WO1993015558A2 - Information compaction system - Google Patents

Abstract

This invention relates to devices to compact data, and to transmit, store, compute with and cross-reference this compacted data. It is particularly concerned with a system of computing devices which accepts any quantity of information from any number of sources (1), combines specific blocks of this data and turns the combination of blocks into a single expression which can be coded by a device for transmission or storage in a more compact form than the data originally fed into this coding device but which because an algorithm has been used to perform this coding can be decoded by a decoding device (7) to regenerate on a one-to-one basis the original data provided to the input device (4). Another device which compacts the information further by approximate methods is particularly applicable to the transmission and storage of graphical and acoustic information. This invention comprising coding, decoding and information storage devices has particular application to the transmission, storage, recall, computation and referencing of information in which the compactness and security of the information is a requirement. Examples of applications in which this is needed or is desirable include the analogue and digital telephone systems including voice and data transmissions, television and radio transmission, disc and tape recordings, computers, computer networks and electrical, optical and chemical memories for the storage of information in digital or analogue form.

Description

Information Compaction System

This invention relates to apparatus for and methods of compacting data, and also to apparatus for and method of transmitting, storing, computing with and cross-referencing this compacted data. It is particularly concerned with a system of computing devices which accepts any quantity of data from any number of sources, combines specific blocks of this data and turns the combination of blocks into a single expression which can be coded by a device for transmission or storage in a more compact form than the data originally fed into this coding device but which by reason that an algorithm has been used to perform this coding can be decoded by a decoding device to regenerate on a one-to-one basis the original data provided to the input device. Another device which compacts the information further by approximate methods is particularly applicable to the. transmission and storage of graphical and acoustic information. This invention comprising inter alia coding, decoding and information storage devices has particular application to the transmission, storage, recall, computation and referencing of information in which the compactness and security of the information is a requirement. Examples of applications in which this is needed or is desirable include analogue and digital telephone systems including voice and data transmissions, television and radio transmission, disc and tape recordings, computers, computer networks and electrical, optical and chemical memories for the storage of information in digital or analogue form.

The term "compact" used in the description of this invention means that by comparison with the time period occupied by the transmission of one bit of digital information and within the frequency bandwidth required for the transmission through the particular medium more than one bit of compacted information can be passed by means of this invention.

As applied to the storage of information it means that by comparison with a conventional memory s t ore which wil l ac c ept one e lement or bit o f information the compact memory store of the present invention will hold more than one bit of information in one location and it will both accept and release all that information in one procedural step such as the period between two successive ticks of a computer ' s clock. State of the Art Transmis sions of data by physical means whether down a conduit such as a wire or optical fibre or through a medium such as air , water , or space outside the atmosphere has taken one of two forms . Either the data is represented by a code of pulses and absence of pulses in a sequential stream, the pulses being some form of energy level above a ground value like increased voltage , a flash of light or other electro-magnetic radiation or else it is represented by puls e s whos e energy level repre sents a quantity according to a code . The former is commonly understood to be digital transmission in which it is the presence or absence of the increased energy level and not its value which is the indicator of information . The l att er i s c ommonly unders tood to be analogue transmission and the information content is in the value of the energy level measured above or below some stated level .

In the case of digital transmission, because it is only the presence or absence of a pulse which is required to indicate information and not its strength. this is a reliable system especially over lines which are of poor quality or routes which are noisy. It is the reason why Morse code was used on ship's radio communication world-wide for many decades. However, sequential digital transmission of bits of elemental information is essentially slow and makes inefficient use of the capacity of the medium and the available information bandwidth. The term "information bandwidth" is used hereinafter to mean all the distinguishable properties of the medium within the compass of the channel allocated to that user and the technical facilities available.

This relative inefficiency can be seen when considering one bit of information to be sent down a telephone line with the constraint of a frequency bandwidth between 300Hz and 3400Hz.

Ideally, if the information to be sent is composed of alternate ones and zeros then 1700 ones and 1700 zeros per second could be sent using just the highest frequency. If consecutive ones or zeros occurred then additional means would be needed to count those periods during which no change occurred but at which the voltage level remained well above the ground state of zero. However, this makes use of only about 2.0805 x 10 of the theoretical information carrying capacity of this channel if means are available for distinguishing every integral frequency in the band and every cycle in those frequencies separately. If other physical properties of an oscillating electrical voltage and divisions of the frequency into parts of cycles were also distinguishable then this theoretical capacity would be increased accordingly.

Further restrictions are placed upon the information carrying capacity of the telephone line by reason of the form of the signal used to represent one bit. The current practice is to attempt to approximate a square voltage/time curve with an abrupt rise from zero to a given voltage and an abrupt fall back to zero after a brief interval of approximately 1/2400th of a second.

To get a square wave would require the incorporation of an infinite bandwidth of frequencies so the approximation which results from its confinement between 300-3400Hz is a waveform with a significant slope to the leading and trailing edges of the pulse. This in itself reduces the maximum number of pulses which can be sent to less than 1700 per second if the intervals between pulses are to be equal to the time occupied by the pulse from the moment the value of the voltage rises above zero to the moment when it returns to zero.

