NL8402009A - SIGNAL PROCESSING DEVICE. - Google Patents

Description

ί v VO 6373 3etrf; Signal processing device.

The invention relates to a signal display device for analyzing and / or synthetically building up signals. More specifically, the signal processing apparatus of the invention uses a pipeline architecture to analyze the frequency spectrum of an information component (with one or more dimensions) of a given temporal signal, the highest of interest, in a delayed real time frequency is not greater than fQ, and / or for building up such a temporal signal synthetically from the analyzed frequency spectrum thereof in delayed real time. Although the invention is not limited thereto, the invention is particularly suitable for delayed real-time image processing of the two-dimensional spatial frequencies of television images, which are determined by a temporal video signal.

Sr has done much work in modeling the operation of the human visual system. It has been found that the human visual system appears to calculate a primitive spatial frequency discharge of light images by splitting spatial frequency information into a number of contiguous, overlapping spatial frequency bands. Each band has a width of about one octave, and the centrecentence of each band differs from its neighbors by substantially a factor of two. Investigations suggest that there are approximately seven bands or "channels" present which span the spatial frequency range of 0.5 to cQ period degrees of the human visual system. The importance of this conclusion is that spatial frequency Information that is more than a factor of two from other spatial frequency information will be processed independently by the human visual system.

It has further been found that the spatial frequency distortion takes place in the human visual system. has been localized in space. Therefore, the signals in each spatial frequency channel over small subareas of the image are calculated. These subareas overlap and have a width of about two periods at a given frequency,

Indian uses a sine wave roster pattern as the test pattern. It appears that the threshold contrast sensitivity function for the sine wave 8402009

W.

£ i - 2 - raster image decreases rapidly as the spatial frequency of the sine wave raster image increases, D, w, that high spatial frequencies require a large contrast to be observed (^ 20% at 30 periods / degrees), but lower spatial frequencies require that a relatively low contrast be observed (^ 0.2% at 3 periods / degrees),

It has been found that the ability of the human visual system to detect a change in the contrast of a sine wave lattice image, which is above a threshold value, is also better at lower spatial frequencies than at high spatial frequencies, More specifically requires an average human to properly discriminate a changing contrast 75% of the time, approximately a 12% change in contrast for a 3 periods / degrees sine wave lattice, but a 30% change in contrast for a grid of 30 periods / degrees,

Dr, Peter J, Burt, who knows the above-described properties of the human visual system, has developed an algorithm (hereinafter referred to as the "Burt Pyramid"), which he realizes in real time with the aid of a calculator to calculate the analyze two-dimensional spatial frequencies of an image in a number of separate spatial frequency bands, Each spatial frequency band (different from the lowest spatial frequency band) preferably has an octave width, If therefore the highest spatial frequency, which important, of the image not greater than 25 fQ, the highest frequency band will cover the octave from fg / 2 to fg (with a center frequency at 3fgA); the second highest frequency band will cover the octave from fgA to fg / 2 (with a center frequency at 3fg / 8), etc,

Referred to the list below of articles which are of the 30 hand or co-owned by Dr, Burt, which articles detail various aspects of the Burt Pyrsmide; "Segmentation and Estimation of Image Region Properties Through Cooperative Hierarchial Computation." by Peter J. Burt. e, a, IEEE Transactions on Systems, Man, II-1. I ΐ | · - ^ '· ι- ··' "ί> 'ι ·» P'> Ί 7, i * i 1 35 and -Cybernetics, Vol, SMC-11, Ho, 12, 802-809, December 19Ö1.

"The Laplacian Pyramid as a Compact Image Code," by

Peter J, Burt, e, a ,, IEEE Transactions on Communications, Vol, 8402009 £ * - 3 - CGM-31, No. 4,532-540, April 1983.

"Fast Algorithms for Estimating local Image Properties," by Peter I. 3urt, Computer Vision, Graphics, and. Image Processing 21, 363-382 (1983).

5 "Three and Pyramid Structures for Coding Hexagonally Sampled Binary Images," by Peter J. 3urt. Computer Graphics and Image Processing 14, 271-280 (1980), "Pyramid-based Extraction of Local Image Features with Applications to Motion and Texture Analysis," by Peter I. Burt, SPIE. Vol 36θ, 10 114-124.

"Fast Filter Transforms for Image Processing," by Peter J. Burt,

Computer Graphics and Image Processing 16, 20-51 (1981).

"A Multiresolution Spline with Applications to Image Mosaics," by Peter J. Burt, et al., Image Processing Laboratory, Electrical, Computer, 15 and Systems Engineering Department, Rensselaer Polytechnic Institute, June 19, 03.

"The Pyramid as a Structure for Efficient Computation." by Peter J,

Burt, Image Processing Laboratory, Electrical and Systems Engineering Department, Rensselaer Polytechnic Institute, July 1982. 1 2 3 4 5 6 7 8 9 10 11 8402009

The Burt Pyramid algorithm uses Determine 2 the sampling methods to analyze an original image 3 with relatively high resolution in a hierarchy of 51, (where 51 is a 4-pi integer) separate component images (each of which has 5 component images). is a Lapiae image, which is a different octave of 6 comprising the spatial frequencies of the original image) plus a 7 residual G-auss image (containing all spatial frequencies of the original 8 image below the lowest octave component of the Laplace 9 image includes). The term "pyramid" as used herein refers to the successive reduction in spatial frequency band width 11 and sample density of each of the component hierarchies from the highest octave component image to the lowest octave component image.

- 1 * - ♦ s

A first advantage of the Burt Pyramid algorithm is that it makes it possible to synthesize the original high-resolution image synthetically from component images and the remaining image without introducing disturbing spatial frequencies as a result of aliasing. A second advantage of the Burt Pyramid algorithm is that the spatial frequency bandwidth of one octave of each of the component images' hierarchy is adapted to the properties of the human visual system discussed above, thus enabling the spatial frequencies of indivir-10 selectively process or modify dual images of the hierarchy of component images in different independent ways (ie, without the signal processing of any component image significantly affecting another component image) in order to produce another desired effect in the synthesized component images to strengthen or obtain a built image. An example of such a desired effect is the "SPIE method" with a number of resolutions, which is detailed in the article "fA Multiresolution Spline with Applications to Image Mosaics" mentioned above, 20 So far it has been Burt Pyramid algorithm in non-wara time realized by a general purpose digital calculator The level of each pixel (pixel) sample of an original image is represented by a number of bits (for example 8 bits), which is on a individual address location of a computing device memory is stored, for example, a two-dimensional original image of relatively high resolution, comprising 2 ^ (512) pixel samples in each of its two sizes 18, requires a large memory with 2 {262, [kk] address locations for storing each of the multi-bit numbers, respectively, representing the levels of the respective pixel samples proposals, which include the original image,

The original image, which is stored in memory, can be processed by a digital calculator in accordance with the Burt Pyramid algorithm. This processing includes the iterative execution of steps, such as convolution of pixel samples with a predetermined core weighing function, sample decimation, sample expansion by interpolation and sample subtraction, The size of the core function 8402009 9 * - 5 - (in either one or more dimensions) is relatively small (in terms of the number of pixels) compared to the size in each size of the whole image, The bottom area or window of image pixels (which is equal in size to the core function and in turn symmetrical about each image pixel is multiplied by the nuclear sweeping function and summed in a convolution calculation,

The nuclear sweeping function is chosen to be a low-pass filter for the multidimensional spatial frequencies of the image being convolved. The nominal "pinch" - (also known in the 1C filter technique as "" angle "or" break "-) frequency of the low-pass filter characteristics provided in each size by the core function is chosen to be substantially equal to half of the highest frequency of interest in that size of the signal to be convolved. This low-pass filter characteristic likewise need not have a brick vail, especially at a given cutoff frequency, but may have a relatively gradual decrease. , in which case a nominal cutoff frequency is defined as the frequency, with a preselected rate (e.g. 3 d3) of damping in the gradual decrease taking place.

Filters with smoother drop characteristics can be used because the Burt Pyramids inherently introduce interfering frequencies, resulting from alias distortion, caused by a gradual decrease in low-pass filter characteristics. compensates. The convoluted image is decimated by successively considering in each of the respective dimensions of the image, each cm, to eliminate the other convolved pixel present, thereby increasing the number of pixels in the convoluted image in each size. half of that is reduced. Since an image is conventionally a two-dimensional image, an uncompressed-decimated image comprises only one-fourth of the number of pixels present in the image for such decimation. The reduced number of pixel samples of this convoluted-decined image (which is called a gauss image) is stored in a second memory. 35 Based on the unbeaten pixel samples of the original image, the above-described convolution-decimation procedure is iteratively H times executed (where il is a full integer 8402009 - 6 - number) leading to (H + 1) heals, which include the original high resolution image and a hierarchal pyramid of II additional reduced resolution guis images, where the number of pixel samples (sample density) in each size of each distant image is only half the number of pixel samples in each size of the immediately preceding image. If the original large resolution stored image is indicated by GQ, the hierarchy of N stored further images may be indicated by G ^ to G ^ respectively, the successively reduced number of pixel samples of each of these II further images is stored in a separate memory of N memories, therefore, there is, counting the original stored image. a total of H + 1 memories are present. '

According to the inexact time realization of the Burt Pyramid algo-15 rhythm, the following calculation procedure is to generate further pitch tests, with interpolated values between each pair of stored G ^ pixel samples in each size, thereby reducing the sample density of the saved G ^ image is expanded again to the sample density of the original saved Gq image, The 20 digital value of each of the pixel samples of the expanded G ^ image then becomes of the saved digital value of the corresponding pixel subtracted sample from the original GQ image to provide a difference image (known as a Laplace image), This Laplace image (denoted by Lq), which has the same sample density as the original GQ image, includes those spatial frequencies, which are present in the original image within the octave fQ / 2 to fQ - plus often a small error-compensating component with 1 lower spatial frequency, corresponding to the loss of information caused by the decimation step, respectively, which is used in deriving the reduced stesk sample density from the G ^ image and introducing interpolated value samples , which takes place upon expanding the sample density to that of the original G0 image. This Laplace image Lq then replaces the original image Gq stored in the first memory of the ff + 1 pyramid memories.

Similarly, by an iteration of this procedure, a hierarchy becomes which further Laplace images L. through L ™.

8402009 * ε - τ - includes, sequentially derived and recorded in a corresponding memory of the respective further 21-1 memories, in which the * auss images 31 to '1 ^ 1 are stored (whereby the memory is the • iSaussheide G to 0ΛΤ. are replaced). The sauce image (with the most reduced sample density) is not replaced in its memory by a Laplace image, but remains in this memory as a Gauss remainder. comprising the lowest spatial frequencies (nfl, below the L ^ octave) present in the original image. The 3-hour Pyramid algorithm allows the original image to be restored without an aliasing by an iterative arithmetic process, which is successive steps from expanding the stored residual image G ^ to the sample density of the Ljj. ^ picture and then adding it to the saved Laplace 15 picture L ^. ^ for obtaining a sum policy. This son image is expanded in a similar manner and counted for the Laplace image Ljj 2, etc., until the original high-resolution image is synthetically constructed by summing all the Laplace images and the remainder image. Furthermore, according to the analysis of one or more 20 original images in 21 Laplace images and a G; au.ss residual image, possibly producing a certain desired image processing or modifying step (such as "splining, r") before a full high-resolution image is synthetically constructed therefrom,

The true-time realization of the Burt Pyramid algorithm 25 by arithmetic processing is effective in processing fixed image information. Therefore, it does not apply to the analysis of a stream of consecutively occurring images, which may continuously change over time (e.g. successive video frames of a sea vision image), Sen real-time realization of the 3urt Pyramid al-30 gorithm, as provided according to the invention is necessary to analyze such successively occurring time-changing images.

More particularly, the invention is directed to a signal processing apparatus using a pipeline architecture to analyze, in delayed real time, the frequency spectrum of an information component of a given temporal signal with the highest frequency of interest of this frequency spectrum 8402009 jv - 8 - does not exceed f ^, furthermore this information component of the determined temporal signal corresponds to information with a certain number of dimensions, The device comprises a set of ordinally arranged sample signal interpreters (where N is a plural integer 5), Each of the interpreters includes first and second input terminals and first and second output terminals, The first input terminal of the first interpreter of the set serves to receive the determined temporal input signal, The first input terminal of each of the second until the Hth's interpreting bodies are linked 1d with 10 the first output terminal of the immediately preceding interpreter of the set, so that each of the second to the U-th interpreters sends therefrom a signal to the immediately following interpreter of the set, The second input terminal of each of the interpreters of the set serves to receive a separate sample clock. In this construction, each of the interpreters of the set at the first and second output terminals thereof receives signals at a frequency equal to the sampling frequency of the sample. clock signal applied thereto,

Furthermore, each of the interpreters of the set has a low-pass transfer function between the first input terminal and its first output terminal for the information component of the signal applied to its first input terminal. The low-pass transfer function of each interpreter of the set has a nominal cut-off frequency, which is a direct function of the sampling frequency of the clock signal applied to the second input of that interpreter of the set. Furthermore, the clock sew, applied to the second input terminal of the first interpreter of the set, has a sampling frequency which (a) is double that of f ^ and (b] provides a no-3Q minimum pinch-off frequency for said information component. for the low-pass transfer function of the first interpreter of the set, which is less than fQf. Furthermore, the clock signal applied to the second input terminal of each of the second to the JT interpreters of the frame has a sampling frequency, which is less than the clock frequency which is applied to the second input terminal of the immediately preceding interpreter of the set, (b) is at least equal to twice the maximum frequency of the information component, which is supplied to the first 8402009> t-9 input terminal and (c) provides a nominal cutoff frequency for its low-pass transfer function, which is smaller is then that of the immediately preceding representation of the set, 5 The signal taken at the second output terminal of each of the sets of the set corresponds to the difference between the information component applied to its first input terminal and a direct function of the information component, which is taken from its first output terminal.

