FR2717281A1 - Coding system for characters of ASCII and foreign language symbols - Google Patents

Abstract

The system uses coding elements (C0-C4) each of which covers a specified hexadecimal range. The coding elements are arranged in groups of varying length (L) such that their combination provides ranges up to 224. Each group is symmetrical commencing and ending with the same element. The length (L) of each group and the hexadecimal range of the leading coding element defines the sub-domains in the whole available range. The characters to be coded are expressed in terms of tokens each comprising a category and a transform and the tokens are allocated to the sub-domains as defined by the coding group.

Description

Coding device for symbols such as characters
The invention relates to the encoding of symbols, such as characters, as well as their decoding.

One of the current bases of computer coding of symbols is the so-called ASCII table (for "American Standard Code for
Information Interchange "). It provides a standard 7-bit standard coding, the meaning of which is well known to specialists. The values 32 to 127 of the code (in decimal), normalized, cover the numbers from 0 to 9, the 26 letters of the Latin alphabet in their uppercase and lowercase versions, the punctuation marks or usual operations, as well as some special symbols and utility symbols of computer science.
ASCII (from O to 31) has a meaning which can differ depending on whether it is microcomputing, workstations, or large systems.

In most computer equipment, the basic storage unit is the byte, which covers 8 bits. As ASCII coding requires only 7 bits, the remaining bit has been used to extend the coding to other characters. When the first bit is 0, the byte is given the meaning of the ASCII code contained in the following 7 bits. When the first bit is 1, a new range of 128 possibilities is obtained, the "extended" ASCII, the meaning of which will vary according to the context. For example in the ISO 8859-1 coding (ISO LATIN 1), the 128 additional codes cover the accented characters of the main European languages, in the ISO 8859-7 coding, the 128 additional codes cover the Greek letters.

The "extended" ASCII meets the needs of European languages relatively well, but its extension possibilities are too limited to cover the character sets of Asian languages (Chinese, Japanese, Korean). Mixed coding systems in which the extended characters are coded on two bytes and the ASCII characters remain coded on one byte have been developed to solve this problem.

Despite significant efforts at standardization, there are still significant differences in the way of coding characters (ISO LATIN 1, OEM and Macintosh for European characters; Shift-JIS and EUC for Japanese characters). In addition, the absence of a coding which makes it possible to cover all the characters used in the world poses a problem for the coding of multilingual documents.

Experts have tackled this problem.

Thus was developed a technique called "UNICODE" which aims in particular to cover all the characters of all languages, with a precise and coherent nomenclature, suitable information as for cross references, dense coding, and a length in principle fixed coding , on two bytes, or 65,536 possibilities.

However, the Applicant has observed that this solution is not completely satisfactory, for reasons which will be developed later.

Consequently, the object of the present invention is to provide an improved symbol coding technique, in order to better satisfy the needs.

The device according to the invention comprises - a first channel, suitable for conveying an order signal which designates a symbol, - a second channel, suitable for carrying a symbol code, belonging to a chosen field of symbol codes, and - processing means for passing from the order signal to the symbol code.

According to one aspect of the invention, the order signal being formed from a predetermined set of digital order codes (called tokens), the processing means comprise, in combination: - first means, capable in response to the presentation of an order code, of providing a category datum, as well as a transform, - second means, suitable for establishing a correspondence between a predetermined plurality of sequences of transform intervals each associated with a category of tokens given (called a sequence of token intervals), and a plurality of respective disjoint sub-domains in the field of code symbols, - third means for sequentially analyzing the sequences of command codes of the order signal, as a function of a predetermined analysis rule, by obtaining from the first means the category data and the transform of each code-order, and from the second means the sub-domain associated with the sequence of intervals of j and are associated with the order signal, and - means for calculating the symbol code associated with the sequence of order codes analyzed as a function of the subdomain and of the transforms of the order codes thus obtained.