As the pulses are not generally counted by timing, as it was suggested above was one way of determining the number of constituent pulses in a consecutive block, it is necessary for their identification that a detectable voltage drop shall occur between each one. This prevents the pulses from being as closely spaced as they might be if they were truly square. A the cost of requiring a better quality of line with a lower level of tolerable noise, _. analogue techniques have been introduced for higher speed facsimile and modem transmissions, which are still bound within the constraints of a telephone line for their bandwidth, but which by using phase shift of the carrier frequency and two levels of amplitude convey four times the amount of information that the digital technique achieves over ordinary lines. Where special conditions exist so that noise levels are very low then sixteen times the quantity of information per unit time can be sent than by digital bits. This is a compaction ratio of times four, obtained by superimposing on the carrier frequency of 2400Hz one of eight phase shifts at one of two amplitudes so that any one of sixteen numbers can be represented by a particular phase shift at a particular amplitude. As sixteen requires four binary bits for its expression instead of the single expression of phase and amplitude of this analogue signal in the same unit time as one bit, so the compaction ratio is times four.

In systems other than telephone lines an analogue system is used which depends upon allocating each one of a number of frequencies within the allocated bandwidth to the representation of a specific symbol or value. Then detection of each separate frequency reveals the item of data and the rate at which it can be transmitted depends upon the time required to identify each frequency. If the devices used to send or receive the signal require a number of cycles to pass through them in order to operate then the data rate will be quite slow even if the actual time taken depends upon the number of cycles rather than being fixed at the numbers of cycles for the lower frequency. However, if the standard by which relative compaction is to be judged is the time taken for one bit of digital information to be transmitted and if this is being done already at the highest frequency of the available band then there is no room for any change of frequency to increase the rate unless the detecting equipment can measure the frequency from the information contained in less than one cycle. Thus compaction will be minimal and will depend upon the ability to discriminate frequencies close to the maximum and over the shortest part of a cycle that can be managed. In practice, frequency modulation on telephone lines has achieved only the same rate of data transfer as digital techniques using two frequencies pulsed in l/1200th of a second.

Another analogue system employed to compact graphical images employs an approximating method in which expressions based upon fractal mathematics are fitted to the data which appear on a screen and it is these expressions which are transmitted or stored to reconstitute an approximation of the scene upon receipt or recall respectively . While this is sufficiently accurate for the transmission of graphical images where point by point accuracy is not es sential it is not tolerable as a method of transmitting data where every bit has to be transmitted or stored if the information is not to be corrupted . It is also a method which demands a large amount o f memory s torage and calculation for its execution so it is slow in its implementation .

In the cas e o f the transmi s s ion of information by pulses of electromagnetic radiation from wave lengths as short as light or as long as low frequency radio waves the problem is accentuated because the available spectrum of wavelengths is so much greater while the signalling techniques remain essential by digital . The result is that while the most modern optical fibre transmits solitons containing a spectrum of wavelengths between lμm to 3μm, it has a theoretical information carrying capacity of about 10 1 ft^o bits per second on the assumption that a difference of 1Hz is distinguishable . However , only about 10 ² is achieved at the moment .

Where amplitude and phase modulation are used, even in conjunction with frequency modulation , the amount of information sent through the medium is only a few times greater than it would be if sent by digital techniques, but the improvement that this increase of information gives to the quality of the signal is significant, especially in radio broadcasting. Turning now to the storage of information, in conventional practice digital information is stored in physical locations by changing a property of that location between one value and another. In the case of analogue information that property may take one of more than two values and may be in the form of electric charge, reflectivity of light, magnetic intensity, or pressure in the medium. It has not been found to be practicable to use more than about 32 levels of distinct value for any one property as far as is known. The primary limitation upon the compaction of information, whether for transmission or storage, and therefore the inability to exploit the information carrying capacity of the medium to the full, arises because none of the known methods used allow two or more items of information to be added together when derived from separate, unrelated observations and then to be uniquely analysed into the component parts on receipt or on retrieval from the store.

Much of digital computing consists in comparing the physical properties stored at specific locations according to rules of logic and then storing the result, which itself generates a value for another physical property to be stored in another location. Analogue computing does the same except that the multi¬ valued properties at the specific locations will give rise to multi-valued results instead of one of two values.

The speed at which computations by such comparison can take place is limited by the sequential nature of the process of comparison in each processor and the division of every task into elemental steps, one bit at a time in the case of digital processing and only a few bits at a time in the case of analogue processing. As in transmission and storage the potential capacity of a system to process information comprising large assemblies of elemental bits at one operational step is not realised because of the inability of the methods or the equipment to add a significant amount of information together in such a way that its components are distinguishable in the composite block at any one moment of the processing as required. If the information was to be merged into a single unit by adding or some other compositing arithmetical operation then the components would be irrecoverable, like adding two volumes of a similar fluid to each other and being unable to separate them again afterwards.

