN).
Direct-sequence encoder (DSE) 20 multiplies each of coded highs signals 16
by a unique pseudorandom (PN) sequence to generate product signals that
are then combined to produce a sum-of-products signal 24. More
specifically, the product signals are added together with different power
levels, generally with higher power being used for lower-frequency
components. Thus, in general, the lower-frequency components among coded
highs signals 16 are transmitted with relatively larger amplitudes.
A modulator 26 accepts sum-of-products signal 24 plus coded "lows +" signal
14 as well as a carrier frequency, f.sub.c, and produces a modulated
carrier 28, which is input to a transmitter 30. Generally, transmitter 30
is simply a frequency shifter and amplifier that produces a high-frequency
modulated signal 32, which is directed to a transmitting antenna 34.
By using spread spectrum, it is possible to generate a transmitted signal
that looks like white random noise. Such a signal is particularly useful
since, for a given transmitted power, this minimizes the visibility of
interference into other signals in the same or adjacent channels.
Scrambling of the samples of the original components before the spread
spectrum operation guarantees that the crosstalk and interference depend
only on the average signal levels, and not the peak signal levels.
Note that when binary digital signals are used in spread spectrum, the
"multiplication" process is actually the exclusive-or process, and the PN
sequence has the logical values 1 and 0. When analog or multilevel
"digital" signals are used, the PN sequence has the values +/-1 and the
operation is ordinary multiplication.
Overview of Receiver
Referring to FIG. 2, at the receiving end, a television receiver 36 accepts
the received signal from a receiving antenna 38. In receiver 36, a front
end 40 frequency shifts and amplifies the received signal and a channel
equalizer 42, if used, compensates for frequency distortion (e.g.,
multipath). The output of receiver 36 (i.e., signal 44) is, ideally, a
replica of modulated carrier signal 28 (See FIG. 1), which has passed
through a channel equivalent to an ideal bandpass filter and which has had
a certain amount of random noise added to it. Signal 44 is demodulated in
a demodulator 46 to produce received lows signal 48 and received highs
signal 50, which are approximations of the corresponding lows signals 14
and sum-of-products signal 24, respectively, at the sending end.
Demodulator 46 recovers the carrier frequency and perhaps other
synchronizing signals by using techniques well known to those skilled in
the art.
At the receiver, the sum-of-products signal within the transmitted signal
is multiplied by each of the PN sequences in turn and integrated over the
period of one sample. If the N sequences are mutually orthogonal, only one
signal is recovered by each demodulation process; there is no crosstalk.
Received highs signal 50 is then separate into its original components by a
channel decoder 52. Channel decoder 52 includes a direct-sequence decoder
(DSD) 54 and a PN generator 56 which generates the same PN sequences that
were used at the sending end. With suitable filtering (e.g., integration
over one sample duration), the original samples are recovered. Received
lows signal 48 and the output of DSD 54 then pass to a source decoder 58
which produces a replica of the input video signal, although, in general,
noisier and perhaps of lower resolution, depending on channel conditions.
As will be described in greater detail later, some scheme is used to judge
the noisiness of the recovered components to establish whether their use
in signal reconstruction will improve image quality and only those that
improve image quality are utilized. With such a procedure, recovered image
quality improves with CNR at the receiver, since more and more components
are usable at higher and higher CNR.
Note that the noise in the recovered signal depends only on channel noise.
In the white-noise case if all products are added with equal wieght, CDMA
produces the same recovered SNR for each component as do TDMA and FDMA,
i.e., the SNR is equal to the channel CNR. However, CDMA has two
capabilities not possessed by FDMA and TDMA. In addition to the N
orthogonal sequences, there are many other sequences of the same length
that are nearly orthogonal. (In the case of very long sequences, almost
all possible sequences are nearly orthogonal.) Thus, additional components
can be multiplexed at the cost of some small loss of recovered SNR.