In a particularly advantageous embodiment, the processing means further comprise decoding means, suitable for matching a symbol code with the corresponding sequence of command codes. These processing means also include transcoding means, suitable for matching symbol coding of a different nature than that used for the corresponding sequence of command codes.

The method according to the invention comprises the following two steps a) a first step during which an order signal is established which designates a symbol, and b) a second step during which this signal is made to correspond order a code-symbol, belonging to a chosen field of code-symbols.

According to the invention, in step a), the order signal is established from a predetermined set of digital order codes.

Furthermore, step b) comprises the following operations b1) sequentially analyzing the order codes of the order signal, as a function of a predetermined analysis rule, b2) making a category data correspond to each order code , as well as an associated transform, b3) determine the sequence of token intervals associated with the sequence of order codes, b4) determine a sub-domain of the domain of symbol codes which corresponds to this sequence of jetOns intervals , and b5) calculating the symbol code associated with the sequence of order codes analyzed as a function of the subdomain and of the transforms of the order codes thus obtained.

Other characteristics and advantages of the invention will appear on examining the detailed description below, and the appended drawings, in which - FIG. 1 schematically illustrates an installation with a device according to the present invention; - Figure 2 schematically illustrates data exchanges to which the present invention applies; - Figures 3 and 4 define the coding ranges used in an example of implementation of the present invention - Figure 5 illustrates an example of a tree table usable for the implementation of the invention; - Figure 6 illustrates, in a simple example, two particular cases of coding implemented by the device according to the present invention.

In the present description, any sign used by humans to transmit information is considered to be a "symbol". A priori are targeted graphic symbols, without excluding symbols of another nature, for example sound.

When the context allows, the word "symbol" will be replaced by "character", and a series of symbols will be called "character string" or "text", according to the habit of those skilled in the art.

We will distinguish - the representation of a symbol in a text, which involves "symbol codes", or character codes, and - the order signal designating the symbol, which involves order codes ("tokens" , or "token" in English), of a different nature.

There is an essential difference between coding and decoding operations.

Coding is the operation which consists in associating a unique symbol code with one or more tokens emitted by a source of tokens which can be either a peripheral (such as for example a keyboard), or a software component.

Decoding is the reverse operation which consists in issuing one or more tokens from a unique symbol code stored in a memory (live or mass, in a file) of
central unit. Tokens can be sent to a specialized peripheral (such as a printer) or else processed by an appropriate software component which can extract useful information such as a lexical category (for example simple letter, accented letter, number, etc.). .) or an upper / lower case state.

As a general rule, the symbol codes will rather be used for the storage and exchange of textual data, the representation in the form of sequences of order codes will generally be temporary and internal to a software process. For example, a keyboard controller can synthesize the symbol codes to be sent to the processor from the command codes which correspond more or less directly to the keys activated by the user. Conversely, a printer controller can decompose the symbol codes into sequences of order codes in order to determine the constituents of a composite character. Any software can use a combination of decoding and coding operations, with an intermediate processing on the command codes to convert character strings (upper / lower case conversion).

The input and output channels of the coding and decoding processes are generally "logical" (software) channels, and not necessarily "physical" (hardware) channels.

The Applicant has studied in detail the UNICODE technique already mentioned. It focused in particular on the properties set out below.

The first property is the specificity of the coding. The same character can be represented on the screen or on paper in several different ways, depending on the font and the style (bold, italic) chosen. A distinction must be made between the glyph (the typographical variant) and the character. The role of a character encoding system is to assign a unique code to each character, which will cover all of the glyphs for that character. However, the glyph / character distinction is not always easy to make.

For example, subtle distinctions should be made between ligatures (fi, ffi) and digraphs (oe, ae). In most existing codings, digraphs are considered as original characters while ligatures are treated as typographical variants.

Sometimes two characters are represented by the same glyph.