Two functions of a computer to which the present invention is relevant and which devices to be described hereinafter are designed to fulfil are the ability to cross-reference data that has been stored in a memory and the related ability to recognise patterns and relationships between entities.

All observable phenomena can be measured in a variety of quantifiable dimensions. For phenomena to be comparable they must share one value in at least one dimension, yet to be distinguishable each phenomenon must differ from all others in a value in at least one dimension. Within these constraints the relationship between phenomena is a function of their adjacency in value in the dimensions that are relevant to the purpose of the comparison and in descending order of importance to that purpose.

For a digital computer even quite simple referencing, or pattern recognition, which is what it __g_

is, and which for animals capable of choiceful behaviour is easy, requires such an amount of computation that few such tasks can be carried out in real time. This is because the problem has to be described in elemental terms and these terms computed in sequence. Even for a number of processors working in parallel the kind of actions executed by a lowly animal like the sand wasp (A mophila sabulosa) represent so vast a computational exercise that it would be too protracted to be practicable. Analogue computers able to manipulate more bits of elemental data at each transaction might ease this burden if the speed at which they could operate each one could be made to equal that of the digital computer and with the same reliability but this is not so at present.

The problem is exacerbated by the analytical procedure necessary to digital computing in which each bit, or even each small group of bits, is processed quite independently of its relationship to the whole pattern of which it represents a part. To some extent these difficulties are ameliorated by means of the kind of electrical circuit known as a "neural network" after its supposed analogous working to a biotic neural system. In this arrangement sensors, which may also be memory cells containing the record of some physical quantity, are connected to a second layer of cells. Every cell of the first, sensory layer is connected to every cell of the second layer. A third layer beyond the second is likewise connected, every cell of the third layer being connected to every cell of the second layer. Some connections may be made between cells of the first and third layers. The connections between the cells of the layers are changed according to the Hebb rule which dictates that the threshold of acceptance is lowered between all the connections linking the activated sensor cells of the first layer and the activated output cells of the third layer when the output agrees with the model being sensed and is raised when the output does not agree. The amount by which the threshold is altered on each occasion of the model being presented during the training period is in accordance with some rules of logic determining the process. The network is then trained by offering a known pattern to the sensors repeatedly until the output corresponds sufficiently accurately to the model. However, this is a protracted procedure and far exceeds the training required by most biotic neural systems which have evolved with a capacity for learning. Further objections to the system are that while in nature such intense connectivity is possible between very large numbers of cells because each is a self-constructing unit, in the sequential manufacturing processes which form artificial devices comparatively few units can be connected so the inherent capacity of the device remains small. Also because artificial systems are used in such a way that they work in a discontinuous context, in contrast to the natural organism whose context is almost never discontinuous, the data on which they work is relatively sparse. The cells are constrained to communicate by means of the inherently slow digital system and the pattern of connection thresholds which emerges from the training process is non-rational, depending upon the detailed physical state of each cell at the start of the procedure.

In conclusion there does not exist a system of cross-referencing or pattern recognition by artificial means which does not require either the manipulation of very large numbers of elemental bits in s equence or i s restricted to a smal l capacity in principle because it cannot be constructed in the manner of its biotic exemplar .

Fundamentally these objections arise from the inability at present to add together signals either in the transmission or the storage in such a manner that on receipt or recall respectively the component signals can be distinguished . A solution to this problem is what this invention achieves . Objects of the Invention

It is one object of this invention to provide a data compaction system which operates at very high speed with compaction ratios of a very high order approaching the theoretical maximum of the capacity of the medium achievable by any current practical technique .

It is a further object to provide a cross- ref erencing system which is inherently densely connected throughout the memory store in terms of the adjacency of the items and patterns held in that memory store .

It is another obj ect to provide a memory store which is able to store complex patterns at each location so that on recall comparison of these stored contents according to desired official rules can be carried out treating each as a single entity .

I t i s yet another ob j ec t t o provide a computing system which is able to compare the items of c ompacted inf ormation delivered to it as s ingle entities according to any desired logical rule and to deliver the result of such comparison in a similar form to devices f or transmis sion , reception , storage or further computation .

It is a further object to provide a secure , confidential method of transmitting and storing data. It is another ob j ect to provide an approximate method of transmitting or storing graphical images or acoustic information.

These objects of the invention may not all be achieved by all embodiments of the invention.

Examples of applications in which these obj ectives are needed or are desirable , singly or collectively , inc lude the analogue and digital operations of telephone systems in transmitting and receiving audio, visual or data signals ; analogue and digital disc , tape , film and electronic recordings ; computers , computer networks and electronic , optical and chemical memories ; and control and operation of automated or artificial intelligence systems . Advantages of the Invention

The f ollowing advantages may not all be achieved by all embodiments of the invention.

1. Devices in accordance with the invention may be used to compact information to approach the maximum theoretical capacity available in the medium that can be achieved by current technical methods . Thus the capacity of existing communication links can be greatly expanded.

2. The system of the invention allows for the comparison of complex items and patterns containing many elements of information as a single entity in one logical procedural step. This enables truly parallel processing to be achieved on a scale that can exceed this ability as it is displayed by biotic organisms , in a comparable time or faster.