10 Although not limited to that, the information component of the :; certain temporal signal processed by the signal processing device according to the invention, for example corresponding to the two-dimensional spatial frequency components of each of successive frames of a television picture scanned in series in each of two dimensions,

In general, the invention is useful in analyzing the frequency spectrum of a signal taken from a source of spatial or non-spatial frequencies in one or more dimensions, regardless of the particular nature of the source. Therefore, the invention useful for example in analyzing one, two,. three or more dimensional complex signals, which come from audio sources, radar sources, seismographic sources, robot sources, etc., in addition to two-dimensional visual image sources, such as television images,

Torch, the invention is also directed to a signal processing apparatus using a pipeline architecture and which synthetically builds such a complex signal in response to a set of delayed real time signals,

The invention will be explained in more detail below with reference to the drawing, in which: Fig. 1 shows a functional block diagram showing the invention in the most general and generic form. Fig. 1a shows a digital embodiment of a first specie of a single member of the set of sample signal interpreters of FIG. 1a, a digital embodiment of a second species of a single member of the set of sample signal interpreters of FIG. 1; 34 02 0 0 9 10 -

Λ Q

fig. an alternative digital embodiment of the end member of the set of sample signal interpreters of the first or second species of FIG. 1; Fig. 2 is an illustrative example of a core weighing function which can be used in realizing the invention; FIG. 3 is a block diagram of a one-dimensional array of a spectrum analyzer, a spectrum changing circuit, and a signal synthesizer according to the invention, having text identifying certain blocks therein; Fig. 1+ shows a block diagram of one of the analysis steps used in the iterative calculations of the spectral analysis process according to Fig. 3, which analysis takes place according to the invention; FIG. 5 is a block diagram of a modification that can be applied to a consecutive pair of the analysis steps of FIG. 15 in another embodiment of the invention. FIG. 6 is a block diagram of one of the synthesis steps used in the iterative process of the signal synthesis according to Fig. 3 from spectral components is used; Figures T, 8, 9 and 10 are block diagrams of representative spectrum change circuits of Figure 3 for use in the invention; FIG. 11 is a block diagram of a modification of the system of FIG. 3, which is applied when it is desired to center spectrum samples over time for processing according to the invention; FIG. 12 is a block diagram of a two-dimensional spatial-frequency spectrum analyzer using a pipeline architecture to perform a real-time spectral analysis; and FIG. J3 is a block diagram of a synthetic signal building apparatus which determines the sample field analyzed by the spectrum analyzer of FIG. 12 from its output spectra.

As can be seen from fig. 1? each of a set of KT ordinally arranged sample signal interpreters 100-1 to 100-N (where I is a plural integer) has two input terminals and two output terminals. A certain temporal signal GQ, which determines information, is applied as an input signal to a first of the two input terminals of the first interpreter 100-1 of the set. The temporal signal 8402009) * - 11 -

Gq can be a continuous analog signal (such as an audio signal or a video signal) or the temporal signal GQ can be a sampled analog signal. Furthermore, in the latter case, any sample level can be directly represented by an amplitude level or indirectly represented by a digital number (ie, by passing each sample amplitude level through an analog-to-digital converter, not shown in FIG. 1) the temperature signal GQ is applied to the first input terminal of the interpreter 100-1). The frequency spectrum of G_ includes a region that extends between zero (i.e., DC) and the frequency f ^ (i.e., a region that includes all frequencies of interest corresponding to information of a given number of dimensions). More specifically, Gq may be a pre-filtered signal that does not contain a frequency greater than fQ. In this case, the clock frequency 2f ^ of the rendering device 10G-1 satisfies the Hyquist criterion for all frequency components of xq. However, gq may also contain some frequency components, higher than fQ, that are not important. In the latter case, the Hyquist criterion is not met and some aliasing occurs. From a practical point of view, although undesirable, such aliasing (if not large) can often be permitted.

In FIG. 1, the first input terminal of each member of the other interpreters 100-2 100-21 of the set is coupled to the first of the two output terminals of the immediately preceding interpreter of the set. More specifically, the first output terminal of the signal interpreter 100-1 is coupled to the first input terminal of the interpreter 100-2; the first output terminal of the interpreter 102 is coupled to the first input terminal of the interpreter 100-3, which is not shown; ... and the first output terminal of the interpreter of 100- (11-1), also not shown, is coupled to the first input terminal of the interpreter 100-2T. Therefore, the signal processing device shown in Figure 1 uses a pipeline architecture in coupling each of the respective barn interpreters with another.

A separate sample frequency clock signal is applied to the second of the two input terminals of each member of the set of interpretation members 1QC-1 ... 100-21. More in particular, a sample-frequent clock signal CL is applied to the interpreter 100-1.

3402009

* V

- 12 - supplied as a second input signal; a sample-frequency clock signal CLg is applied as a second input signal to the interpreter 100-2 ... and a sample-frequency clock signal CL1 is applied to the interpreter 100-N as a second input signal -5. The relative values of the clock signals CL ^ ... CLjj relative to each other are "limited in a manner indicated in FIG. 1. The importance of these limitations will be discussed in more detail later.

Furthermore, in FIG. 1, the interpreter 100-1 generates a second output signal LQ at the second output terminal thereof. In a similar manner, the other interpretation bodies wake up 100-2.,. 100-II of the set of respective second output signals. 1 ^. ^ to their respective second output terminals.

Each member of the interpreters 100-1 ... 100-IT of the set, regardless of its particular internal structure, can be considered a "black box" which performs a low-pass transfer function between its first input terminal and its first output terminal for the frequency spectrum of the information component of the input signal applied to its first input terminal. Furthermore, this low-pass transfer function of each member of the interpreters 100r1, 100-2 100-N of the set has a gradient that has a nominal cut-off frequency, which is a direct function of the sampling frequency of the clock signal, which is connected to the second input terminal. of which is supplied. As discussed above, in the case of the Burt 25 Fyramid, the reduction can be gradual rather than being a "brick wall",

More specifically, the input signal Gq discussed above is applied to the first input terminal of the interpreter 100-1.

The highest frequency of interest in the frequency spectrum of GQ 30 is not greater than fg. Furthermore, the sampling frequency of the clock signal CL.j, which is applied to the second input terminal of the interpreter 100-1, is equal to 2fQ (ie has a frequency which meets the Nyquist criterion for all frequencies of interest within the frequency spectrum of Gq). Under these conditions, the low-pass transfer function between the first input terminal and the first output terminal of the interpreter 100-1 is such that only those frequencies within the frequency spectrum of Gq, which are not greater

§ 4 0? _ 0 0 S

Then (where f ^ is less than f ^) are passed to the first output terminal of the interpreter 100-1. Therefore, at the first output terminal of the interpreters 100-1, an output signal G ^ has a frequency spectrum (determined by the determination of the characteristics of the low-pass transfer function) f, which substantially comprises the lower part of the frequency spectrum of C-q.

This signal G is then applied as an input signal to the first input terminal of the interpreter 100-2.

As shown in Fig. 1, the sampling frequency of the clock signal CL2 (applied to the second input terminal of the interpreter 100-2} is less than 2fQ (the sampling frequency of the clock signal CL ^), but at least equal to 2f ^ (twice the maximum frequency f ^ in the frequency spectrum of G ^). Therefore, the sampling frequency of the clock signal CL ^ is still high enough to satisfy the 3yquist criterion for the frequency spectrum of Gj, which is at the first input terminal of the interpreter 100 -2 is supplied, although the frequency is not high enough to meet the Nyquist criterion for the highest possible frequency of interest f ^ in the frequency spectrum of G ^, supplied to the first input terminal of the immediately preceding interpreter 1QC -1. This type of relationship (where the sampling frequency of the clock signal applied to the second input terminal of the interpreter of the set is lower. rdt as the ordinal position of this translating member of the set increases, generally applies. More specifically, the clock signal has that on. the second input terminal of each of the sets 100-2 100-ST of the set is supplied - a sampling frequency. which (a) is smaller than that of the clock signal applied to the second input terminal of the immediately preceding interpreter of the set, (b) is at least 30 times the maximum frequency of the information component of the signal, which is its first input terminal is applied, and (c) decreases the nominal cut-off frequency for its low-pass transfer function to a value less than that of the set's immediately preceding interpreter. Therefore, the maximum frequency 22 of the signal G ^ occurring on the second output terminal of the interpreter 100-2 is less than f. and finally the maximum frequency f 'is in the frequency spectrum of * 2i. 14 - A ^ the signal G ^ (which occurs on the first output terminal of the interpreter 100t.IT) is lower than the frequency f ^ of the frequency spectrum of the. signal G ^ (which occurs on the first output terminal of the interpreter - not resolved - of the set, which immediately precedes the interpreter 100-1 and is applied to the first input terminal of the interpreter 100-1T).

When each interpreter 100-1 100-N is considered a "black box" again, each of the respective output signals Lq. ..., which respectively occur on the second output terminal of each interpreter 100-1 ... 100-IT of the set, corresponding to the difference between the information component of the signal applied to the first input terminal of that interpreter, and a direct function of the information component of the signal which occurs on the first output terminal of that interpreter. Thus, when 15 is indicated in Fig. 1, Lq is equal to (or at least corresponds to) the difference Gq, where g (G ^) or G ^ is itself or is a particular direct function of G ^, on a similar L1 is equal to (or at least corresponds to) Gj-g (G2);., ^ equal to (or at least corresponds to) G ^ ^ -giG ^), 20 The signal processor shown in FIG. 1 analyzes the original signal Gq in a number of parallel output signal representing the Laplace output signals Lq, L1 ,,. L ^ 1 (which appear on the second output terminal of each of the respective pipeline interpreters 100-1 100-N of the set, respectively) plus a residual Gauss output G, T (which is applied to the first output terminal of the final interpreter 100-11 of the couple occurs,

In general, the only limitations imposed on the relative values of the respective sample clock frequencies fQ ... f ^ ^ are those indicated in FIG. However, it is usually preferred to specify values of the sample clock frequencies applied to the second input terminal of each of the respective interpreters 100-1 ... 100-N such that the respective ratios CL ^ / CLp CL ^ / CL OL ^ / CL ^ ^ can be equal to 1/2 (or an integral power of 1/2 may correspond to the number of dimensions 35 of the information component of the signal being analyzed.

This results in the analyzed frequency spectrum output of the original signal GQ in the separated, pa-8402009 - 15 reverberation frequency democrat bands of Laplace component signals Lq ...

T. are divided, which (neglecting any sampling errors, which are due to the loss of signal information caused by a reduction in the sample density or due to the addition of interfering alias frequency components) have one octave width in bandwidth for each dimension of the information component and include only those frequencies that are in the frequency spectrum of the original signal G ^ that are within this particular octave. Those frequencies 10 of the frequency spectrum of the original signal Gq, which lie below the Laplace component signal ^ with the lowest octave, are then located in the remaining Gaussian signal G ^ of the analyzed output signal,

Generally, IT is a plural integer with a given value of two or more. Sr, however, are types of information, where a relatively small determined value of H may be sufficient to analyze all of the frequencies of interest in any dimension of the frequency spectrum of the original signal GQ with sufficiently high resolution. Thus, in the case of visual images, it often appears that a value of seven for II is sufficient, so that in this case the frequencies in each dimension of the residual signal G ^ are less than 1/128 (1/21} of the highest of frequency fQ of the frequency spectrum Gq of the original signal,

In Fig. 1a, in generalized form, a digital embodiment of a first specimen of the respective sample signal interpreters 100-1 ... 100-21 of the pipeline set shown in Fig. 1 is taken off. In Fig. 1a, the first specimen embodiment of one of the interpreters 100-1 ... 100 (11-1) of the set is designated 1 GOa-K and the first specie embodiment of the immediately following representation of the set is indicated by 1Q0a - (& - 1},

The interpreter 100a-Κ comprises a digital convolution filter 102 with m taps (m being an integer multiple of three or more - preferably odd), a decimator 1G-, an expander 1C6, a digital interpolation fiter 108 with n taps (where 35 n is an integer multiple of three or more - preferably odd), a interception device 109 sa a subtractor 110. A sample-frequency clock signal ZL · (ie, the clock signal V "; , V rJ; 1, '.

O L: w i .. -J l. - 4 - which is supplied to the second input terminal of each interpreter of the set of interpreters (100a-K) as a control input signal to each of the respective elements 102, 104, 106, 108, 109 and 110 supplied from it.

5 The G ^ signal. The input to the first input terminal of the interpreter 100a-K is applied as an input signal to the convolution filter 102 and after the delay device 109 as an input signal to the subtractor 110. The sample densities shown in Figure 1a are the sample densities per dimension 10 of the information signal. In particular, the signal G ^. a sample density in each information signal dimension, which is indicated in the temporal domain by the sample frequency of the clock signal CLg of the interpreter 100a-K. Therefore, each of the samples, which includes G ^, will be affected by the filter 102.