For example, the Latin letter "a" and the Greek letter "alpha" have the same uppercase representation. However, these are two different characters to which the coding must assign different codes. The decision to distinguish or unify characters with the same glyph is less obvious when dealing with characters such as the Greek letter "mu" and the prefix "micro" of a unit of measure or cases such as the exclamation point "!" and the mathematical symbol "factorial".

A third example is that of the "half-width" and "normal width" variants of the Asian characters. The codings currently in force assign different codes to these variants and, therefore, a universal code must provide a mechanism to distinguish these variants so that the texts obtained using the current codings can be converted without losing information. .

As these three examples show, any coding implies a certain number of arbitrary decisions as to the degree of specificity of the characters. One of the characteristics of the proposed coding system is to allow characters to be specified as much as desired and therefore to provide consistent and complete character sets (for example a set of punctuation marks containing the exclamation point and a set of mathematical symbols containing the factorial sign) with effective means for categorizing characters from their codes and for decomposing composite characters (digraphs, accented characters) into their constituent elements.

The combinatorial nature of the proposed coding and the large size of the code space available are fundamental to achieving this goal.

A second property is compatibility with the above-mentioned ASCII base table. A string consisting only of ASCII characters is coded by UNICODE in a different way from the equivalent ASCII string, because, in UNICODE, each ASCII character is transformed into double byte coding (the first of which is Oh, that is say zero in hex).

A third property is uniqueness. It is highly desirable that there is only one way of representing a given symbol, at the level of specificity chosen (which is not true in general, especially in UNICODE). As an example of non-uniqueness, the character "to" can be represented either as ['to Latin' lowercase] or as E'a Latin 'lowercase] + [' grave accent 'without spacing]. This has many annoying consequences, especially when comparing strings to detect their strict identity. Indeed, it is necessary to interpret beforehand any character capable of at least two different representations.

Another property is called "locality". Fishing for lack of locality a coding technique which requires knowing information contained elsewhere in the text (character string) processed. So although the coding
UNI CODE has in principle a fixed length of two bytes, it nevertheless authorizes coding on several groups of two bytes for certain characters. (See the above example: ['a Latin' lowercase] + ttaccent grave 'without space], which is a single character encoded in two groups of two bytes). This also has many harmful consequences. For example, always when exploring two strings for comparison, you have to look further forward, to avoid considering as identical a "to" and an "a".

As another property, we also note the extensibility. Most of the time, a few places are reserved for characters to choose from. But any coding principle of fixed length (as on two bytes) prohibits the extension beyond a given limit (like 65536), except to introduce non-local modal mechanisms, with the serious disadvantages which are attached to it.

Under these conditions, the object of the present invention is in particular to provide a coding means which makes it possible to "specify" any symbol up to the desired level of sophistication, without unduly complicating its subsequent processing.

The invention also aims to allow easy extensibility of the running-in technique, strict ASCII compatibility, with of course good compactness and efficiency, as well as coding.

The present invention also aims to provide a coding means which can easily and systematically be made "unique", that is to say in which there is only one way of representing a given character, up to the level of specificity desired.

The invention finally aims to provide a coding means which reduces as much as possible the need to use expedients which make an exception to the desirable properties of "locality" and "non-modality".

According to an important aspect of the invention, the proposed coding technique is associated with an algebraic description of the characters. This means that a great deal of information can be obtained directly from the sequence of code codes designating the character, without the need to inspect an additional table.

The invention applies to many types of computer installations.

A basic installation (FIG. 1) comprises a central unit 1, associated with a keyboard 2 and a screen 3, as well as a mass memory 4 such as a hard disk and a printer 5, it being observed that, if being a network, the mass memory and the printer can be located remotely.

In operation, the applicable principle diagram becomes that of FIG. 2, in which the processor 10 and the random access memory 12 are distinguished in the central unit 1. There is the keyboard 2, the screen 3, the printer. 5 and the mass memory 4, possibly a modem or a network link 6.

The processor 10 may be implementing several software; a distinction was made here between software A or 91 and software B or 92.