3 . Devices incorporating the system of the invention require a small fraction of the machinery currently necessary to achieve the same computational speed . This will reduce the cost of computer and computing . By reducing the size of the computer many more applications become feasible .

4. Devices incorporating the system of the invention wil l be c apabl e of c ompl ex pattern recognition and association of memorised items out to distant values of adjacency in such short periods of time that effective responses to real time situations can be executed.

5. The enhancement of computing capacity due to the ability to manipulate large blocks of information as single entities will allow the development of truly intel l igent computing devic e s with wide spread application in many fields but in particular to the intelligent exploration of and work in hostile environments such as under ground, under water and in space . It also permits the development of prostheses for disabled brains and innervated organs .

6. Because devices incorporating the system of the invention can convey more information than existing ones , the quality of the communication can be greatly improved or more lines of the existing quality can be catered f or at the same time . This allows the introduction of , for example , high quality video telephoning and video conferencing in real time using only one line , colour facsimile transmission as the standard technique and telephone cable television transmis sion as common s ervices throughout the community over existing telephone wires or optical fibres .

7. Due to the compact nature of the devices incorporating the system of the invention and the more efficient use made of the information carrying capacity of the medium less power will be needed to operate them as compared to existing devices of the same capability.

8. The system of the invention can be applied to existing data compaction and compression systems to reduce the length of the transmission or the capacity of the storage required by the compaction ratio of this system as applied to the already compacted or compressed data. 9. Because so much information is packed within each discrete transmission or memory location and the order and form in which it is sent or located is according to an arbitrary choice made by the legitimate users, so it can be made difficult for any unauthorised user without the key to this choice to decipher the transmitted or stored data.

10. Because the information can be greatly compacted accurately and in a form which is suitable for a subsequent approximating method of further compaction any general body of information such as a graphical image or body of acoustical information can be conveyed or stored in a very compact state.

The design of devices in accordance with this invention is based upon some or all of the following principles although any particular application may embody only one or more and not all of them. The claims appended hereto define the scope of the invention.

1. All the values of the properties measured in _a single observation can be expressed by the occupation of a single cell in a structure of cells representing all possible measured values of the properties by choosing a common point of origin for all the dimensions along which each property is measured. Each measurement diverges from the common origin.

2. All observations that are distinguishable differ by at least one measurement in one dimension from all other observations. Thus for instance observations which are equal in their physical values will be distinguishable only if their values in the dimension of time are different. That is they are observed at different moments.

3. The occupation of a cell implies that a measurement has been made in each of the constituent dimensions of the structure of which that cell is a member .

4. For a cell to be occupied the observation which it represents must take a value in each of the constituent dimensions of the structure of which that cell is a member.

5. Each cell of the structure can be in a state of being occupied or unoccupied and in no other state.

6. Each cell in this structure of cells can be represented by "y" as a continuous function of "x", including its directly connected modulating constants, each cell being allocated a different valued function. For example the general form of the functions allocated to the cells of a structure might be y=ax+b, or y=a sin(bx+c), in which the constants a,b,c take different but fixed values for each of the cells of the structure. These values are chosen so that in adding, or whatever is the arithmetic procedure of association that may be chosen, each of the separate continuous functions of those cells which are occupied, the presence or absence of these cells in the final composite result of the arithmetic procedure can be uniquely determined from consideration of the whole or a part of the composite result as a pattern of values.

In order that the invention may be fully understood, a number of embodiments in accordance with the invention will now be described by way of example and with reference to the accompanying drawings, in which :

Figure 1 is a diagram illustrating the principle behind the present invention; Figure 2 is a diagrammatic illustration of waveforms associated with the system of Figure 1 ;

Figure 3 is a diagrammatic illustration of a compaction system according to the invention using an optical fibre;

Figures 4 (A) and 4 (B ) illustrate the storage o f c ompacted inf ormation in accordance with the invention; and.

Figures 5 (A) and 5 ( B ) illustrate a cross- referencing system in accordance with the invention.

In the f irst example three sensors make measurements of three properties in the course of five observations A, B, C , D & E . Each measurement can take one of three values in the dimensions Dl , D2 and D3 . These are illustrated in Figure 1.

Table 1 shows the values of the observations expressed in binary notation: Observation

Dl 10

B 11

C 01

D 01

E 10

in this visualisation the three dimensions

Dl, D2 & D3 in which the measurements of the five observations A,B,C,D & E are made, are counted from the common origin which is at the bottom, back, right hand corner of the cube of cells, increasing in value to the left, upwards and forwards.

It can be seen from Figure 1 that each observation can be described by the position in the cube of the appropriate occupied cell.

If each cell in the cube depicted is allocated a distinct set of values for each of the constants in the general expression y=a sin (bx+c), then in this example and with reference to Table 2 each curve in a graph depicting "y" against "x" for each observation would be unique. An oscillating electrical signal in which the voltage in the circuit is represented by "y" and "x" represents the phase angle in radians between 0 and 2π of the sine curve y=sinx forms the basis of Figure 2. The constants a,b and c in Table 2 below correspond to amplitude, frequency and phase respectively, expressed in arbitrary units.