The purpose of the convolutive filter 102 is to reduce the maximum frequency of its output signal relative to the maximum frequency of the input signal G „. thereof (as discussed above with reference to Fig. 1), As indicated in Fig. 1a? the sample density at the output of the filter 102 is still the sample frequency of CL ^,

This output from the filter 102 is applied as an input to the decimator 10k. The decimator 10U supplies at its output only certain samples (not all) of the successive samples in each dimension supplied to its input from the filter 10 10. Therefore, the sample density in each dimension at the output of the decimator 10U is reduced as compared to the sample density in that dimension at the input of the decimator 10U. More specifically, as indicated in Fig. 1a, the sample density CLg. + 1 in each dimension at the output of the decimator 10 ^ 30 is such that in the temporal domain it can be indicated by the reduced amount determined by the reduced sampling frequency of - the clock signal CL ^, which is applied to the second input terminal of the immediately following interpreter 100a (K + 1). Furthermore, the reduced sample density samples enter in each dimension of the 35 G 1 signal at the output of the decimator 10U. as indicated in the temporal domain, in phase with the occurrence of the clock signal CL ^ + ^ with the sampling frequency, applied to the second input terminal of the immediately following interpreter 100a (K + 1) at. 1a 'becomes 8402009; ♦ - 17 - G output signal from deeiaator 10k (which is the signal on the first output K.

output terminal of the interpreter 1 comprises QOa-K) applied to the first input terminal of the immediately following interpreter 100a (X + 1). Therefore, the isochronous relationship between the reduced sample density of the samples from G ^ the first input terminal comes and the kick signal CL. . with reduced sampling frequency at the second Δ. + 1 input terminal of the interpreter 1Q0a- (K + 1) corresponds to the isochronous relationship between the larger sample density of the samples of G „at the first input terminal and the higher sampling frequency of the Δ.— i 10 clock signal CL ^. on the second input terminal of the interpreter TOGa-X (described above).

Although not limited to this, a preferred embodiment of the decimator 10b is that which reduces the sample density at its input in that dimension by half in each dimension of the signal information. In this case, the decimator 10k in each dimension feeds each sample every other input to its output. Therefore, for one dimensional signal information, the sample density CL.i + 1 is equal to (1/2} 1 or half the sample density CL ^.

For two-dimensional signal information, the sample density CL ^ + 1 20 in each of the two dimensions is equal to one-half. thereby providing a two-dimensional sample density of (1/2) or one-square.

Although the baseband frequency spectrum of Cy is the same at the input of the decimator 10k and at the output of the decimator Wh, the reduced signal density L signal at the output 25 of the decimator 10h leads to the loss of a certain amount of phase information , which is present in the G ^ signal of greater pitch test density, which is applied to the input of the decimator 100K.

In addition to the first input terminal of the immediately following interpreter, they are supplied, the output of the decimator 10k is also supplied as an input to the expander 106. The expander 106 serves as an additional sample of zero (a digital a number representing a zero level) in any pitch control position of the clock signal CL lacking a sample from the output of the decimator 104. In this way, the sample density at the output of the expander 10b is returned to the sample density at the input of the decimator 104. In the preferred case, the sample density in each G 4 L L 'J 0 - 18 dimension having half is reduced, the expander 106 in each dimension introduces a zero between each pair of adjacent samples in that dimension at the output of the decimator 100i. Although the expander 106 increases the sample density at its output relative to its input, the device in no way changes the GT signal information at the Ά.

its output from that at its entrance. However, the introduction of zeros has the effect of adding images or repetitions of baseband G r signal information which occurs as CL harmonics of sideband frequency spectra.

The G1 signal at the output of the expander 106

Jx is then passed over an interpolation filter 108. The interpolation filter 108 is a low-pass filter, which transmits the baseband Gg signal, but suppresses the CL harmonics of the sideband frequency spectra. Thus, the filter 108 replaces each of the samples with a value of zero with samples of an interpolation value, each of which has a value determined by the respective values of the information carrying samples surrounding it. The effect of these samples with interpolation value is that the envelopes of the information-carrying samples are determined with a higher resolution. In this manner, the high-frequency components of the G1 signal at the output of the expander 106, which are located above the baseband, are substantially removed by the interpolation filter 108. However, the interpolation filter 108 does not add any information to the interpolated Gv. signal at its output, which is not already present in the G_r signal with reduced sample density at the output.

ÏX

of the decimator 10U, and cannot add it. In other words, the expander 106 serves to expand again the reduced sample density in each dimension of the G ^ signal 30 to the sample density in each dimension of the G ^ signal at the output of the convolution filter 102.

The subtractor 110 serves to subtract the G1 signal which occurs at the output of the interpolation filter 108 from the GT. signal, which is connected to the first input terminal of the interpretation device 35.

100ar-K is supplied and is supplied as input signal to the convolution filter 102 and via the delay device 109 to the subtractor 110. The delay device 109 provides a delay, q λ t; ·. ·, 1 · _ f "'V tL v * · - t9 - which is equal to the total delay introduced by the convolution filter 102, the decimator 104 , the expander 106 and the interpolation filter 108. Therefore, since the heather signals applied to the subtractor 110 as input signals have the same sample density CL1 and have equal delays in each dimension thereof, the subtractor 110 subtracts level represented by the digital number of each sample of the .G signal applied to it from the level represented by the digital number of the corresponding sample of the G signal applied thereto. supplied. Therefore, the output of the subtractor 110 constitutes the Laplace signal 100 which occurs at the second output terminal of the interpreter 100.

Only those signal components of G ^. Also, which are not present in the GT signal applied to the subtractor 110, will be present in the Laplace signal LT at the output of the subtractor 110. The first of these components includes the high-frequency portion of the frequency spectrum of the G1 signal located above the passband of the convolution filter 102. Thus, for example, if the interpreter 100a-K 20 corresponds to the interpreter 100-1 of Fig. 1, the first component of L., (L-.) Includes those frequencies of the frequency spectrum of G.. (Gq) within the passband f ^ to f ^. In addition to this component, the Laplace output signal L1-1 from the subtractor 110 also includes an error-compensating second component, which includes frequencies within the passband 25 of the convolution filter 102; which essentially correspond to the phase information contained in the G ^. signal with a higher sample density at the output of the convolution filter 102v which phase information is lost in the decimation process (discussed above).

Therefore, the lost phase information in the G .. signal with reduced sample density (decimated signal) fed to the first input seed of the immediately following interpreter 100a (S + 1) is substantially preserved in the Laplace signal Ly. which occurs at the second output terminal of the interpreter 100-K.

of the interpreters 100-1 ... 100 — may have the configuration 35 of the interpreter 1GOa-K of Figure 1a. In this case, the rest signal GAj. of the analyzed output signal, occurring at the first output terminal of the last interpreter 100-51 of the 8402009-20 set, have in each dimension thereof a sample density which is less than (preferably half of) the sample density in any dimension of the G ^ ^ signal added to its first input. However, since by definition no rendition 5 of the set follows the rendition 100-H, for most applications (with the exception of the compressed information transfer applications) it is not essential that the sample density of the residual signal G is smaller than the sample density of the Gjjr signal applied to the first input of the interpreter 100-W, therefore, in this case, instead of having the structure of the interpreter 100a-K, the set's final rendition 100-f also have a structure shown in Fig. 1c (although each of the other rendition members 100-1 ... 1Q0 (H-1) of the first-spec set is still constructed in the manner of the interpreter 100a-K). In Fig. 1c, the G ^ output of the convolution filter 102 (having in each dimension thereof the same sample density as the G. signal applied to the input of the convolution filter 102) is not decoder, but passed directly as the residual G 2 output from the last interpreter 100a-H of the first specie set. Since no decimation is present in this case, there is no need for expansion and interpolation. Therefore, the G1 signal at the output of the convolution filter 102 is supplied directly as the G1 input signal to the subtractor 110. In other words, the configuration of the interpreter 100a-H in FIG. 1c differs from that of the interpreter 100a-K in FIG. 1a in that the decimator 10U, the expander 106, and the interpolation filter 108 are not present. In this case, the delay device 109 provides a delay which is only equal to that introduced by the convolution filter 102,

The first specie, shown in Figure 1a (or alternatively in Figures 1a and 1c) provides a real-time realization of the Burt Pyramids algorithm. of course, in its most useful form, each of the Laplace components of the analyzed output signal from the Burt Pyramid algorithm has a bandwidth of one octave in each dimension thereof. This most useful form of the Burt Pyramid algorithm is used in the real-time realization according to Fig. 1a obtained by making the sampling frequency of the clock signal CL ^ in a small dimension equal to half the sampling frequency of the clock signal CL ^ in that dimension.

Reference is now made to another type of hierarchal pyramid 5 which is an alternative to the Burt pyramid. This alternative pyramid is referred to as a "filter subtraction decimer" (FSD) pyramid. Although the FSD pyramid does not possess some of the desired properties of the 3urt pyramid, the FSD has certain other desirable properties which the Burt pyramid does not have. For example, a desirable property of the Burt pyramid (which does not possess the FSD pyramid) the inherent condensation, in the synthesis of the rebuilt original signal, with respect to disturbing alias frequencies, which are present in each of the respective Laplace and residual components of the analyzed output signal. the applications the FSD pyramid has fewer parts and can therefore be realized cheaper than the Burt pyramid,

The signal processing apparatus of the invention, using a pipeline architecture, is also useful in providing a cross time realization of the FSD pyramid. The FSD-20 pyramid includes a second species of the structural configuration of the respective members of the set of sample signal interpreters 100-a ,,, 100-Ii, shown in Figure 1, using interpretation means or stages such as 100 ° K given in Figure 1b (instead of stages such as the Interpreters 1COa-X described above which are used in the Burt pyramid).

Interpreters 11Ob-K of Figure 1b show a digital embodiment of the above second specie, each member of the interpreters 100-1100 (2T-1} of the set of interpreters shown in Figure 1, such as 100b-1C and 100b- (X + 1) spring-30 given in Fig. 1b. Furthermore, the interpreters ICQb- (£ + 1) in Fig. 1b represent that member of the interpreters 100-1 ... 100 —1ST of the proposal immediately following the interpretation organs 100b-K.

As shown in Fig. 1b, the interpreters 10Qb-K comprise only the digital convection filter 102 with m taps, decimator 10 *, delay device 109 and subtractor 110, The constructional configuration of the second species of the interpreters 1Q0b-1 shown in Fig. 1b corresponds to the constructional configuration - 22 - of the first specimen of the interpreters 1 QOa-K (Fig. 1a) in the sense that the G „. signal (with a sample density CL„) as ür- 1 ft.

input signal to the filter 102 eh is supplied through the delay device 101 as an input signal to the subtractor 110 and that the output signal -G '(which also has a sample density with x ft a value CL) is applied over the decimator 10U to every di-ft mensie the sample density of the G 1 signal up to CLr. reduce

ft ft “M

before the G-r signal with reduced sample density is applied to the first input terminal of immediately following interpreter 100b-10 (K + 1).

The second specie of the interpreters 100b-K differs from the first specie of the interpreters 100a-KL in that at the G1 input of the subtractor 110, the G1 signal directly with the sample density CL1. (in any dimension) is supplied, which is supplied from the output of the filter 102 to the input of the decimator 10k. More specifically, it differs from the first spec of the interpreters 100a-K, using the G ^ signal with the reduced sample density CL, +1 (in each dimension) at the output of the decimator 104. Therefore required first, expander 106 and interpolation filter 108 return the sample density CL, thereof (in each dimension) to the G signal before the signal at the G input of the subtractor.

IV

device 110 is supplied. Since the GT input signal of the remote

IV

puller 110 at the second specimen of the interpreters 100b-H does not come from a decimated sample density source, there is no need for the expander 106 and the interpolation filter 108 in the configuration of the interpreters 100b-K. Therefore, in Fig. 1b, the delay device 109 provides a delay that is only equal to that introduced by the convolution filter 102. Furthermore, the L, uit output signal of the subtractor 110 includes only relatively high frequency components of the frequency spectrum of the GT. .signal, which are not also present

iVfI

in the G ^ signal at the output of the convolution filter 102,

According to the second specie, the final representation member 100-N 35 of the set may also have the constructional configuration of the interpretation member 100b-K or the constructional configuration of FIG. 1c. The respective embodiments of the first and second species, 8402009-23 shown in Fig. La 113, are digital embodiments. In such digital execution standards, an analog-to-digital converter is initially used to convert an analog signal to digital level samples, the level of each sample normally being represented by a multi-bit binary number. However, it is not essential that both the aerate and the second species according to the invention be realized in digital form. Sample signal interpretation devices using charge-coupled devices (CCD) are known per se. For example, transverse CCD filters, such as split-port filters, can be designed as convolution flashes and as interpolation filters. CCD signals include a series of discrete samples. However, each sample has an analog ampi rate level. Therefore, the invention can be applied both in digital form and in analog form.