In addition, the product signals in the transmitter can be added with
different relative amplitude, so that the relative SNR of the signals at
the receiver can be adjusted. This is highly advantageous in TV since all
components do not require the same SNR. When different amplitudes are used
for the various products in such a way as to maintain the total
transmitted power, then the SNR of some components is raised while that of
others is lowered, thus producing a substantial improvement in perceived
SNR of the reconstructed image. In addition, when it is desired to add
components in order to raise the spatial resolution, very little power is
required for these additional components as they do not require high SNR
at the receiver. Thus, there is little loss of overall perceived SNR.
Source Coder
The invention is applicable to TV source-coding schemes that divide the
signal into a number of frequency components, and in which some or all
(typically not including the lowest-frequency components) of these
components are of the type in which small errors in amplitude produce
correspondingly small deterioration of image quality.
Source coder 12 divides the video signal 10 into a group of spatial or
spatiotemporal frequency components or subbands. It may use any of a
number of techniques known to those skilled in the art. In the described
embodiment, source coder 12 uses a filter bank in which the filters are
quadrature-mirror filters such as are described by P.P. Vaidyanathan, in
"Quadrature Mirror Filter Banks, M-Band Extensions and Perfect
Reconstruction Techniques," IEEE ASSP Magazine (July, 1987) pp. 4-20.
These filters have the advantage that the sum of the data rates for the
sampled filter outputs (i.e., frequency components) is the same as the
data rate for the original signal. Unless the filters have infinitely
sharp cutoff characteristics, which is neither possible nor desireable,
the subbands will overlap in frequency and therefore there will be
aliasing in each component. This aliasing due to the operation of the
analysis filter bank is exactly cancelled in the synthesis filter bank
that is used to reconstruct the signal at the receiver.
Note that the discrete cosine transform (DCT) is a special case of subband
coding in which the image is divided into blocks, typically 8.times.8
picture elements, before transformation. DCT coefficients correspond
exactly to subband samples if a particular set of analysis and synthesis
filters is used. Thus, hereinafter, when the word "sample" is used, it is
to be considered interchangeable with "coefficient".
Direct-Sequence Encoder
Referring to FIG. 3, direct-sequence encoder (DSE) 20 processes the m highs
components or samples, x.sub.1 through x.sub.m, that are produced by
source coder 12. Each train of samples, x.sub.i, passes to an associated
multiplier 70 where it is multiplied by a different unique one of a group
of pseudorandom sequences, {PN.sub.i }, generated by PNG generator 22.
Each pseudorandom sequence PN.sub.i is of length N, is a member of a set
of orthogonal sequences, and has a clock rate that is twice the channel
bandwidth in Hz. (Note that there are exactly N orthogonal sequences cf
length N.) Multiplying each signal sample by a pseudorandom sequence
expands the signal to the full channel bandwidth.
The output of each multiplier 70 passes to a weighting circuit 72 that
multiplies the product by an appropriate weight. In other words, the
products (i.e., the outputs of multipliers 70) are assigned different
power levels, generally with higher power being used for lower-frequency
components. The weighting process is symbolized here by adjustable
resistors, but any appropriate means may be used to achieve the desired
weighting. The weighted signals are then combined in a adder 74 to produce
a weighted sum-of-products signal (i.e., sum-of-products signal 24 shown
in FIG. 1) which passes to the modulator.
Direct-Sequence Decoder
Referring to FIG. 4, direct-sequence decoder (DSD) 54 processes the output
of the demodulator in the receiver to generate the m frequency components
that were sent. The output of the demodulator including the
sum-of-products signal is input to m multipliers 80, one for each of the
frequency components which is to be extracted from the received signal. In
each multiplier 80, the demodulated signal is multiplied by a different
one of the pseudorandom sequences {PN.sub.i } that were used in DSE 20.