Table 2: Allocation of Physical Properties to Cell Number

22

23

24

25

26

27

Referring to Figure 1 and Table 1 it can be seen that the cell numbers corresponding to each of the observations A,B,C,D and E are 5,9,10,25 and 23 respectively. Converting these into the expressions for sine curves with the constants replaced by the values according to Table 2 gives the entries for Table 3 below. Table 3: Conversion of Observations to ϋniσuelv Valued

Expression y(A)=2sin(2x-^*τ) y(B)=sin(3x-3tr/2) y(C)=2sin(x-tr/2) y(D)=sin(x-7tr/4)

y(E)=2sin(2x-5tr/2) The respective curves are depicted in Figure 2.

These modulated frequencies can be combined into a composite curve such that y(n)=y(A)+y(B)+y(C)+y(D)+y(E) which can be transmitted as a voltage varying with time through a medium with a limited frequency bandwidth between y=sinx and y=sin3x. At the same time it should be noted that now, as distinct from conventional techniques, because the amplitude and phase shifts can be attached to distinguishable curves they can be added together with these curves to increase the amount of information that can be carried by the composite curve and hence increase the compaction ratio. Furthermore the only limit to the compaction that can be achieved is set by the ability of the current technology to discriminate between the difference in the pattern of voltages detected by the decoding device that the presence or absence of a constituent curve causes in the composite curve.

When this composite curve is received, or recovered from storage, it can be decomposed into its component curves by any one of several methods.

Two are considered by way of example. The first is to pass the signal to a device comprising a bank of tuned frequency filters each of which will pass preferentially a single frequency, a phase discriminator which will register its phase with respect to the start or finish of the signal or some marker within or accompanying it, and a voltmeter to measure the peak voltage with respect to the minimum voltage for that frequency or the ground voltage.

In this case each frequency which is present in the transmission or store would be picked out by the respective filter and permitted to pass its energy on for analysis of phase and voltage in that channel.

The values determined for phase and voltage, together with the particular frequency with which they were associated, would be passed to the decoding device holding the reference values as shown in Table 2 from which the occupied cell number would be obtained and from that by reference to a list or algorithm contained in this decoding device the quantities in each dimension for each of the five observations would be determined.

However, this method suffers from various practical disadvantages which would reduce its actual• performance with regard to reliability and obtaining a high compaction ratio. These are the difficulty of obtaining a precise and stable performance from the f liters , phase discriminator and voltmeter, and the requirement of these instruments for the passage of s everal cycle s of the compo site curve bef ore the constituent pure sine curves can be detected reliably. The second, preferred method relies upon high speed, short duration sampling of the composite curve and analysis of the pattern of the values of these s ample s as they are acquired with the ob j ect of determining the components of the composite curve with sufficient certainty for the purpose of the work. At a point in the analysis , generally expected to be before the ττ/2 phase of the highest frequency in the frequency bandwidth of the medium has been reached , this certainty will have been attained, so allowing that analysis to be stopped and attention to be turned to the next signal for decoding.

This proces s has been set out in Table 4 below which shows that the abs ence of any of the constituents or the presence of an additional one of s imilar value would c ompri s e a distinguishable proportion of the compo site curve . Furthermore , although the addition or subtraction of constituents at any one or several points may cancel each other this will not happen over the range from x=0 to x=0 .5236 where 1/4 of a cycle of the highest frequency (y=sin3x) occurs . te curve D + E = Comp.

0.707 -2.0 -2.293 0.753 -1.982 -2.512

0.795 -1.929 -2.724

0.834 -1.841 -2.924

0.869 -1.721 -3.105

0.900 -1.570 -3.260

0.927 -1.390 -3.3S3

0.950 -1.186 -3.465

0.969 -0.961 -3.501 The decoding proce s s which the decoding device would be required to execute could be of several kinds . For exampl e in thi s cas e where the f ive observations may take values in a block of 27 possible values as shown in Figure 1 there are 80730 possible composite curves which could be drawn . This is a small enough number for a look-up table to be used to find a solution, particularly if a compact memory store was available which allowed the comparison of the shape of the whole composite curve over the range shown in Table

4 in one procedural step .

For larger numbers of observations or blocks giving rise to a more extensive list from which to select the solution such mathematical techniques as the application of Fourier analysis by means of an additional computing device incorporated into the decoding device would be more suitable to separate out the component f requenc ies and their modulating properties . Having found the appropriate cell number for the f requency , phase shif t and amplitude of each component of the composite curve from Table 2 then by reference to a list or the working out of an algorithm contained in the decoding device the quantities of each dimension f or each of the f ive observations can be determined .