The filter characteristics of a branched filter depend on factors such as the number of branches, the effective time delay between the branches, and the determined value levels and the polarity of the respective weighting factors, which manage individually at each of the branches. For illustrative purposes, it is assumed that the convolution filter 102 is a one-dimensional, five-branch filter. Fig. 2 shows an example of the determined value levels of weighting factors, all of which have the same polarity (positive in Fig. 2) and belong to the five individual branches, respectively. The figure also shows the effective time delay between each pair of adjacent branches. More specifically, as shown in FIG. 2, the effective time delay between each pair of adjacent taps 1 / CL ^., The sampling period is determined by the clock signal CL · ^ with the sampling frequency, which is individual to the convolution filter 1C2 of each of the interpreters 100-1 100-21 of the first 30 or the second species (indicated in Figs. 1a, 1b and 1c) is performed, Hence, the absolute value of the time delay CL1 of the convolution filter 102 of each interpreter is 100-2 100-IT greater than that of the immediately preceding interpreter of the set r 35. In Fig. 2, the weighting factors associated with the five branches have all positive polarities and certain value levels distributed symmetrically with respect to the third branch . More specifically, in the illustrative example shown in Fig. 2a, the weighting factors associated with the third branch have the determined value six, the respective weighting factors associated with each of the second and the following. fourth taps the same determined lower value four, and the weighting factors associated with each of the first and fifth taps, the even lower determined value is one. The envelope 202 of the weighting factors 200 determines the core function (and therefore the shape of the filter characteristics in the frequency domain) of the convolution filter 102 of each of the interpreters 100-1, 100-N of the set. More particularly, since all samples 200 (1) have the same polarity (positive in Fig. 2a), (2) are symmetrically located around the central (third) sample, and (3) the sample level decreases as the sample continues remote from the central sample, the convolution filter 102 has a low-pass filter characteristic in each of the respective interpreters 100-1, 100-1T of the set. Although in fig. 2 all weighting factors have the same polarity (positive) „this is not essential with a low-pass filter. Some of the weighting factors can have an opposite (negative) polarity as long as the algebraic sum of the weighting factors differs from zero. -20 function waveform (such as that of the envelope 202 of FIG. 2, for example) may be identical for all convolution filters 102 of the respective interpreters of the set, so that the relative low-pass frequency characteristics (the shape of the filter characteristics in the frequency domain) are the same for all filters 102 (although this is not essential), however, the absolute value of the nominal low-pass cutoff frequency of the filter for each individual member of the interpreters has a scale depending on the sampling frequency period 1 / CLT for that filter. The levels of

JX

suitably selecting the weighting factors 200 (which need not have the determined values 1? k and 6 shown in FIG. 2a) can be used for the signal G ^ at the output of the convolution filter 102 (which in each dimension has a sample density CL „) a nominal layer

V1 cross-cut-off frequency is obtained, which is substantially equal to half the maximum frequency (or in the case of GQ the highest possible frequency of interest fQ) of the G ^ -segral. {That at the convolution filter 102 is supplied, In this case the decimator 10U in each dimension reduces the one-dimensional pitch 8402009 - 25 - test density of the G ^ signal to CL ^ / 2 by eliminating any alternate sample in this dimension, The G „signal (determined by the sample envelope 202), however, remains essentially the same at the input and output of the decimator 10 ^ (although there is some loss of phase information due to the lower sample density at the output of the decimator IGk).

Certain preferred embodiments of the real-time realization of the Burt pyramid, which forms the first species (shown in Fig. 1a) of the genus of Fig. 1, will now be described. Referring to Fig. 3, which is a block diagram of shows a spectrum analyzer, spectrum change circuit, and signal synthesizer influencing an electrical signal representing one-dimensional information (such as any type to the time-varying information signal, for example). 3 shows the original electrical signal, the spectrum of which is to be analyzed, which signal is supplied in analog form for digitization to an analog-to-digital converter 305. The sampled digital response of the ADC 305 is indicated by.

The higher frequency response to GQJ, a high-pass spectrum Lq, 20 is subtracted in a zero-order analysis stage 310, leaving G ^ a low-pass filtered response to GQ. The higher frequency portion of G ^, a band-pass spectrum L ^ t is extracted in a first order analysis stage 315 for transferring a low-pass filtered response to G1. The higher frequency portion of G5, a band-pass spectrum below the band-pass spectrum, is extracted in a second order analysis stage 320 for leaving G ^, a low-pass filtered response to Gg. The higher frequency portion of a bandpass spectrum below the bandpass spectra and is extracted in a third order analysis stage 325 for transferring G ^, a low pass filtered response to G ^ · The higher frequency portion of Gi, a bandpass spectrum L. under the bandpass spectrum L-, 4 * 4 * j is extracted in a fourth order analysis stage 330 to leave G ^, a low-pass filtered response to G ^. The higher frequency portion of G ^ t a band pass spectrum among the other band pass spectra is extracted in a fifth order analysis stage 335 for transferring G ^. a residual low-pass filtered 84 O? o 9 - 26 r * ter response to G ^, The response Gg is essentially a six-fold low-pass filtered response to the original signal Gq.

The analysis stages 310, 315, 320, 325, 330 and 335 comprise initial low pass filter stages 311, 316, 321, 326, 331 and 5 336 respectively with successively narrower pass bands, The low pass responses of these filters 311, 316 '321, 326, 331, 336 are so much narrower than their input signals that they can be resampled at a reduced rate before being fed to the next analysis stage. The sampling is reduced by 10 on a regular basis, for example by decimation - in the decimation chains 312 # 31Ts 322, 327. 332, 337 following the respective filters 311, 316, 321, 326, 331, 336 to make a choice. In octave spectrum analysis, which is of particular utility, alternating samples are eliminated by the decimation process. The high frequency portions of the input signal applied to each analysis stage are extracted by the low frequency portions of its input signal from its input signal. to confiscate. The lower frequency decimated portion of the input signal has problems that it is undesirable in a sample pitch matrix of smaller resolution than the input signal and is undesirably delayed from the input signal. The first of these problems is solved in the expansion chains 313, 318, 323, 328, 333, 338 by introducing zeros in the missing sample pulses in the low-pass response sample matrix and then by low-pass filtering the interfering harmonic spectra, which are introduced simultaneously , to eliminate. The second of these problems is solved by delaying the input signals of the analysis stages before subtracting the expanded low-pass filter responses provided by the expansion chains 313, 318, 323, 328, 333, 338,

The delay and subtract processes are performed in the respective chains 31319, 324? 329, 334, 339 in analysis stages 310 (315, 320 # 325, 330, 3351 (in certain circumstances, as will be described later), elements can be advantageously shared between the initial low-pass filter and the delay and subtraction circuit of each analysis stage.)

The spectral analysis just described has a pipeline character 8402009 - 2T - and there is a gradually longer time shift of the samples. Lg samples, l_ samples, L ^ samples, and L ^ samples relative to the LQ samples, The "time shift" * used here refers to the different time delays with vocal definitions known between the corresponding samples of vessel information as to related, parallel signals occur - such as between the corresponding samples of the analyzed output signals LQS L, ^, Lg, and Gg of the spectrum analyzer, shown in FIG. 3. The signal synthesis is yet to be calculated. writing spectra procedures require an opposite time shift of the respective sets of samples, which can be obtained by delay lines 3b0, 341 _ 3h2 ,. 343 and 3½ (which more particularly include shift registers or some other type of memory that performs the equivalent function - for example, reading out and then registering - series 15 memory) for L ^, L ^ and L ^ samples for their modification in respective chains 3 ^ 5, 34β, 3 ^ 7, 3 ^ 3 and 3 ^ 9, as shown in Fig. 3.

The spectra can also be altered and the altered spectrum sample can be delayed thereafter. The delay can also be split before and after change in different ways - for example, to allow spectrum changes to take place in parallel with time. It is clear that the different delays in the change chains 3, 5, 346, 3 ^ 7, 3½ and 3 ^ 9 themselves can be used in certain circumstances as parts of the totally different delays, The Lj and Gg spectra are changed in the change chains 350 and 351 In some signal processing applications, it may be that certain chains of change chains 3 ^ 5 - 351 are not needed and can be replaced by respective direct connections. The spectral analysis procedures described so far can be extended if further analysis steps are used or if fewer analysis steps are used. The residual low-cross spectrum, G at the end of the spectral analysis, will not equal Gg in these cases,

In the synthesis of a signal by recombining the 35 spectrum analysis requirements, if any amended, the decimation of the sample array from analysis step-to-analysis step must be eliminated so that the spectrum samples can be summed using 8402009-28.

the options 353, 355 s 357, 359, 361, 363. This is in addition to correcting a time-fresh housing in the delay chains 3 ^ -0 - 3 ^. The decimation is eliminated using expansion chains 352, 35 ^, 356, 358, 360 and 362, which essentially correspond to respective ex-5 expansion chains 338, 333, 328, 323 f 318 and 313. By multiplexing, a single chain can fulfill double function. The remaining low-pass spectrum, G ^, is shifted forward with respect to the adjacent band-pass spectrum, L ^, such that the expansion of the sample in question centers the time in question relative to that of ^), In FIG. 3 is G_p Gg, which has been modified (new Gg,) and expanded in expansions 52 and then added in adder 353 to a changed L rs in FIG. 3), resulting in a synthesized new G ^ ( new G ^). The output of the adder 353 is expanded in the expansion circuit 35 ^ and added in the adder 355 to the delayed and modified to synthesize the new G ^.

The output of the adder 355 is expanded in the expansion circuit 35 ^ + and added in the adder 357 to the delayed and modified to synthesize the new G ^. The output of the adder 357 is added in the expansion circuit 358 expanded and added in adder 359 to the delayed and modified to synthesize the new Ggft. The output of adder 359 is expanded in expansion circuit 60 and added in adder 361 to delayed and modified L2. to build the new G1, synthetically. Finally, the output of the adder 361 in the expansion circuit 362 is expanded and added in the adder 363 to synthetically build the new Gq, the new Gq,, G,,, Gg,,, G,. Gj. (And Gg, are indicated with 30 accents in the signal synthesis circuit of Fig. 3. The new Gq can be converted into analog form by an (unshed} digital analog converter), if desired,

The expansions in chains 352, 35, 356, 358, 360, 362 provide for an elimination above the band at every step of the synthesis process.

35 Where the band-pass spectra are no wider than an octave, this suppresses any harmonics generated by the change circuits 3 ^ -5 -35-1 which might otherwise adversely affect signal synthesis by damaging the signal synthesis 8402009-29. introducing disturbing alias frequencies.

ig, 4- more explicitly shows the construction of a spectrum analysis stage for dimensional information, such as 310, 315, 320, 325, 330 or 335,5 used for spectrum analysis with octaves, The stage is the spectrum analysis stage of the S-th order, where £ is zero or a positive integer,

In the case of the zero-order spectrum analysis stage, the clock frequency for the stage will have a frequency H for sampling the original input signal, Cq, the spectrum of which must be analyzed, In the case where £ is a positive integer the clock frequency is reduced by 2,

The input signal G ^ of the spectrum analysis stage of FIG. H is applied as an input signal to a shift register V70 with M stages, which shift register is clocked at the clock frequency 15 R / 2 * ^ t. The (M + 1) samples with gradually greater delay are supplied to the outputs of the shift register h? 0. that if the delay line operates with a number of branches of a low-pass delay line filter, the samples are weighted and summed in the circuit 171 to provide samples of a low-pass filter response with linear phase. For all analysis stages, except the first one, in which stages K is greater than zero., The halved clock frequency (compared to the previous stage clock frequency) decimated in the initial shift register 70 and the add-ons in the weighting and summing circuit U71 is used with respect to G. ^, The response G <£ + ^ \ is applied as a single input signal to a multiplexer which provides a varying choice between its input signal G, ^ and an input signal zero, the switching takes place with a frequency of H / 2K, to generate a signal Cv-,,, / 30 The signal has a baseband frequency spectrum, which is twice the size of the G ^ + ^ spectrum, mixed with a first double sidebands harmonic spectrum with suppressed carrier wave with an oi channel flue. Teriooos it is noted that the next structural analysis stage can use appropriately damped instead of 35 a signal. The Gj-+ signal is input as input signal to a shift register 1? 3 with a number of stages (which may be the same or different from M) and with a frequency of R / 2. clocked, The (M + 1) samples supplied by the input signal and the output signals of each of the stages of the shift register 473 are applied to yet another sweeping and summing circuit 9 corresponding to the circuit 471, The circuit 474 suppresses the former 5th harmonic spectrum of and delivers an expanded version of j to a sample matrix with as many samples as the sample matrix of G ^,

In an addition chain 475, this expanded version of Gv. subtracted from G ^ after G ^ has been delayed in shift register 470 and delay chain kj6. The delay of M periods of G1 in the shift register 470 compensates for the delay of M / 2 periods of the central sample for the weight and summing circuit 471 and for the G1 input signal at the spectrum analysis stage of FIG. 1 *, respectively. and the similar delay of M / 2 periods between 15 G ^ g + .yyK and the central sample for the weight and summing circuit 474. The delay device kj6 introduces a delay to compensate for the delays in adding the weighting sum and summing circuits 471 and 474, and the delay device 476 can be easily obtained by extending the shift register 470 by the required number of further stages. The output signal.

Dg, of the adder circuit 475 is one of the searched spectrum analysis components, the lower frequency limit of which is set by the low-pass filtering f which takes place in the K-th spectrum analysis stage shown in Fig. 4 and whose upper frequency limit is set by the low-pass filtering of the preceding spectrum analysis stage, if any,

Fig. 5 shows a way to reduce the number of shift register stages used in a spectrum analyzer according to the invention. The samples for x-stacking must be weighted and summed to obtain the low-pass filtering associated with the interpolations of G ^ + i ^, are obtained from the scanned delay line system, which is used for the initial low-pass filtering of G ^ + 1 \ in the next spectrum analysis stage instead of using a shift register 473.