Text data, that is to say containing symbols within the meaning of the present invention can be exchanged not only between the processor, the keyboard, the screen, the printer, the random access memory, the modem and the memory. mass, but also from software to software. In addition, such text data can be encountered in quite different situations: a so-called text file will in principle only contain text data. On the other hand, a program file may contain instructions on the one hand, and text data on the other.

Similarly, orders designating symbols may be defined not only at the level of the peripherals such as the keyboard, the screen or the printer, but also in the internal dialogues with the memories and / or between software.

In general, this can be defined by considering - a first channel suitable for conveying an order signal which designates a symbol or "character", and - a second channel suitable for conveying a symbol code.

In known manner, this symbol code must belong to a chosen domain of symbol codes, and it is the role of the processor 10 to pass from an order signal to the corresponding symbol code, and vice versa.

For this purpose, it is conventional to use tables, which are generally stored in files.

In most cases, the table will directly provide the correspondence between the order relating to a particular symbol and the associated symbol code.

The Applicant has observed that this method of proceeding has a major drawback.

Indeed, there are many situations where a symbol presents variants, such as the case of "a" and "to, already mentioned.

In advanced computer applications, it is desirable to be able to distinguish between these variants, since this distinction constitutes useful information. But it is also desirable to be able to measure their similarity, which is also useful information.

With classic table systems, when you want to distinguish two variants of the same symbol, you necessarily end up with two different codes. And the only way to additionally have similarity information is to create other tables, which group the codes corresponding to similar symbols. But this solution is difficult to implement when it is necessary to treat not only the usual accented characters, but also complex sets such as the syllabic characters of the oriental languages.

But the first problem which arises is to ensure the specificity of the coding, and its extensibility. Indeed, as soon as we want to complicate the specification of a symbol, we must be able to extend the coding technique used.

For this, the invention firstly provides that the order signal designating a symbol consists of a predefined set of digital order codes, which will be called hereinafter "tokens".

This game must be the same regardless of the source of the data to be processed in the application concerned. That is to say that if, for example, a keyboard is used which has its own coding mode, this order coding mode of the keyboard should be converted into the set of command codes of the present invention.

For a good understanding of the present invention, it is preferable to first describe the field in which the symbol codes are defined (FIGS. 3 and 4).

The extensibility of the symbol codes according to the present invention is ensured by the fact that the coding is not carried out over a fixed length. By way of example, FIG. 4 illustrates the ranges of definition of bytes CO to C4, with the range limits in hexadecimal, and the numerical extent of each range in decimal. Here, CO is free, and can range from
OOh to FFh (according to the current notation of a byte in hexadecimal, identified by the suffix "h"). FIG. 3 illustrates four possible coding lengths of symbol codes L, and the number of possibilities that they offer. By definition, we associate a given subset with a given coding length. Each subset offers a number of possibilities for code-symbols which grows with the length it represents. This notion of subset, if it makes it possible to clarify the understanding of the present description, is however not necessary for the definition of the invention.

We define a subset as consisting of terminal bytes belonging to ranges of identical values. However, these two bytes will not necessarily have identical values.

- L = 1, with a single byte C1, defined on the range OOh-7Fh.

(this single byte is both the beginning and the end of the symbolic code).

- L = 2, with a start byte belonging to the range C2 and an end byte C2, both defined in the range 80h-BFh.

- L = 3, with a start byte belonging to the range C3 and an end byte C3, both defined on the range COh-DFh, the intermediate byte CO being free.

- L = 4, with a start byte belonging to the range C4 and an end byte C4, both defined on the range EOh-EFh, the two intermediate bytes CO being free.

These four subsets already offer 128, 4096, 218 and 224 respectively for coding different symbol codes. The free range from FOh to FFh allows the extension to other subsets even richer in possibilities.

Note that the C1 symbol code consisting of a single byte can use the standard ASCII code.