A s ec ond embodiment o f thi s invent ion comprises coding and decoding devices which compact data being provided to the coding device as a stream of ones and zeros from such sources as a conventional computer or facsimile apparatus for transmission to adecoding devic e attached to other c onventional computing or facsimile apparatus . In this case the coding device is fitted with a counter and a sufficient memory to count off a predetermined number of digits to c orrespond to the number o f distinct continuous functions and their modulating constants each of a value to give rise to a unique curve allocated to the po s ition which wil l be taken by a digit in the predetermined block of digit s , and to store the results . Those cells occupied by a one will then contribute their unique curves to the composite curve which will be transmitted or stored while the counter and memory are made ready to accept the next block of data.

For instance , if the data stream was that describing the observations A, B , C , D and E of the previous example then the thirty bits of information containing the measurements of each observation in each of the three dimensions could be translated into a single composite curve using the Tables 1 , 2 and 4 (with a supplement to Table 2 to allocate values for the last three positions in the block. Numbers 28 to 30 inclusive ) . This arrangement would be known to the receiving station excepting the actual values which any five observations might take within the compass of the system.

It is apparent that more observations could be arranged to be contained within the block if the constraint is observed that no two observations can occupy the same cell and if the procedure to allocate a s ingl e cell to each observation is f ollowed as described previously. However, Fourier analysis would be required to analyse the signal arriving at the decoding machine as the number of possible combinations o f the bit s o f data in thi s f orm amount s to approximately 6. 872x10 ^•*- .

I n both embodiment s o f thi s invention described above the period required to send one bit of digital information is that of one cycle expressed as y=sinx (frequency 1 in Table 2) . Thus the compaction achieved in both examples is by times thirty as it would require thirty binary bits to send the data describing the five observations A,B,C,D and E by conventional digital transmission while it requires only a single digital period to incorporate it within the constraints of the available frequency bandwidth by means of this invention.

Furthermore, if the sensitivity of the detecting and analysing equipment allows, then as Table

4 shows this compaction ratio can be increased by a factor of twelve, making an overall compaction ratio of 360 because only one twelfth of a complete cycle of y=sinx (itself equal to the period required in which to send a distinguishable binary digit) is required in which to identify each component function of "x" allocated to each of the thirty cells.

A third embodiment of this invention describes the compaction of information to be sent through the medium of an optical fibre and is illustrated by Figure 3.

In this case the observations made as in the first embodiment of the invention as described above are arranged so that each cell of the cube depicted in Figure 1 is allocated a separate and limited spectrum of wavelengths of electromagnetic radiation, which may or may not be visible, and which will be transmitted as a component of the transmitted beam if the cell is occupied and will be omitted from this beam if the cell is unoccupied.

Referring to Figure 3, the composite beam of radiation is made up by shining beams 2 of electromagnetic radiation, each consisting of one of the component wavelengths, from sources 1 through an electronically controlled shutter 3 which will admit or obstruct the beam 2 according to whether the cell which it represents in the cube depicted in Figure 1 is occupied or not.

Those beams 2 admitted by their shutters 3 pass to another electronically controlled shutter 4 which admits or obstructs the composite beam 5 made up from all the admitted beams 2 according to when it is desired to pass a pulse of the composite radiation of beam 5 down an optical fibre 6 to a decoding device 7. Where it is considered appropriate other properties may be imposed upon any one of the beams 2 such as a change in intensity or difference of polarisation of the radiation to differentiate it further from the remainder of the beams 2. At the decoding device 7 the incident beam 5 is passed through a dispersing device 8 such as a prism or a length of optical fibre which is not corrected for dispersion or is of uniform refractive index in its core wherein the component wavelengths become separated due to their different velocities through the transparent material of the fibre.

The dispersing device 8 casts the separated components 2 of the beam 5 upon detectors 9 which then register the fact of their illumination to a decoder 10.

In this embodiment of the invention the analytical function of a look-up table, Fourier analysis or other mathematical process is carried out automatically by the dispersing device 8 so that the component beams 2 falling upon the detectors 9 are a direct indication of which cells in the cube depicted in Figure 1 are occupied or not, leaving the decoder 10 with the task of relating the occupied cells to the dimensional values as depicted in the cube shown in this Figure 1. A fourth embodiment of this invention relates to the storage of the compacted information resulting from the application of the principles set out above to data which may be obtained directly from observations or which is provided f rom s ome other s ource in conventional digital or analogue f orm . This is illustrated in Figure 4.

In Figure 4 a beam 20 of electromagnetic radiation compo sed of a number o f component and separable limited and distinct spectra of wavelengths is incident upon a small domain 21 of a screen 22 . Over this screen 22 is evenly spread a thin deposit comprising an intimate mixture of various molecules each type of which will respond to a component of the beam 2 0 by such a rearrangement of it s internal chemical bonds that the molecule changes its shape , disposition of electric charge or chemical bonds so that its response to subsequent irradiation by a specific wavelength of electromagnetic radiation is different to what it was before being irradiated by beam 20 , particularly with regard to its reflectivity, transmission or absorption of that wavelength .