FIG. 5 shows by way of example how this is done, for example between the zero-order analysis stage used to generate Lq and the next analysis stage. Elements 8402009 - 31 - 570-Q, 571-0, 575-0, and 576-0 are those elements in the zero order split-accuracy analysis stage, which correspond to elements 470, 471, 465, and 476 of the X-order spectrum analysis step according to FIG. 4.

Elements 570- * and 571-1 of the first-order spectrum analysis stage are analogous to elements 570-0 and 571-0 of the zero-order spectrum analysis stage except for being clocked at half frequency, The four samples from the input and the first three outputs of the shift register 570-1 will be supplied in parallel with the clock frequency 3/2, they will be interleaved with zeros and the results will be weighted in two phases by the seven filter weight cartridge A3CDC3A for forming the pair of consecutive samples to be subtracted by the clock frequency 3 from the delayed GQ in the subtractor 575-0,

The previous sample of each pair of consecutive samples, which must be subtracted from the delayed GQ, is obtained by multiplying the input signal of the shift register 570-1 and its first three outputs by filter weights A, G, C and A in weighing chains 5θ0 , 531, 532 and 583 and then sum the weighted samples in the summing circuit 537, The interleaved zeros occur at 2G in points, which must be weighted with B, D, 3 for this positioning of Gj relative to the filter weight pattern, The last sampling of each pair of consecutive samples to be subtracted from the delayed Gq is obtained by the input signal of the shift register 570-1 and its first two inputs with the filter weights 3, D and 3 in the weighing chains 534, 585 and 536 multiply and then sum the weighted samples in the summing chain 588, The interleaved zeros occur in points, which must be weighted eyes with A, C, C, A for this positioning of G, relative to the filter weighing pattern, Sen multiplexer 539, which operates with the clock frequency 3, alternately selects between the samples at the outputs of the summing cats 537 and 588 to provide from the flow of samples subtracted from the delayed Gq in subtractor 575-0,

Fig. 6 shows in more detail a stage of the signal synthesizer of FIG. 3, Samples of G ^ (or delayed and modified G r>) are zeroed in a multiplexer 692 and the resulting expanded signal becomes as an input signal 3402009 - . 32 - applied to a shift register 693 with M (or other multiple) stages and clocked at the expanded sampling frequency, The input signal of the shift register 693 and the outputs of each of its stages are applied to the weight and sum rasta & 9b, Het 5 (or 8: J spectrum? if sampled again at the double frequency and then freed from harmonics, it is then fed from the weighing and summing circuit 69 ^ to an adder 695 to be combined with the modified j ^ delayed in time. for centering with the re-sampled and filtered G ^ ((or G q) samples, adding them to this. The multiplexer 692, the shift register 693 and the weighting and summing chain 69 ^ can be multiplexed? subject to serve as the elements bj2 473 and 474 in the spectrum analysis process,

It is worth considering at this point the characteristics of the low-pass filtering to be used in the low-pass filtering step of the spectrum analysis procedure and in the expansion steps of the spectrum analysis and signal synthesis procedure. The low-pass phasing has a linear phase, so that the pattern of filter weights is symmetrical about the central sample (tests). The filter weights together are equal to the unit for the low frequencies in the high-pass spectrum Lq and suppress the band pass spec tra L ^, L · ^, L ^, ... as much as possible. If the spectrum analysis is to be done in octaves, the decimation of which takes place in the re-recording of the lower band removed by low-pass filtering in each spectrum analysis stage, the frequencies below two-thirds of the octave center frequency during the low-pass filtering are desired. , A step frequency response in the filter (called a "brick wall" *) introduces an overshoot in the filtered signals, reducing the dynamic range of both the ¢ (£ +]) function, obtained by the spectrum analysis stage functions obtained by subtracting the expanded G ^ ^ j from is increased Bit is an example of the Gibbs phenomenon, which can be modeled using a less abrupt termination of the Fourier series, A number of truncation windows, which deliver in filter response with reduced Gibbs phenomenon is known; for example, those which can be assigned to Bartlettaan Hanning to Hamming, to 31ackman and to Kaiser. By way of example, reference is made to section 5.5 of 8402009-33 - the book "DIGITAL SIGNAL PROCESSING" by A. 7. Oppenheim and RW Schafer, published by Printice-Hall Inc., Englewood Cliffs, NJ in 1975, which section is titled "Design of PIS Filters Using Windows" and which section can be found on p. 239 - 251.

In practice, a number of samples in low-pass fixing are usually limited to only a few. In a filter using an odd number of samples, the filter response will include a DC component and a series of cosine harmonics, and in a filter using an even number of 10 samples, the filter response will include a DC component and a series of sine harmonics. The desired response curve is most closely approximated using a calculator to attempt to make a selection of weight coefficients.

Set is possible to obtain equal Q spectra of non-octave widths according to the invention, although such an approach appears to have limited utility. Decimating the low-pass filter response to choose every third sample and filtering out frequencies below half the band-pass center frequency to obtain the low-pass response produces a set of band-pass spectra, the bandwidth decreases successively by one-third instead of, for example, a half,

The sample change chains 3 ^ 5 - 351 of FIG. 3 can take a wide variety of forms, and certain chains thereof can be replaced by direct throughput. For eliminating low level background noise in the different spectra, each of the change chains 3 ^ +5 - 351 include, for example, a respective base line cutter J0 of FIG. Such a cut-off device 7 can be simple with a truncation of the less significant bits of the signal, FIG. 8 shows a circuit which can be used for each of the change circuits 3-5351 to provide a spectrum equalizer. A rotary switch 897 is wired to provide binary code for each of a number of axis displacements.

This code is applied through a multiplier 898 to a two-quadrant multiplier to multiply input spectrum samples to generate output spectrum samples, which must be synthetically constructed to provide 8402009-3k ^.

The latch 898 maintains the code input of the multiplier 889, while the setting of the rotary switch 897 can be changed. One can make sure that each of the octave spectra is subdivided, using digital filters: in which the same 5 sampling frequency is used as that used to form an octave spectrum, or a halved sample, after which the spectral subdivision gains are individually determined. be set. A subdivision of the octaves ώ. Twelve provides individual tone and hal o settings and signals for encoding 10 music, for example,

The change chains may consist of dead memories (ROMs) for storing non-linear transfer functions. For example, a ROM 990 in which a logarithmic response to the input signal of FIG. 9 is stored can be used in any of the transducer sampling circuits 3-5351 and a ROM 10919 in which an exponential response can be stored on the input signal of FIG. 10 are applied in each of the corresponding sample change circuits of a receiver, thereby providing a push-up of the signal before transmission and a push-back after reception. Other complementary push-up and push-back characteristics may be alternately stored in the ROM change chains of the transmitting and receiving spectrum analyzer signal synthesizers,

Fig. 11 shows a modification of the spectrum analysis and signal synthesis system of Fig. 3, wherein the delays between analysis and synthesis are divided to provide spectral samples without time shift for processing. Such centering is desired, for example, in a z, g, expansion system, where spectrum analysis is used to separate signals for expansion into spectra, so that the expanded spectra can be filtered to suppress distortions generated during rapid signal compression or expansion. The amplitude of the original signal applied to the ADC 305 of FIG. 3. can be detected to derive in the circuit 1.130 a compensation control signal CC, which is applied to each of the expansion devices 1110, 1111, 1112, 1113, 1115, 1115, 1116 to provide a fast and slow tapered compensation of the signals which these devices compose. The compensators 1111 - 1116 may consist in fibers of two quadrant digital multipliers, the control signal CC being generated by an analog-to-digital converter cascaded with conventional analog circuits for just detecting the signal to be compressed and responding to generating this detection of an analog compensation control signal.

The expansion devices 1110, 1111, 1112, 1113, 111k, 1115 and 1116 affect the LQ, L, s Lg, L · ^. L, and Cg spectra after they have been delayed differently using the delay TO chains 1100, 1101, 1102, 1103, 11Cl, and 1106 to center their respective samples to time, Be delay chains 1120, 1121, 1122, 1123, 112k, and 1125 then shift the comandated LQ, 5 I * 2, * 1 ^,, L ^, and Gg, signals appropriately for the signal synthesis procedure using elements 352-3 ^ 3 ^ of fig. 3,

The delays in the delay circuits 1106 and 1125 are essentially M / 2 periods at the frequency of R / 2; where E equals five, or 16 M, periods of the base clock frequency R, which is delayed in the construction of the samples for the weight and sum 20 chain kjU of the last spectrum analysis stage 335 · These 1oM delay periods are increased with a delay time D ^ to accommodate the addition times in the expansion circuits 338 and 352 and by a delay time Dg to accommodate the addition times in the delay and subtract chain 33¼ and in the adder 353. It is assumed that all addition processes are performed at the base clock frequency R, and that C'i and Dg are expressed as numbers of these clock periods.

The delay in the delay chain 110¼ will be greater than 16m + D, + D2 periods at the clock frequency R by the difference between the time required to generate L- from G ^ ° P, and the time required 30 to I from G, - to generate, The time required cm. L, - to be generated from G_ is M periods at the clock frequency R / 2 to collect samples for weighing and summing twice, or 32 A periods at the base clock frequency, plus 2D ^ for two sets of sample summation. plus Dp for sample subtraction, The time required to generate L ^ from G_ - I4. ^ 35 is M / 2 periods at the clock frequency of R / 2 for sample collection for weighting and summing, or 8 M periods at the base clock frequency, plus D.j is for sample summing. plus for stitch deduction 8402009 - 36 - * ψ. 2 ^ M + D ^ periods at the base clock frequency additional delay are required to center L ^ samples in time with L. samples. Therefore, the delay circuit 10U will have a total delay of h-0 M + 2D ^ + periods at the base clock frequency R.

Similar calculations determine that the periods at the base clock frequency R by which the samples in delay circuits 103 102, 101 and 100 are to be delayed are 52M + 3D1 + Dg, respectively. 5ÖM + ^ + Dg, 6m + SD1 - Dg and (β2 |) Μ + 6ö1 + Dg.

The delay required at the delay circuit 112i in addition to that provided by the delay circuit 1125 is the time required for expansion in the circuit 35 * + and the Dg delay associated with the addition in the counter 55. The first delay is 1; M / 2 periods at the clock frequency of R / 2, needed to collect samples for weighing and summing. 8M periods at the base clock-15 frequency R, plus D1 associated with the summing in the weighing and summing process. The total delay in the delay chain 112U is then 2kU + D1 + Dg. By similar calculations, the total delays in the delay chains 1123, 1122, 1121, and 1120 in terms of periods at the base clock frequency R are 28M + 3D1 + 3Dg, 20OM + UDj + * + Dg, respectively. 31M + 5Dt + 5Dg and (31g) M + 601 + 60g.

Similar calculations can be used to determine the total delays in the delay circuits 3Uo-3Uh of FIG. 3. assuming that the modification circuits 3 * + 5 - 351 all have the same delays. The delay chains 3 * + 0, 3 * + 1, 3 * + 2, 3 * + 3, 25 3W + and 3 * + 5 respectively have delays in periods at the base clock frequency R of 77M + 12D ^ + 7Dg, 76m + 10D ^ + oDg, 72M + 8d1 + 5Dg. 6Um "+ 6d + tog and U8m + UD1 + 3Dg.

The digital filtering employed in the spectrum analyzer is a species of a hierarchal filtering of general interest in that the low-pass or band-pass filtering extends over many samples. takes place such that relatively small numbers of samples are weighed and summed at a given time.

Although the invention is applicable to use the spectrum of a signal representing one-dimensional information, the Burt pyramid was developed to substantially analyze the spatial frequencies of two-dimensional image information. The invention allows for real-time spectral analysis of the spatial frequencies 8402009 * - of changing image information as it occurs in successive video frames of a television display.

As is known in the field of television, consecutive video frames (of the NTSC type} occur consecutively with a ras frequency of 30 frames per sec. Each frame comprises a frame of 525 interlaced horizontal scan lines. The successive odd-numbered horizontal scan lines of a grid are sequentially transferred during a first field period The consecutive even-numbered scanning lines of a grid are averaged sequentially over a second field period following the first field period, followed by the first field period of the next grid. duration of each field period is 1/60 sec. Sr, however, a collection must be applied for at least the number of pixels in a field time in order to be able to define the full spatial frequency spectrum of the image in delayed real time.

A method, known as progressive scanning, is known in television technology to extract from a ZIT C-vi the successively full frames of 525 lines at a rate of 0 frames per second. to confiscate. This method includes the delay of each consecutive HTSC-20 field over a 1/60 sec field period. Therefore, the consecutive scan lines of a consecutive odd field are interleaved with the consecutive scan lines of an immediately preceding even field delayed by one field period to provide a full frame of image pixels during the then-occurring odd field of each of successive grids. Similarly, the consecutive scan lines of an instantaneous even field are interleaved with the consecutive scan lines of an immediately preceding odd field that has been delayed by a field period, to provide a full frame of 30 pixels during the currently occurring provide even field period of each of the successive grids,

The progressive scan method. described above is of particular use in obtaining high resolution image displays in what is known as high definition television (HDTV), 35 'currently being developed on the television field. The invention is also useful in HDTV for providing improved picture displays. ? ig. 72 shows a spectrum analyzer according to the invention for 8402009 *> -38.

influencing signals representing two-dimensional information, such as the spatial frequency image information contained in successive progressively scanned television video frames. However, such two-dimensional information can also be obtained from non-interlaced television chambers, or from a line-interlaced television camera followed by a suitable buffer memory.