The symmetry of definition of the two terminal bytes (start and end) has two advantages - it allows a verification first; - secondly, it makes it possible to browse symbol code chains in both directions (left to right and right to left) with the same efficiency.

It should however be noted that it is not strictly speaking the symmetry of the coding which intervenes here but the fact that for a given code length (L), the start and end bytes are forced to belong to ranges very precise values and that the ranges corresponding to different lengths are disjoint. The fact of having chosen the same ranges for the initial and terminal bytes is not essential but simplifies the coding and makes it possible to use a single table instead of two for the traversing of strings in both directions and the verifications.

Sub-domains are defined in the subsets of L = 2 to L = 4, using information associated with the first byte of the code. For example, the subsets can be partitioned as follows.

L Value of the 1st byte Organization 2 80h - 9Fh 16 sub-domains of 128 2 AOh - AFh 16 sub-domains of 64 2 BOh - B7h 16 sub-domains of 32 2 2 BFh 64 sub-domains of 1 3 COh - c7h 8 sub-domains of 8192 3 C8h - CFh 16 sub-domains of 4096 3 3 D8h - DBh 1 sub-domain of 32768 (hangul) 3 DCh - DFh 1 sub-domain of 32768 (kanji)
This means makes it possible to arbitrarily partition a very large coding space by means of a simple table with 256 entries, each entry containing - the number of bytes on which the character is coded (L), - the number of sub- domains for this value of the initial byte or the sequence number of the initial byte inside a sub-domain extending over several initial bytes (case of the kanji sub-domain).

By means of a simple table, it is thus possible to partition the coding space into a large number of subdomains of various sizes. In practice, the sizes of sub-domains will be powers of 2, in order to allow more effective treatments.

These subdomains will be used to code sets of characters forming a coherent whole. For example, the 24 lowercase Greek letters and the 24 capital letters will fill a sub-domain of 64 codes, the hangul syllables formed by a hangul consonant (30), a hangul vowel (21) and a hangul consonant (30 ) will (partially) fill in a sub-domain of 32768 codes.

The next step in coding is to express the symbols to be coded as a sequence of tokens. Some tokens will designate basic forms or characters (for example the Latin letter "A", the grave accent, ...), other tokens will designate operations altering the basic forms (for example upper / lower case, rotations, ...).

Each token will consist of a category and a transform.

For example, the token "Latin letter A" will be the transform number 0 of the category "Latin letter", the token "Latin letter Z" being the transform number 25 of this same category. The upper and lower case operators will be two transforms of the category "upper / lower case operators".

By expressing the symbols to be coded in the form of sequences of tokens thus formed, we will obtain a set of sequences in which we can identify "regular" subsets. For example, the upper and lower case Latin letters will all be expressed in the form of a token of the "Latin letter" category followed by a token of the "upper / lower case operators" category. Typically, the regular sequences will consist of tokens belonging to well-defined categories but having different transforms.

Coding will therefore consist in associating each regular sequence with one of the coding sub-fields defined above. A description of the continuation of the categories will be attached to the coding sub-domain. Each symbol to be coded will have a unique sequence number within its sub-domain. This sequence number will be obtained using a simple and generic calculation on the transforms of the tokens constituting the symbol (for example for the upper and lower case Latin letters, we can multiply the upper / lower case transform (1 or 0 ) by 32 and add the transform "latin letter" (0 to 25). It will of course be necessary to choose the size of the sub-domain associated with a regular sequence so that this sub-domain can contain all the possible transforms of the regular suite.

The coding mode described above is well suited to regular and dense character strings. There are however a certain number of singular characters or very scattered sequences (for example, the digraphs of Latin letters are very few in comparison with the 26 x 26 x 2 possible digraphs). To take these special cases into account in the coding system, regular sequences will be allowed to describe sequences of token intervals and no longer simply sequences of token categories, a token interval being an interval of transforms for a category given tokens. The token intervals may be reduced to a single transform, for example in the case of the encoding of singular characters. The organization of the subdomains makes it possible to define a large number of small subdomains (possibly a single code) which will be associated with these particular cases.