I n thi s manner the re spons e o f the illuminated domain 21 to illumination by a beam 23 , which contains all the components of wavelengths of electromagnetic radiation to which the sensitised molecules of the various types composing the mixture covering the screen 22 respond, has been changed. Only those component s of the incident beam 23 will be reflected, transmitted or absorbed, to be detected by detectors 24 , which are related to those components of the beam 1 which altered the characteristic s of spec i f ic molecul e s coating the screen 22 and in particular the domain 21 , which were sensitised. Both beams 20 and 23 can be swept across the screen 22 illuminating domains such as domain 21 sequentially. When beam 20 sweeps across the screen 22 the characteristics of the constituent molecules of the coating of the screen are set according to which components of specific wavelengths comprise beam 20 .

When beam 2 3 sweep s ac ro s s the s creen 22 then components of beam 20 can be inferred from the response of the detectors 24. Thus an optical memory is created which can retain information which may be relayed to it in the compacted form generated by a device such as that described above in the third embodiment of this invention.

Where the screen 22 is of such a shape that when it is illuminated by beam 4 the reflected or transmitted components of this beam from each domain such as the domain 21 , impinge upon a different group of detectors 24 then it can be seen that if beam 23 encompasses a large part of or the whole of the screen 22 , as in Figure 4 (B ) , then all of the groups of the detectors 24 will be illuminated at once . Similarly if beam 20 likewise encompasses a large part of or the whole of the screen 22 then all the domains such as the domain 21 over the area of the screen 22 , will be sensitised. This provides a memory which can release the whole of its contents f or immediate and mas s ive parallel processing by subsequent devices .

The f ifth embodiment o f this invention relates to a cross-referencing device which depends upon the physical storage of the composite waveform taken by the compacted information generated by such devices as those exemplified by the first embodiment or- retained in a memory as in the fourth embodiment of this invention. Cros s -ref erencing is a function of the adjacency of one event in at least one dimension to the adjacency of another event and this will vary from equality of value, indicating coincidence in that dimension, to the maximum displacement in value allowed in the portrayed dimension. Thus a waveform which can express the value of any observation or event in all its measured dimensions can be compared directly with another observation or event, the differences between them becoming immediately apparent and being an expression of their relatedness by reason of their adjacency in each of their several dimensions.

This facility, which this invention bestows upon a computing device designed to carry out the necessary procedure, is suited to the massive parallel processing capability of the memory device as described in the fourth embodiment of this invention.

Reference is made to Figure 5. In this example of a device for revealing the relationships between events or between one event and events held in memory, a source 30 of electromagnetic radiation at a number of chosen wavelengths as described in the third embodiment above illuminates a screen 31 divided into domains 32 each of which comprises or is coated with a mixture of molecules with the reflective, transmissive or absorptive properties described in the second embodiment above.

The composite beam of electromagnetic radiation from the source 30 is composed of those components of wavelength which describe the selected event which is to be matched for likeness or relatedness with another event or with the events held in memory. The domains 33 of the screen 31 are individually controlled as described in the fourth embodiment to reflect, transmit or absorb those components of the radiation which together, as a de¬

composition of wavelengths, describe an event which is to be compared for likeness or relatedness to the event represented by the radiation from the source 30, or each domain may be set to represent a different event drawn from a memory store.

The particular pattern of electromagnetic radiation from the source 30 illuminates the screen 31 evenly or the domains 32 in sequence. Each domain 32, now set to reflect, transmit or absorb those components of any other electromagnetic radiation with wavelengths which accord with the pattern set to represent the particular event allocated to that domain 32, duly reflects, transmits or absorbs just those elements of the incident radiation from the source 30. Immediately adjacent to the back of the screen 31 is an array of electromagnetic radiation detectors 33 each of which will respond to any of the wavelengths of electromagnetic radiation which may be transmitted through the screen 31 in the particular domain 32 to which the relevant detector in the array of detectors 33 is allocated.

In the case where the sensitive material comprising the substance of the screen 31 is reflective in its response, then the disposition of the array of detectors 33 will be such, relative to the illuminated surface of the screen 31, as to make use of the reflected radiation as it would have done of the transmitted radiation. This is shown in Figure 5(B) by the alternative dispositions of the source 30 and the array of detectors 33.

The relationship between the event characterised by the component wavelengths of the composite beam emitted from the source 30 and another event or events held in memory and characterised by the settings of the domains 32 in the screen 31 will be revealed by the characteristics of the radiation detected by each of the detectors of the array 33. The remaining component wavelengths of the composite incident radiation from the source 30 are prevented by absorption, reflection or transmission from influencing the detectors 33.

The domains 32 which represent those events which are exactly the same as the event being represented by the pattern of component wavelengths in the composite beam emitted from the source 30 will reflect or transmit all of the incident beam to the detectors allocated to them in the array 33. Those which bear no resemblance will not pass on any illumination to their respective detectors. Intermediate relationships will be revealed according to the characteristics of the radiation received by the respective detectors in the array 33 and the detail of this relationship will depend upon the number of component wavelengths which the material of the screen 31 and of the domains 32 can be set to differentiate.