A monochrome processing of luminance signals will be described for the sake of simplicity with reference to Fig. 12, but the methods to be described can be applied individually to the primary colors of color television signals or to signals generated therefrom by an algebraic matrix operation. An original. video signal is sent in raster scan format to an analog-to-digital converter 120? for sampling, if the signals have not been sampled, for resampling if a sampling has already taken place, and for final digitization. The digitized video samples are designated as a signal by Gq and contain the full two-dimensional spatial frequency spectrum of the original signal and its harmonic spectra. which can be assigned to the sampling processes. These harmonic spectra are symmetrical with respect to respective spectra with the sampling frequency and its harmonics. These harmonic spectra will be discussed separately in the following explanation: The general fact of their existence is mentioned, because the harmonic spectra must be taken into account in the design of the two-dimensional low-pass filters with spatial frequency, which are analyzed in the regret at or according to Fig. 12, due to the fact that these harmonic spectra give rise to alias frequencies during spectral analysis and during signal synthesis from spectral analyzes.

In the zero-order spectral analysis step 1210, a high-pass spectrum Lq is separated from Gq. The high-pass operation is performed essentially by low-pass filtering of Gq. delay of Gq relative to its timing from the ADC1205 to the same extent that the lower frequency portions of Gn are delayed at the low-pass filter response, and subtract the low-pass filter response from the delayed Gq, when assumed. that the spectral analysis takes place with octaves, the 8402009 "- 39" cut-off frequency in the livestock-dimensional low-pass filter 1211 non-spatial frequency is chosen such that it is equal to the "upper frequency of the next transmitted spectrum with an octave hand width to be analyzed - ie four-thirds 5 of the center frequency. In the decimator 1212, alternate sample rows and columns are eliminated to sample the low-pass filtered G ^ at the spatial frequency B / 2, providing this reduced sample rate signal as a low-pass output response of the stage 1210 for further Spectral analysis. The low pass filter GQ at the reduced sampling frequency is then interpolated according to the methods indicated by RW Schafer and LR Habiner in the article "VA. Digital Signal Processing Approach to Interpolation ". page 692 - T02 in PROCEEDINGS OF THE IEEE, vol. 61, Ho. June 6, 15 19T3. In the expansion circuit 1213, the samples eliminated in the decimator 1212 are replaced by zeros to provide an Input for another two-dimensional spatial frequency low-pass filter 121U. In this filter, the same sample weighting coefficients as with the initial low-pass filter can be used, but in any case, this filter has substantially the same cutoff frequency as the initial low-pass filter. The resulting signal has a sample matrix, which is co-extensive with that of G ^ ra delayed in the delay circuit 12151 and subtracted from the delayed Gq in the subtractor 1216 to obtain the high-pass output response is not just the high-pass portion of G0. but also includes lower frequency phase error correction terms, as discussed above, which are to be used during the synthesis of the video signal from spectral analyzes for the complex response of the errors, which are sampled again by G30 at the lower sampling frequency in the decimator 12 are introduced.

This split the signal into a low-pass portion. resampling at half frequency and a high-pass portion is altered in each spruce analysis stage. Each successive spectral analysis delay receives as its input signal the resampled low-pass output response of the previous spectral analysis stage, the sampling frequency in each subsequent spectrometer stage being halved with respect to the frequency in the pre-8402009 - Uo - descending stage trumanal. high-pass output response of each sweetruan analysis stage 1220, 1230, 12 ^ 0, 1250, 1260 after the initial stage 1210 has an upper limit imposed by the low-pass response characteristic of the previous stage, so that it allows "high-pass!" output responses in reality are bandpass spectra with equal Q and decreasing spatial frequency, The decimation of the responses of the initial low-pass filters in each stage by a factor of two> and the cut-off frequency of the low-pass filters in each stage equals two-thirds of the center frequency of the spectrum Lysis obtained thereby are the factors which cause these spectra to be equal Q decreasing octaves of the two-dimensional spatial frequency,

The decimated low-pass output response of beat trumanalysis stage 1210 is fed from the associated deimator 1212 as input 15 signal to the next bitruman analysis stage 1220. The spectrum analysis stage 1220 includes elements 1221t 1222, 1223, 122¼, and 1226, analogous to the respective elements 1211, 1212, 1213, 121¼, 1215, and 1216 of the spit rum analysis stage 1210; the differences in operation are due to the fact that the sampling frequencies in the stage 1220 are halved in both 20 dimensions relative to the stage 1210. The low-pass flashes 1221 and 122¼ have weighting coefficients corresponding to those of the low-pass filters 1211, respectively. 121¼, but halving the sampling frequency at stage 1220, compared to stage 1210. Halves the cutoff frequencies of filters 1221 and 122¼ compared to filters 1211 and 121¼. The delay for subtraction in delay chain 1225 is twice as long as in delay chain 1215; when these delays are assumed to be clocked delays in a shift register or the like, the delay schemes correspond, the delay ratio of 2: 1 being obtained by the ratio of 1: 2 of the respective delay clock frequencies in the delay circuit 1225 and the delay circuit 1215. The high-pass output response of the sweetener analysis stage 1220 is a bandpass spectrum of spatial frequencies immediately below the spectrum.

The decimated low-pass output response of the spectrum analysis stage 1220 is applied from the associated deimator 1222 as an input signal to the next stage tru analysis stage 1230. The 8402009 -1 * 1 - bandpass spectrim L., which is one octave below. is the high-pass output response of the spectrum analysis stage 1230 to the corresponding input signal G ^. The spectrum analysis stage 1230 includes elements 1231, 1232, 1233. 123 **, 125 and 1236, which correspond to elements 1221, 1222, respectively; 1223, 1221 *, 1225 and 1226 of the spectrum analysis stage 1220 with the exception of the halved sampling frequencies.

The decimated low-pass output response G ^ of the spectrum analysis stage 1230 is input from its decimator 1232 as an input signal to the next spectrum analysis stage 121 * 0. The bandpass spectrum, which is one octave below. is the high-pass output response of the spectrum analysis stage 121 * 0 to its input signal G ^. The spectrum analysis stage 121 * 0 includes elements 12 ** 1_ 12k2w 121 * 3 # 12W *, 121 * 5 and 121 * 6, which correspond to elements 1231 # 1232, 1233, 123 **, respectively. 1235 and 1236 of the spectrum analysis step 15 1230, except for the halved sampling frequencies.

The decimated low-pass output response G ^ of the spectrum analysis stage 12 ** 0 is applied from the associated decimator 12 ** 2 as an input signal to the next spectrum analysis stage 1250.

The band pass spectrum L ^, which is one octave below L ^. is the high-pass output response of the spectrum analysis stage 1250 to its input signal G ^. The spectrum analysis stage 1250 includes elements 1251 1252, 1253, 125 ** 1255 and 1256, which correspond respectively to elements 12 ** 1 # 121 * 2, 121 * 3, 12l * l * „12l * 5 and 121 * 6 of the spectrum analysis step 12i * 0. excluding the halved sample rates.

The decimated low-pass output response G_ of the spectrum analysis stage 1250 is supplied from the associated decimator 1252 as an input signal to the next spectrum analysis stage 1260. The band-pass spectrum that is one octave below L <is the high-pass output response of the spectrum analysis stage 1260 to its input signal-n-flfti G_. The spectrum analysis stage 1260 includes elements 126l. 1262, 1263, 126 **, 1265, and 1266 corresponding to elements 1251, 1252, 1253, 1254r, 1255, and 1256, respectively, of the spectrum analysis stage 1250, except for the halved sample rates.

The decimated low-pass output response G, which is supplied from the decimator of the final spectrum analysis stage, where Gn here is G ^, which originates from the decimator 12Ö2 of the spectrum analysis stage 1260 is a residual layer-law spectral response »8402009 - U2 -, t

This serves as the basis for the renewed synthetic build-up signals by summing interpolated band-pass spectral responses from the later spectrum analysis stages and the "capstone" high-spectrum response from the initial spectrum analysis stage. Lq,, L ^, L ^ L ^ 5 and L ,. have shifted over time and are being delivered with increasing delay. The remaining low-pass spectrum G (here Gg) precedes the last band-pass spectrum L ^ (here L · ^) in time with an opposite time shift.

As will be described below, iterative methods of signal synthesis from spectral components also require that the spectral components Lq. , L ^, L ^: and have this oppositely directed time shift relative to each other. Before describing the processing of spectral analyzes and the synthetic construction of signals from the processed spectral analyzes, a more detailed description of the structure of the spectrum analysis steps follows. Attention will first be paid to the initial two-dimensional low-pass filter systems.

As is known in the field of filters, two-dimensional filter systems can be either inseparable in nature or also separable in nature. Separable filtering in first and second dimensions can be achieved by first filtering in a first direction using a first. one-dimensional filter and then filtering in a second direction perpendicular to the first direction using a second dimensional filter, Since the respective low-pass filter characteristics of the two separate, one-dimensional filters cascaded to form a separable two-dimensional low-pass filter late filter, be completely independent of each other. therefore, the core function and structure of each of these low-pass filters may correspond to that described above with reference to Figures 2a and 2b and Figures 3-11.

In the case of television images, which include a grid of horizontal scanning lines, the two orthogonal directions of a separable filter are preferably horizontal and vertical. If separable two-dimensional low-pass filtering is used in realizing the invention, certain advantages can be obtained by performing the horizontal low-pass filtering prior to the vertical low-pass filtering, while other advantages can be obtained by the vertical low-pass filtering. for horizontal low-pass filtering. For example, when horizontal filtering and decimation is first performed, the number of pixel samples per horizontal scanning line to be affected by the vertical core-5 function is reduced by half during the subsequent vertical filtering. However, when performing the vertical filtering first, it is possible to use the same delay system as the system necessary to provide the relatively large delay required for vertical filtering, 10 and also to provide the respective compensatory delays ( 1215, 1225, 1235, 1245, 1255, and 1265) to the respective signals - G ^ to the positive terminal of each of the respective subtractors 1216, 1226, 1236, 1246, 1256, and 1266 of stages 1210, 1220, 1230.

1240, 1250, and 1260 of the accelerant analyzer shown in FIG. 12.

The overall filter responses of separable two-dimensional spatial frequency flashes may be square or rectangular in cross section parallel to the spatial frequency plane. The filter responses of inseparable filters may, however, have other cross sections. Circular and elliptical cross-sections are of particular interest for filtering field-scan television signals since filters with responses with such cross-sections can be used to reduce the excessive diagonal resolution in the television signals. For example, uniformity of the image resolution in all directions is important in television systems where the image must be rotated between the camera and the spring display device.

Below is a matrix of filter weights with a pattern showing quadrantal symmetry and linear phase response filter characteristics. which are particularly suitable for use with the 2-D low-pass filters 1211, 1221, 1231, 1241, 1251 and 1261 and the 2-D low-pass filters 1214, 1224, 1234, 1244, 1254 and 1264 of Fig. 12, A 3 C 3 ADSF Ξ D 7 GEJH 5

D Ξ F E D A 3 C 3 A

8402009 4 - Owl · -

A core function matrix with this pattern of weighting factors successively affects each of successive image samples. each pixel sample, when affected, location wise corresponding to the centrally located weighting factor J of the matrix. In a low pass filter, the weighting factor J has the highest relative value level and each of the other weighting factors has a value level which decreases the further away from the central position. Therefore, the angular weighting factors A have the lowest value level,

In the case of a non-separable two-dimensional filter 10, the determined selected level values A, Bj C, D, K, F, G, H and J are completely independent of each other, In the case of a two-dimensional ion separable filter however. since the level values of the weighting factors are a result of the cross product of the respective values of the horizontal and vertical one-dimensional core weighting factors, the respective values of A, B, CL D, E, F, G. S and J not completely independent of each other.

An apparatus for synthetically building an electrical signal from component spectra, which may have the general form shown in Fig. 13, is important to the invention. The spectrum components Gg ', L ^', L ^!, Dg1 ',: 1 and are responses to the un-accentuated counterparts thereof, which come from the spectrum analyzer of Figure 12, The spectrum components LQ. L · ^,

Lg,} Gg and are time progressively supplied later by the spectrum analyzer of FIG. 12 and must be delayed in several ways to make GQ ", L ^", Dg "," and L0 "progressive later for the signal synthesizer of FIG. 13.

Fig. 13 shows a signal synthesizer with a number of consecutive signal synthesis stages 1360 1365 1370, 1375. 1385. Each stage 30 interpolates to expand the sample matrix of a spectral component to make it coextensive with that of the spectral component. which in addition has a higher spatial frequency and allows the component to be added to this spectral component.

The expansion of the sample matrix is accomplished by overshooting the sample points 35 in the matrix with zeros and subjecting the result to low-pass filtering to remove harmonics. Low-pass filtering preferably exhibits the same filter characteristics as the low-pass filtering. which belongs to the corresponding interpolative process to the spectrum analyzer of FIG. 12.

The low-pass filtering associated with the interpolation in the signal synthesizer. suppresses harmonics associated with the G ^ or 5 1 ^ signals. which are altered by a non-linear process which may occur with change chains (as described above with reference to Fig. 3) present between the spectrum analyzer of Fig. 12 and the synthesizer of Fig. 13. Such non-linear processes give rise to visible alias artifacts 10 in the synthetically constructed composite image when the low-pass filtering is not used, associated with the interpolative processes used in the signal synthesizer.