For its part, the central processing unit is designed to constitute - a first automaton, capable of responding to the presentation of an order code, by supplying therefrom, on the one hand, a so-called category datum , on the other hand a transformed, - a second automaton, which has a predetermined plurality of sequences of token intervals, to which it is capable of matching a plurality of respective sub-domains disjoint in the chosen domain of codes- syibo- les. Consequently, when presented with a particular sequence of tokens, this second automaton will identify the sequence of corresponding token intervals and designate one of the sub-domains in the field of symbol codes, - a third automaton, capable to sequentially analyze the order codes of a received order signal. The sequential analysis of these order codes does not pose a problem of beginning or end, because it is an order, or if we prefer an instruction, in which case any computer system knows a priori how to determine what are the beginning and the end of the order or instruction.

This analysis is carried out according to a predetermined analysis rule. The easiest way is to take the order codes in the order in which they appear.

For each order code, this third automaton pulls the category data from the first automaton and transforms it correspondingly.

When the sequential analysis of the order codes is finished, the third automaton draws from the second automaton the sequence of token intervals and the indication of the sub-domain associated with the sequence of couples (category sub-domain and transforms of the codes -Orders thus obtained.

In this way, it is ensured that it is possible, by lengthening at will the sequence of order codes, to be able to specify any symbol code with the desired level of sophistication.

This calculation can be made very simple. Indeed, the category data, as well as the transformed are a conversion of each code-order. The corresponding token interval must cover a certain numerical range, which will be rounded up to the power of 2 higher. To calculate the sequence number of the symbol in its subdomain, we will start with a value of O and, for each code, we will multiply the value of the sequence number previously obtained by the numerical range of the interval of corresponding token (rounded to the power of 2 higher) and we will add the difference between the transform of the order code and the lower bound of the corresponding token interval.

A typical example of sequence number calculation is illustrated in Figure 6. The symbols E and è in this figure should be considered as theoretical examples, not the Latin letters, which are already part of the ASCII table.

In practice, the Latin letter è is coded according to the invention by LA + 0 * 25 + 4.

From this sequence number and from the data in the sub-domain, the symbol code can be calculated.

The uniqueness of the coding according to the present invention can be ensured at the base by the fact that the predetermined plurality of intervals of recognized tokens is chosen so that there exists a single series of intervals of tokens for a single and same symbol. specific to code. This in itself is sufficient to ensure uniqueness.

Incidentally, if it happens that a series of order codes results in a series of token intervals that do not belong to said predetermined plurality, we are then in an error situation for which there are general techniques known processing methods, of which it is not necessary to describe here how they could apply to the case of the present invention.

It is possible to distinguish operator order codes and operand order codes, and to consider sequences of order codes as algebraic formulas. This distinction is perfectly known to those skilled in the art, so it will not be the subject of a detailed description.

To decode, you must first extract the first byte from the symbol code. From this first byte, the length of the symbol code (in bytes) and information describing the organization into sub-domains are obtained by simple table consultation. From this information and the content of the following byte (s), the coding sub-domain and the sequence number inside this sub-domain can be determined. Then, we consult a table which provides for each coding sub-domain the sequence of corresponding token intervals (inverse of the second coding machine).

From the sequence number in the sub-domain and the sequence of token intervals, we can recalculate the transforms corresponding to each of the tokens in the sequence (by a series of divisions and modulo extractions, inverse of the series of multiplications and additions made by the third coding machine). The decoded sequence of tokens is thus obtained.

For transcoding, we will use tables that can be generated automatically from descriptions of correspondences: sequences of tokens <-> external symbol code.

It is the same at the level of transcoding which may be required to pass from code-symbols expressed by another technique than that described here, which can also be done using a simple table.