The sixth embodiment of this invention relates to the transmission and storage in memory of graphics or similar assemblies in which it is the generality of the patterns formed by the sensing devices which convey the information and which are not critically dependent upon the accuracy with which every element is portrayed for the meaning to be conveyed to the observer or observing device. This applied for instance to text readers, television pictures or recordings of speech or music.

In this embodiment a sensing device reads the physical measurements of observations and allocates a cell to each in a structure of cells as described previously in the first embodiment. Each of these cells is allocated a unique modulated continuous function as in the first embodiment and these functions are combined by addition into a single composite continuous function which can be expressed as a curve on a graph in which time is plotted against some other value. This composite curve is then passed to another proces sor which mathematically approximates this composite curbe over a period that equals a determined part of one cycle of the highest frequency that can be employed in transmitting or storing the information. This approximation may be done by the application of algorithms describing curves which follow the composite curve sufficiently closely but which can be translated into the behaviour of electrical systems , such as sine curve s , and transmitted or stored in memory accordingly.

On receipt or recall of this composite curve, which approximates the composite of the constituent curves of the continuous functions , it is analysed by any of the methods described in the first embodiment to reveal each of these curves . Alternatively , the unanalys ed compos ite curve constructed f rom the approximating curves , such as the sine curves to suit electrical transmission systems , may be used directly for comparison with other such curves to derive the results of comparisons without recalculation of the re spective representations of the component observations and their dimensional values . This proc e s s then give s a method of executing the calculation of the interaction of large assemblies of data in parallel and a very compact form.

Claims

CLAIMS :

1. A method of compacting data, whether stored or transmitted, which comprises allocating a uniquely valued modulated continuous function to each of a plurality of information channels, and combining these functions in a composite of modulated continuous functions .

2. A data compaction device which codes, decodes, stores or transmits data which it compacts or decompacts, comprising means to allocate a particular value to the constants of a modulated continuous function to relate one-to-one to each distinguishable channel of information, and means to combine these functions by an arithmetic procedure to produce a composite of said functions which is the property to be transmitted or stored or to be coded from or decoded to binary or analogue information.

3. A data compaction device comprising a coding apparatus which accepts a stream of data in the form of randomly ordered sequential ones and zeros and compiles them into a composite of continuous functions of 'x' with respect to 'y' with the dimensions of time and some measurable physical property.

4. A data de-compaction device comprising a decoding apparatus which accepts a composite of continuous functions of 'x' with respect to 'y' from a coding apparatus and decompiles it into a stream of ones and zeros which correspond on a one-to-one basis with the data which was given to the coding apparatus.

5. A data compaction device according to claim 2 or 3, which compares, adds, subtracts, multiplies or divides composites of continuous functions in one clock cycle of a computer.

6. A data compaction device comprising a coding apparatus which accepts a stream of data in digital form and compiles it into an electromagnetic spectrum of discrete, distinguishable energy levels.

7. A data de-compaction device comprising a decoding apparatus which accepts a spectrum of electromagnetic energy and relates the separable, component energy levels of that spectrum to a stream of digital data in one-to-one correspondence with the data which went to compile the spectrum in the coding apparatus .

8. A decoding apparatus which accepts a composite of continuous functions comprising compacted data from a coding apparatus and decompiles the composite into a digital stream in one-to-one correspondence with the digital stream which the coding apparatus compiled into that composite of continuous functions after receipt of less than the period of one cycle of the highest frequency that is in use by the coding apparatus or is incorporated into the formulation of any periodic function which is one of the components of the composite of continuous functions .

9. A device which compares stored composites of continuous functions and orders them with respect to the value of the adjacency of the data they incorporate.

10. A device which compares stored spectra of electromagnetic energy and orders them with respect to the value of the adjacency of the data they incorporate .

11. A data compaction device comprising a coding apparatus which orders the measurements of quantities in different dimensions pertaining to a specific observation, from a common origin.

12. A decoding device which samples the incoming voltage of a signal at intervals of time which are short in relation to the period of one cycle of the highest frequency in use in the decoding device and relates the pattern of voltages as they are sampled with respect to the voltages and the time at which they occurred to the pattern of continuous functions and their modulating properties which comprise the group of allocated functions to each of the observations or the positions of the digits in the originating digital signal.

13. A memory storage device which accepts a composite of continuous functions generated by a coding apparatus and stores the composite in such a manner that it can emit that composite of continuous functions later without significant distortion in the same or a lesser period of time and within the same constraints of frequency in the transmission medium as the coding apparatus.

14. A system of information or storage which can be made confidential to the legitimate users of the system by means of the arbitrary choice of the values of each of the modulated continuous functions allocated to each cell representing a multi-dimensional observation and itself a part of a structure of such cells.

15. A device as claimed in claim 2 or 3, in which an approximation to the composite of modulated continuous functions is compiled to further compact the body of information represented by that composite.

16. A device as claimed in claim 4, in which an expression which approximates the composite of a number of modulated continuous functions is resolved into this composite and from that into the component continuous functions.

17. A device as claimed in claim 2,3 or 4, in which expressions which are formed as approximations to composites of modulated continuous functions are compared and their relatednes s according to chosen rules of logic is derived.