In the synthesizer of FIG. 13, samples of the low-pass spectrum Gg? in the expansion circuit 1361 zeroed 15 and passed over a two-dimensional low-pass filter 1362 with spatial frequency corresponding to the filter 1265 of the spectrum analyzer of FIG. 12. Samples of the response of the filter 1362 are added to samples of V0Qr in an adder 1363. kg ·); providing G_1 _ corresponding to or identical 5 "20 v to a hypothetical time-delayed replica of G_. Thereafter, random samples in the expansion chain 1366 are interleaved with zeros. This signal is passed through a low-pass filter 1367, corresponding to the low-pass filter 125½ of FIG. 12, and added in one adder 1368 to provide G ^; . corresponding to or identical to a time-delayed replica of G ^.

The samples of G ^ j are zeroed in the expansion chain 1371 and the result is subjected to a low-pass filter in the filter 1372, corresponding to the filter 12½½ of Fig. 12. The response of the filter 1372 is added to L ^ in an adder 1373. for providing G ^? matching or identical to a delayed replica of. The samples of! zeros in the expansion circuit 1376 and the result is subjected to a low-pass filter in a filter 1377 corresponding to the filter 123½ of Fig. 12. The response of the filter is 1377

3C

J is added in an adder 1373 to provide Gg * j corresponding to or identical to a delayed replica of G ^. Between the G ^ samples, zeros are introduced in the expansion chain 1331 8402009 * - 1 + 6 - * and the result is subjected to low-pass filtering in a filter 1382. The response of the filter 1382 is added in an adder 1383 to provide Gj1 corresponding to or identical to. G1 with delay. The 5 samples of 'are applied for interpolation to an expansion circuit 1386 and a low-pass filter 138T, corresponding to the filter' 1211+ of Figure 12. The response of the filter 1387 is summed in an adder 1388 with Lq 'for providing G ^ the synthetically constructed signal, which describes the same image as that produced by Gq, optionally with modifications.

Although the two-dimensional realization according to the invention is particularly suitable for use in image processing of the spatial frequency spectrum of real-time images, it is clear that the two-dimensional information concerned with the invention is not limited to the spatial frequency spectrum of two-dimensional images. For example, one of the two dimensions may correspond to spatial frequency information and the other of the two dimensions may correspond to temporal frequency information.

Furthermore, the invention is useful for analyzing the real-time frequency spectrum of information defined by more than two dimensions. For example, in the case of three-dimensional information, all three dimensions may correspond to spatial information, or two of the dimensions may correspond to spatial information, while the third dimension may correspond to temporal information. Of interest in this case, the image processing device, which responds to the occurrence of a movement in a displayed television picture. In this case, the portion of the spatial frequency spectrum of the displayed picture corresponding to stationary objects remains the same of video frame to frame of the video information, while the portion of the spatial frequency spectrum of the displayed image corresponding to moving objects changes from frame to frame of video information. A spectrum analyzer according to the invention can be used in such an image processing apparatus using 3-D low-pass filters. Two of the three dimensions of these low-pass filters are spatial and correspond to the two spatial dimensions of the 2-D low-pass filters used in each stage 8402009 - ^ 7-- ........

of the two-dimensional spectrum analyzer shown in FIG. 12. The third dimension is temporal and corresponds to the fine-structure characteristics of the three-dimensional spectrum due to changes caused by moving objects in the values of the value levels of corresponding pixels of the displayed image from frame to frame.

In the above description of embodiments of the invention, it is assumed that the temporal signal G-1 is a baseband signal with a frequency spectrum that determines information of one or more dimensions. As is known, such baseband information is often communicated in frequency-multiplex format, the baseband information comprising the sidebands of a carrier frequency modulated by a baseband information component. By using suitable modulators and demodulator and in the respective interpreters 100-1 ... 100-IT of FIG. can Gq and / or any of .., Gjj. and / or any of Lq ... L ^ ^ frequency are multiplex signals.

The term! * Shift register "should also include those means that perform an equivalent function - for example, a first read and then record serial memory.

3402009

Claims (29)