In other words, the machine coding of symbols is carried out according to the invention in the following manner
In a first step, an order signal is established which designates a symbol. This order signal is established from a predetermined set of digital order codes.

In a second step, this order signal is made to correspond to a symbol code, belonging to a chosen field of symbol codes. This second step comprises the following operations: 1) the sequence of order codes constituting the order signal is analyzed sequentially, as a function of a predetermined analysis rule; 2) a category datum, as well as an associated transform, is made to correspond to each order code of said sequence; 3) the sequence of token intervals associated with the sequence of order codes is determined; 4) a subdomain of the domain of symbol codes is determined which corresponds to this sequence of token intervals; and 5) the symbol code associated with the sequence of order codes analyzed is calculated as a function of the subdomain and of the transforms of the order codes thus obtained.

Claims (9)

Claims 1. Electronic symbol coding device, comprising: - a first channel, suitable for carrying an order signal which designates a symbol, - a second channel, suitable for carrying a symbol code, belonging to a chosen field of codes - symbols, and - processing means for passing from the order signal to the symbol code, characterized in that, the order signal being constituted from a predetermined set of digital order codes (called tokens), the processing means comprise, in combination: - first means, capable in response to the presentation of an order code, of supplying a category datum, as well as a transformed, - second means, suitable for establishing a correspondence between a predetermined plurality of sequences of transform intervals each associated with a given category of tokens (called sequence of token intervals), and a plurality of respective disjoint sub-domains in the domain of codessymbols , - third means for sequentially analyzing the sequences of order codes of the order signal, as a function of a predetermined analysis rule, by obtaining from the first means the category data and the transform of each order code, and second means the subdomain associated with the sequence of token intervals associated with the order signal, and - calculation means for calculating the symbol code associated with the sequence of order codes analyzed as a function of the subdomain and transformed from the order codes thus obtained. 2. Device according to claim 1, characterized in that the sub-domains are respectively associated with intervals of values of the elementary coding byte, each sub-domain being constrained so that the byte of coding start and the byte of end of the coding belong to the same range of values, whereby the data of a sub-domain makes it possible to obtain the length of the coding. 3. Device according to claim 2, characterized in that the coding with a single byte corresponds to the standard ASCII code at 7 bits. 4. Device according to one of the preceding claims, characterized in that the transforms are directly obtained from among the least significant bits of the order code, while the categories are obtained from among the most significant bits of the code- order. 5. Device according to one of the preceding claims, characterized in that said predetermined plurality of sequences of token intervals is chosen so that there is a single sequence of token intervals for a single specific symbol to be coded . 6. Device according to one of the preceding claims, characterized in that the calculation means are suitable for calculating and keeping a quantity equal to the sum of the difference between the transform of a token and the lower limit of the interval of this token, and the product of the previous value of this quantity by the numerical extent of said interval. 7. Device according to one of the preceding claims, characterized in that the processing means further comprise decoding means, adapted to make a corresponding symbol code sequence correspond to a corresponding code code. 8. Device according to one of the preceding claims, characterized in that the processing means further comprise transcoding means, adapted to correspond to a coding of symbols of a different nature than that used following the sequence of command codes corresponding. 9. Machine coding method, in which: a) an order signal which designates a symbol is established, and b) this order signal is made to correspond to a code-symbol belonging to a chosen field of symbol codes, characterized in that - in step a), the order signal is established from a predetermined set of digital order codes, and - step b) comprises the following operations:  bl) sequentially analyzing the order codes of the order signal, according to a predetermined analysis rule,  b2) match each order code with a category datum, as well as an associated transform,  b3) determine the sequence of token intervals associated with the sequence of order codes,  b4) determine a sub-domain of the domain of sy-codes} - boles which corresponds to this sequence of token intervals, and  b5) calculating the symbol code associated with the sequence of order codes analyzed as a function of the subdomain and of the transforms of the order codes thus obtained.