Each of the resulting product signals is then integrated by integrators 82
for a period equal to one pseudorandom sequence length. Each integrator 80
applies a weight that is the reciprocal of the weight applied by weighting
circuits 72 in DSE 20. Thus, to the degree that the sequences are
orthogonal, each multiplier 80 plus integrator 82 combination picks out
one of the original trains of samples from the demodulated signal,
producing signal x'.sub.i, which is a replica of the corresponding signal
x.sub.i at the encoder, but with the addition of some noise.
Modulator and Demodulator Circuits
FIG. 5 shows a typical quadrature modulator that may be used in the system.
If such a modulator is used, the product signals produced by multiplying
the frequency components by the corresponding PN sequences are weighted
appropriately and then placed into two equal-sized groups, identified as
G.sub.I and G.sub.Q, where "I" signifies in-phase and "Q" signifies
quadrature-phase. The G.sub.I and G.sub.Q signals are added by adders 90
and 92 to multilevel digital signals DS.sub.I and DS.sub.Q. This embedding
of the analog signals carrying the high frequency information in the
digital signal is done in accordance with the methods described in U.S.
Pat. 4,979,041 issued Dec. 18, 1990 to W.F. Schreiber, entitled "High
Definition Television System", and incorporated herein by reference. Note
that the peak-to-peak amplitude of the G.sub.I and G.sub.Q signals must be
less than half the level spacing of the multilevel digital signals
DS.sub.I and DS.sub.Q. The two signals containing the embedded analog
information are then multiplied in multipliers 94 and 96 by a carrier
frequency, f.sub.c, and a replica thereof shifted 90 degrees. The products
are added in an adder 98 to produce the input signal to the transmitter
stage.
Of course, any modulation scheme can be used that enables one to transmit
highs by spread spectrum and that transmits lows plus decoding data (also,
AM and AS data, if applicable) nearly without error.
FIG. 6 shows a typical quadrature demodulator which may be used in the
system. The received signal is applied to two multipliers 100 and 102, fed
respectively by the carrier frequency f.sub.c and a replica thereof
shifted 90 degrees. The products produced by multipliers 100 and 102 are
passed through lowpass filters 104 and 106, respectively, to remove the
carrier frequency signal, thereby producing the baseband signals 108 and
110. Baseband signals 108 and 110 then pas through analog to digital
converters (ADC) 105 and 107 to produce digital signals 109 and 111,
respectively. In order to separate the multilevel digital signals
containing the low-frequency information and other data from signals 109
and 111, the latter pass to quantizers 112 and 114, where they are
quantized using the same levels used to produce the multilevel digital
signals DS.sub.1 and DS.sub.2 in the transmitter. The resulting quantized
signals are identified as DS.sub.I ', representing the in-phase signal,
and DS.sub.Q ', representing the quadrature phase signal. DS.sub.I ' is
subtracted from signal 109 by subtractor 116 to obtain a G.sub.I ' signal,
representing the in-phase highs signal. Similarly DS.sub.Q ' is subtracted
from signal 111 by subtractor 118 to obtain a G.sub.Q ' signal,
representing the quadrature-phase highs signal. The four signals G.sub.Q,'
G.sub.I ', DS.sub.I ' and DS.sub.I ' are replicas of the corresponding
four signals which were input to the modulator in the transmitter.
What has been described is a hybrid transmission method wherein the spread
spectrum signal is superimposed on a digital QAM signal. The invention may
be used with any other method of multiplexing the signals that must be
transmitted error-free, generally digitally, and those that can experience
some degradation (such as by the addition of a small amount of noise)
without producing catastrophic loss of image quality. For example, the
digital signal can be transmitted by QAM in one subchannel, or the two
signals can be transmitted in the same channel by time division
multiplexing.