A signal processing device for analyzing the frequency spectrum of an information component of a given temporal signal in (N + 1) separated frequency bands, the component corresponding to information of a given number of dimensions, where N is a plural plural number and the highest frequency of interest in the frequency spectrum does not exceed the frequency with the characteristic. that to analyze the frequency spectrum in delayed real time the device is provided with a "pipeline", (fig. 1, 1a, 1b) provided with a set of N ordinally arranged sample signal interpretation units (100-1 ... 100N) each of the interpreters (Fig. 1a) having first and second input terminals and first and second output terminals, the first input terminal of the first interpreter of the set being for receiving the determined temporal signal (GQ) f the first input terminal of each of the second through 15th set of interpreters is coupled to the first output terminal of the immediately preceding member of the set interpreters and from each of the interpreters to the immediately following supply the signal to the interpretation means of the set (G ^. G ^ etc.), and the second input terminal of each 20 of the interpretation means of the set is intended to receive n of a separate sample-frequency clock signal (CL1-CL2, etc.) to provide respective signals sampled at a frequency equal to the sample frequency of the supplied clock-sew at the first and second output terminals of that interpreter, wherein each of the interpreters of the set between the first input terminal and its first output terminal thereof for the information component exhibits a low-pass transfer function, which low-pass transfer function of the interpreter of the set has a nominal cut-off frequency which is a direct function of the sample frequency of the clock signal applied to the second input terminal of that member of the set's interpreters, the clock signal applied to the second input terminal of the first interpreter of the set having a sampling frequency which (a) is twice as is large if and (b) for the information 35 comp onent provides a nominal cutoff frequency for the low-pass 8402009 V - U9 - let transfer 3 function. the first interpreter of the steit which is smaller than fg. wherein the clock signal applied to the second input terminal of each of the second to Nth interpreters of the set has a sampling frequency. which (a) is less than the clock frequency which is applied to the second input terminal of the immediately preceding member of the interpreters of the set, (b) is at least equal to twice the maximum frequency of the information component of the signal applied to its first input terminal, and (c) for its low-pass transfer function provides a nominal cutoff frequency less than that of the immediately preceding set of the set, and the information component of the signal, which on the second output terminal of each of the sets of the set corresponds to the difference between the information component of the sig-15 sewing applied to the first input terminal thereof, and a direct function of the information component of the signal, which occurs at the first output terminal thereof, so that the (ΒΓ + 1) separated frequency bands the N respective signals on the two the output terminals of the H interpreters, together with the signal on the first output terminal of the interpreter, comprise, Device according to claim 1, characterized in that the clock signal which is applied to the second input terminal of each of the second to BT interpreters of the set, relative to the sampling frequency of the clock signal which is applied to the second input - The terminal of the immediately preceding member of the interpreting members of the set is supplied, the sampling frequency having such that each dimension of the information component of the signal which is supplied to the first terminal thereof is sampled at half the frequency with which the corresponding dimension of the information component of the signal applied to the first terminal of the immediately preceding member of the interpreters of the set is sampled. 3. Device according to claim 2, characterized in that the clock signal which is applied to the second input terminal of every second to the interpreter of the set with respect to the sampling frequency of the clock signal which is applied to the second input terminal of the set. the immediately preceding member of the interpreters of the 8402009 4 ~ 50 set is fed with a sampling frequency such that for its low-pass transfer function, for each dimension of the information component of the signal; supplied to its first clamp, a nominal pinch-off frequency is provided. which is substantially half of the nominal cutoff frequency occurring for the. corresponding dimension of this information component by the low-pass transfer function of the immediately preceding member of the interpreters of the set, whereby the information component of the signal on the second output terminal of each of. the interpreters of the set in each dimension thereof correspond to a different octave of the frequency spectrum of the information component of. the determined temporal signal in that dimension. Apparatus according to claim 1, characterized in that the determined temporal signal is an analog signal comprising an information component corresponding to one dimensional information. Device according to claim 1, characterized in that the temporal signal comprises a video signal which determines two-dimensional image information. Apparatus according to claim 5, characterized in that the video signal-2Q sew corresponds to successive grids of scanned television images. Apparatus according to claim 1, characterized in that each of the interpretation members (100a * K, 100b-K in Figs. 1a and 1b) of the set is provided with first members (102, 10H), which are connected to the first and second input terminals and a first output terminal of that one interpreter are coupled to provide a low-pass transfer function of that one interpreter, the first members comprising a convolution filter (102) with m branches. where m is a given plural integer around the information component of the signal that is applied to the first terminal of that one interpretation device. with a predetermined core function to convolve at a sampling sample frequency corresponding to that of the clock signal applied to the second input terminal of that one interpretation device, the predetermined core function being a sampling frequency of the convolution filter of the respective one interpretation device, respectively, and determine the nominal cut-off frequency of the low-pass transfer function of the respective one transfer member in each dimension 8402009 * - - 51 - of the information component "and second members (109. 110). those with the first members having the second input and second output terminals of the respective one interpreter is coupled to provide the difference signal 5 at the second output terminal of the respective one interpreter, these second members being provided with sample subtractive members (110) and third members, which are provided with delay members (106, 108, 109). 1a; 109, Fig. 1h and 1c) to couple the sample subtractive organs through the delay means to first organs, the sample subtractive organs in temporal alignment at the sampling frequency of the convolved samples of the respective one interpreter, each of the successive respective sample levels of convolved samples of the respective subtract one rendition from each of the corresponding successively occurring respective levels of the information component of the signal applied to the first input terminal of the respective one rendition before a convolution with the predetermined core function of the convolution filter of the respective One interpreter takes place, "whereby the output of the sample subtractive members each of successively occurring respective differential sample levels. at the sampling frequency of the convolved samples. n comprises the respective one interpretation device, which respective differential sample levels form the information component of the signal which is present on the second output terminal of the relevant one interpretation device. 8. A device according to claim 7, characterized in that the predetermined core function of at least one of the interpreting members of the set determines a low-pass transfer function form for said transfer member exhibiting a gradual reduction occurring at its nominal cut-off frequency. extends. Device according to claim 7, characterized in that the respective core functions of at least two of the interpretation members of the set are substantially the same. 10. Device as claimed in claim 7, characterized in that the information component has at least two dimensions and the convolution filter of at least one of the interpreting members is an inseparable filter in at least said two dimensions, 84 02 eye * ru - 52 - 11. Device according to claim 7, characterized in that the information component has at least two dimensions and the convolution filter of at least one of the interpretation elements is a separable filter in the two dimensions. Device according to claim 7, characterized in that the first organs (102, 10k) of at least one of the interpretation members of the house are of a certain type, this particular type of first members being provided with the convolution filter (102) and a decimator (10 ^) coupled in series between the output of the convolution filter and the first output terminal of the respective interpretation member of the set, the convolution filter of the particular type of first members at the output thereof provides sample density in each dimension of the information component which corresponds to the sample frequency of the clock signal applied to the second input terminal of the respective one of the interpreters, and the decimator of the particular type of first members in each of the dimensions of the information component only certain, but not all, of the convolved samples, which are output at the output of the convolution filter of the certain type of first means occur, to the first output terminal of the respective one interpreter, thereby reducing the decimated sample density of the convolved sample, in each of the dimensions of the information component on the first output terminal of the respective one interpreter. relative to the determined sample density of the corresponding dimension of the information component at the output of the convolution filter of the relevant one interpretation device, 13. Device as claimed in claim 12, characterized in that the decimator of the determined type of first organs in each of the dimensions of the information component, every other the samples occurring at the output of the convolution filter of the determined type of first organs, to the first output terminal of the respective one interpreter causing the decimated sample density, in each of the dimensions of the information component, to be reduced to half the determined sample density of the corresponding dimension of the information component * .1 * +, Device according to claim 12, characterized in that in said. at least one interpreting organ (206b-K. fig. 1b) the third organs are provided with fourth organs, which are coupled between the output of the convolution filter and the sample subtractive organs to connect to the sample subtractive organs ( 110) supplying the convolved information component directly from the convolution filter. Device according to claim 12, characterized in that the der members (1Q6 # 108, 109. fig. 1a) further comprise fourth members (106, 108} coupled between the decimator and the sample subtractors to the decimated sample density of the convolved samples in each dimension of the information component.10 on the first output terminal of the relevant one interpreter, to be converted back to the determined sample density of the convolved samples in that dimension, which are the fourth organs, which are fourth organs provided with a sample expander (8) for introducing additional samples, corresponding respectively in performance to each convoluted sample output of the convolution filter, which is absent from the decimated sample density, each of the additional samples introduced level with a value of zero, and interpolating means (108) for a stitch substitute level of interpolated value for the level with a value of zero of each of the additional samples introduced. to. Device according to claim, characterized in that the decimator of the determined type of first means in each of the dimensions of the information component. the samples, which occur at the output of the convo-2p lution filter of the particular type of first members, supply the other one to the first output terminal of the relevant one interpretation device, which expander means an extra sample in the space between each pair of consecutive convolved samples of the decimated sample density in each dimension of the information component, and the interpolation means include an n-branch interpolation filter (where n is a given plural integer) with a low-pass transfer function. Device as claimed in claim 11 or 15, characterized in that the information component of the signal on the first input terminal of the respective one interpretation device is supplied to the sample subtracting elements thereof via the delaying means thereof and the delaying elements of the relevant one interpretation body has a time 8402009 »κ it Η. * introduce traction, which is essentially equal to the total time delay. introduced by the convolution filter, the decimator and the fourth organs of just one interpretation device. 18. Device according to claim 12, characterized in that each of the first to the ('N-1) -th interpretation members of the set is provided with first members of the determined type (100Q-K or 100b-K). 19. Device according to claim 18, characterized in that the Nth interpreting members (fig. 1c) of the set also comprise first members of the determined type, 20. Device as claimed in claim 18, characterized in that the Nth interpreting members (Fig. 1c) of the set. are provided with first organs of a different type. wherein the output of the convolution organs is applied directly to the first output terminal of the Nth interpreters. 21. Apparatus according to claim 20, characterized in that the information component of the signal (¾ ^) on <3> a first input of the N-th interpreters of the set to the sample subtractors thereof. is supplied through its retarders (109). wherein the retarders of the Nth interpreters of the set introduce a time delay substantially equal to that introduced by its convolution filter. 22. Apparatus for performing real-time spectral analysis characterized by a cascade connection of low-pass sample filters (102, 10U of: - Figs. 1a, 1b in 100-1, 100-2, etc.), which are always lower sampling frequencies (CL1 (CL2. etc.) are operated, with an input signal (Gq) of the cascade being the signal, the spectral analysis to be performed, and the output signal (G ^.) having a residual low-pass spectrum, organs ( 106_103, Fig. 1a) to zero the samples of a decimation of the response of each low-pass sample filter (102, 10¼) and subject the result to a low-pass filter to obtain a respective interpolation result, means (109) to samples of the input signal of each low-pass filter in the cascade connection by an amount equal to the sum of the 35 delay in response to that filter and the delay at the low-pass filter response response, interleaved with zeros (and means (110) to differentially combine the delayed sampling of the input signal of each 3402009 * * # · -55 - low-pass filter in the c axis gift connection with the interpolation result obtained from its response , providing a respective analysis (11 ^) of the spectrum analyzes of the input signal of the cascade connection. 5 23 · Apparatus for performing a real-time spectral analysis of an electrical signal, which is regularly sampled at a frequency H, characterized by a number, n in number, of analysis steps (100-1, 100-2, etc.) , which are consecutively numbered zero to n, each analysis stage (Fig. 4) providing a first output signal (G ^^) in response to low frequency components of its input signal (Gg.) and a divorced second. output as a response (L ^) to higher frequency components of its input signal (G ^.) ^ where the stage of the analysis stages, numbered with (^ 100-1), the electrical signal for spectral analysis as receives its input signal, and any other of the analysis signal as its input signal receives the first output. of the analysis stage with the subsequent lower cardinals receives number, wherein the second outputs of all said stages and the first output of the analysis stage, numbered n, are provided in the spectral analysis, and each of the number of analysis stages (for example, FIG. * 0 is respectively provided with a first shift register (470) with m stages .. where m is a plural integer, the input signal (Gg.) for the analysis stage being applied to its input and clocked having a clock frequency equal to R / 2 t where k is the cardinal number of the analysis stage, means (471) which is the input signal (Gg.) of the analysis stage and the input signal as delayed in each stage of the first kick-off shift register, weighing a set of coefficients and summing the weighted signals to provide a low-pass filtered linear phase response (Gg + ^) to the input of the analysis stage, w each response forms the first output of the analysis stage, a multiplexer (472). which alternative makes a choice between the first output signal of the analysis stage and a k-1) zero value at the frequency of R / 2, a further shift register 35 (474) with m-stages, the signal selected by the multiplexing input at its input is supplied and at the clock frequency; equal to B / 2 ^ "is clocked means (474), which weigh the weighted 8402009 & -56 signal from that analysis stage and the signal, as delayed in each stage of the further register of m-stages by the set of weighting coefficients and Sum weighted signals to obtain a resampled first output for that analysis stage, and organs (bJ5) which subtractively combine the resampled first output for that analysis stage with the delayed input for that analysis stage to provide the second output. (L „) for that analysis stage. 2k, Apparatus according to claim 23, characterized in that m is the same for every 10 analysis stage and each analysis stage uses equal values of weighting coefficients. 25. Device according to claim 23, characterized in that the delayed input signal for each analysis stage is obtained from the mth stage of the first shift register with m stages and is further delayed (by i + T6). 26. Digital filter with at least one output response and characterized by a number of tapped, clocked delay lines, for example (Ufo, fig, U in each of 100: -1, 100-2, etc.), which are sequentially numbered sequentially and at progressively lower frequencies (R / 1, R / 2, etc.) 20 are clocked as their ordinal numbering increases, means for applying a filtering input signal (Gq) to the input of the first of the number of delay lines, respective means (e.g. U71 in each of 100-1, 100-2, etc.) to weigh samples from the taps of each delay line and combine the weighted samples to obtain the respective filter responses (Gp Gg), and means to measure the respective filter response, derived from the samples taken from each delay line, except those with the highest ordinal numbering, to be input as input signal to the delay line with the subsequent higher ordinal numbering, 30 At least a portion (G ^) of the respective filter response, taken from the samples drawn from the delay line with the highest ordinal numbering, is used in providing the overall response. Digital filter according to claim 26, characterized in that the weighted samples from the branches of each delay line are combined to obtain respective filter responses (Gp Gg, etc.) which have a low-pass character. 8402009 φ V, 5Tr Digital filter according to claim 27, characterized. that one of the respective filter responses (G ^.) obtained by weighing and combining the samples from the delay line with the highest ordinal numbering is used as an output response of the digital filter. Digital filter according to claim 27, characterized by a further branch. clocked delay line (h-73) - which is clocked at the same frequency as a selected line (** 70) of the number of tapped delay lines, means (** 72) to selectively select the output signal of the selected delay line and zeros at the input of supply the further delay line, means (** 7 * 0 for weighing samples from the branches of the further delay line and combining the samples to obtain a low-pass filter response. and means (** 75) for differential combining the low-pass filter response thus obtained with the output signal of the selected delay line in said number of delay lines to provide an output response (Dg.) of the digital filter. 30. A digital filter according to claim 27, characterized in that the weighted samples taken from the branches of a selected delay line (570-1) of the number of branched delay lines (570-0, 57Q-I} with an ordinal number (1) that is one greater than the ordinal number (0) of a preceding delay line: (a) are weighted differently at alternating periods of the clock periods of the preceding delay line and (b) at each of these clock periods are combined for obtaining a low-pass filter response, and wherein the low-pass filter response thus obtained is differentially combined (at 575-0) with the output signal of the previous delay line (575-0) for obtaining an output response (Lq) of the digital filter, 30 Digital filter according to claim 26, characterized by a further branched clocked delay line (1 * 73), which is clocked at the same frequency as a selected line (** 70) when the number of tapped delay lines j means (** 72) to selectively apply the output of the selected delay line and zeros to the input of the further delay line, means (** 7 **) to weigh samples from the branches of the further delay line and combine the samples to obtain a response and organs (** 75) to obtain the response obtained on this 840200§ V * Λ Jk rj - 58 r combine the delayed input signal (G ^) of the preceding delay line to provide an output response (L1) of the digital filter. The digital filter of claim 31 * wherein the delayed input signal from the preceding delay line is obtained by incorporating the delay device (Vf6) into one of its taps, 33. A digital filter according to claim 3, characterized in that the weighted samples from the branches of a selected line of the number of branched delay lines, with an ordinal number one higher than a previous delay line, differ at alternating periods of the clock periods of the previous delay line are weighted and combined at each of these clock periods, combining the result with the delayed input signal of the previous delay line to provide an output response of the digital filter. 3 ^ -. Digital filter according to claim 33, wherein the delayed input signal of the preceding delay line is obtained by incorporating the delay device into one of its branches. 35. Signal processing device for synthetically building a single temporal signal (G) from an ordinally arranged set of ET separated temporal signals (Lq-G), where ET is a plural integer characterized in that for synthetically building the single temporal signal on a delayed real-time basis (1) the single temporal signal includes a given flow of information component samples, which determines the frequency spectrum of information with a given number of dimensions and a given sample density in each of these dimensions, (2) the first ( Lq) of the ordinally arranged set of N30 separated signals comprises a stream of information component samples (which determines an upper portion of the frequency spectrum of the information with a sample density substantially the same as said determined sample density in each of the dimensions. (3) each of the second through (ïF-1) the (L ^ L ^) of the ordinal rank 35 set of ET separated signals comprises a stream of information component samples, which determines an individual portion of the frequency spectrum of the information - of each dimension thereof, which below that of 8402009 * * * * »- 59 - the corresponding dimension of the spectrum of that section is determined by the immediately preceding separate signal from the signals of the system and lies above that of the corresponding dimension of the spectrum of that part determined by the immediately following separate signal from the signals of the set, (k) the flow of information component samples, corresponding to each of the second to (ff-1) the (Lj ... L ,.) of the ordinally arranged set of ff separated signals and a sample density for each of its own information dimension which is less than the sample density 10 of the corresponding information dimension of the information sample stream of information corresponding to its immediately preceding separated signal from the set, and (5) the respective streams of information component samples occur in a predetermined time-shift relationship with each other and the device 15 (in Fig. 3) includes a group of ( ff-1) sample signal combination means (363 - 353) each individually (e.g. 363, 362) individually cooperating with a respective signal (e.g. Lq) from the first to said (ff-1) ordinal signals (Lq-L ^) of the set of separated signals to combine that ordinal signal (for example Lq) of the set of separated signals, which cooperates with said combination organs, with the cumulative total (for instance 1) of all those separated signals (for example, Lg ·· which follow that ordinal one separate signal in the set and wherein each of the combination members (for example 362, 361), belonging to the first (Lq) to 25 (ff-2) -th (L · ^) ordinal signal the separated signal set includes an adder (e.g., 363), first means, (e.g., 3 ^ 0) to apply the associated ordinal separated signal as a first input signal to the associated adder, and second means (e.g., 362) the output signal of the adder (eg 361) from the combiners (eg 361) associated with the separated signal (eg C-1) phon-mid immediately following the associated ordinal separated signal, as a second input signal to the associated adder with the same sampling precision as that of the associated ordinal separated signal Poe, the combination means (353, 352] f corresponding to the (ff-1) -th separated signal (1j_) being provided with the set from an adder (353), the first means (350) for the (ff-1) -th separate signal as a first input signal to the 8402009 é. and adding the adder, and third means (352) for feeding the U-th separated signal (G) as a second input signal to the "associated adder (353) with the same sample density as the (11-1) -th separated signal and the respective first members (3 ^ + 0, 3 ^ 1, etc.), the respective first members. (3 ^ 0, 3 ^ 1, etc.) 4 introduce the respective second members (362 362 etc.), and the third members of the (N-1) combination members of the group of respective predetermined amounts of time delay when feeding of the time-shifted individual signals of the Set, all such that for each of the respective (N-1) combination members corresponding information samples of the respective flows of information component samples at the first input and at the second input of the adder therefrom occur substantially in time coincidence with each other, thereby obtaining the synthetically constructed single temporal signal at the output of the adder of the combiners associated with the first separate signal of the set. 36. Device according to claim 35, characterized in that the second members (eg 362, fig. 3) of the respective combination organs, which belong individually to each of the first to (N-2) -th ordinal signals (for example, L ^ ^) of the set of separated signals comprise a sample expander (692, 693, 69 * 0, which, in response to the flow of information component samples (GT '), has a smaller sample density at the output of the adder are supplied to introduce additional samples into the supplied current in order to increase the sample density at the second input of the adder (695) of the one combiner to the sample density of the ordinal discrete signal (L ^ ^). in that one combiner, each of the additional samples introduced having a level of zero and interpolating means (693, 69 ^) serving to provide a sample level with interpolated commodities to substitute for the zero level of each of the additional samples introduced. Device according to claim 36, characterized in that the Nth separate signal (G _r> _) of the set has a sample density which is smaller than the (N1) -th individual signal (L, -) of the set. set, and the third members (352) are provided with sample expansion 8402009 * * * - 6ι and interpolation members (692, 693, 694), corresponding to those of the second members, the N-th separate signal at the second input of the adder of the third members. 3Ö. Device according to claim 3T, characterized in that the M-th separate signal (GjftJ of the set has substantially the same pitch sample density as the (1-1) -th separate signal (L ^) of the set and the third members supplying the separate signal directly to the second input of the adder of the third members. The device according to claim 36, characterized in that the stream of information component samples corresponding to each of at least the second (L 1) to ( M1) -th (L ^) of the ordinally arranged set of SF discrete signals for each of its own information dimensions has a sample density "which is half the sample density of the corresponding information dimension of the information component sample stream, corresponding with the immediately preceding separate signal from the frame and in each of the combination members the expansion device (692) of the second members an e additional sample between each pair of consecutive samples with said smaller sample density in each dimension of the flow of information component samples at the output of the adder (CL · '), which is introduced, introduces and the expander an interpolation filter (693) with n branches include # where n is a plurally determined number with a low-pass transfer function. The device according to claim 35, characterized in that the third members and each of the respective second members of the group of JT-1 sample signal combination members have their own predetermined amount of time delay in feeding the associated stream of information component samples as a second input signal to the associated adder. and each of the first members of the group of jJ-t sample signal combination means includes delay members (340, 341, etc.) which introduce a certain amount of time delay when feeding the ordinal discrete signal thereof as a first input signal the corresponding adder, which depends on so many (1), the respective time shift between the respective ordinal separate signal and each of those separated signals of the set, which are separated on the corresponding 8402009 - 62 * V * * \ ordinal signal tracking, and (2) the total amount of time delay consumed by the third members and all second members of said combination members associated with the separated signals of the set following the respective ordinally separated signal. 5, wherein the determined amount of delay is such that corresponding information samples of the respective flows of information component samples Tests at the first input and at the second input of the adder in question occur substantially in time coincidence with each other. 8402009