A Spread Spectrum Version of the MIT-CC System
A spread spectrum version of the MIT-CC system (a channel compatible system
developed at the Massachusetts Institute of Technology) will no be
described to further illustrate the invention. The MIT-CC system transmits
images of 720.times.1280 active elements per frame at a rate of 60 fps,
for an uncoded transmission rate of 55.2 Megapixels/sec. It uses hybrid
transmission in which the data that must be substantially error-free is
transmitted digitally using 4-QAM at 9.83 Mb/s while the selected subband
samples, which need not be error-free, are superimposed on the digital
data as analog samples at 9.83 Megasamples/sec. The analog samples are
selected from the 55.2 Ms/s of the original signal as those most important
to image quality. In the original form of the MIT-CC system, the selected
analog samples, which are adaptively modulated and scrambled, are
transmitted two-by-two (i.e., two samples per symbol) through the channel.
Since it uses a QAM signal, that is equivalent to one-by-one data
transmission in each phase of the QAM signal.
Referring to FIG. 7, in the spread spectrum version of the MIT-CC system, a
3.times.3 linear matrix 152 processes an RGB high-definition video signal
150 to generate the standard luminance/chrominance form. A
quadrature-mirror analysis filter bank 154 uses two-dimensional subband
decomposition to divide the spectrum of each video frame into 8.times.8
(i.e., 64) frequency components, each having dimensions 90.times.160 pels
(see FIG. 8). The lowest frequency components, which are identified by the
letter "D" in FIG. 8 are sent error-free at full resolution. That is, they
are processed by a lows coder 156 into a digital signal at 4 Mb/s, while
the other components are treated separately. Lows coder 156 uses a JPEG
(Joint Picture Expert Group) algorithm for coding the lows. See e.g. G.K.
Wallace, "Overview of the JPEG Still Picture Compression Algorithm,"
Elect. Imaging Conference, Boston, Oct. 29-Nov. 1, 1990. Optionally, a
forward error correcting code (FEC) may be used on the lows signal.
In the described embodiment, the highest ten frequency components are
discarded (i.e., not sent). The remaining blocks, numbering 51 components
and representing the highs signal are sent using the spread spectrum
techniques described herein.
There are two methods of dealing with the fact that different SNR is
required for different samples, according to their frequency. For example,
the number of bits/sample may be preassigned to each frequency.
Alternatively, an adaptive modulation (AM) process may be used in which
the smaller samples are multiplied by larger adaptation factors. To
minimize the quantity of side information, each transmitted adaptation
factor is used for many samples. At the receiver, each received sample is
divided by the appropriate adaptation factor to restore the sample to
approximately its original amplitude. The division process also reduces
any noise or interference added in the channel. Either method can be used
with the invention.
Experience demonstrates that with high-definition TV systems, only about
1/8 or less of all the samples are needed for intraframe coding. Thus,
adaptive selection of only these most important samples may be used. Of
course, when samples are adaptively selected, their location in the
frequency plane or the space plane (both methods are used) must be
indicated to the receiver. Such information (referred to as AS data) must
be delivered essentially without error, whereas the sample amplitude can
suffer small errors without catastrophic effect on the reconstructed
image.
In the described embodiment, improved noise performance is achieved by
adaptive modulation, and data compression is achieved by adaptive
selection of about 10 Ms/s of subband samples out of the 55.2 Ms/s in the
original signal. An AM compute module 158 computes the adaptive modulation
factors to be applied to the image frame and an AM encoder 160 encodes the
AM factors. Similarly, an AS compute module 162 computes the adaptive
selection information and an AS encoder 164 encodes the adaptive selection
decisions. (The AM and the AS data are each coded to about 2 Mb/s, using
the techniques to be described shortly.)
Since some distortion occurs in the coding of these two data streams, they
are decoded at the sending end exactly as they are to be decoded at the
receiving end, and the reconstructed AM and AS data is used to process the
video information by applying both the adaptive modulation factors and the
adaptive selection decisions generated for the image. As shown in FIG. 7,
an AS decoder 166 decodes the coded adaptive selection information from AS
encoder 164 and an AS circuit 168 implements the adaptive selection
decisions on the image data. Similarly, an AM decoder 170 decodes the
coded adaptive modulation from AM encoder 160 and an AM circuit 17 | | |
