Data Compression Method and System

Europaisches Patentamt J European Patent Office © Publication number: 0 678 986 A1 Office europeen des brevets

EUROPEAN PATENT APPLICATION

@ Date of filing: 21.04.95

© A lossless type data compression method em- generated by registering the combined strings hav- ploying a dictionary system is suitable for character ing occurrence frequency higher than a given value generator of a game machine and so forth. A work- with a dictionary number. The combined strings in ing data strings are generated from an original data the data stream are replaced with the dictionary stream. Two sequential working data strings are numbers corresponding to the combined strings in combined to form a combined string. A dictionary is the dictionary.

FIG. 1 100

RAM _ (ORIGINAL DATA (DATA STREAM) — (NUMBER OF CYCLES TO — CO STREAM) REPEAT DATA 00 COMPRESSION) Oi , CPU RAM COMPRESSED 00 (DICTIONARY DATA REGISTER " CO DATA)

ROM (DATA COMPRESSION PROGRAM)

Rank Xerox (UK) Business Services (3. 10/3.09/3.3.4) 1 EP 0 678 986 A1 2

The present invention relates generally to a is next age coding system in the facsimile. The data compression method and system. More spe- JBIEG may handle redundant data stream which cifically, the invention relates to a lossless type cannot be handled by Huffman coding system, by data compression method employing dictionary unitary compression based on probability of occur- system suitable for a character generator for a 5 rence of strings. Thus, the JBIEG realizes optimal game machine and so forth. data compression in view of information entropy. Conventionally, various data compression In general. LZ system is a data compression methods have been developed for reduction of system which performs data compression detecting necessary capacity of storage devices in data pro- repetition of strings. The LZ system is applied for a cessing systems and for improvement of data io data compression tool for personal computers or transmission efficiency. The data compression for data backup cartridge tape recording apparatus method may be generally divided into lossy type and other products. and lossless type in the viewpoint of capability of The LZ system generally includes LZ77 sys- bidirectional coding. tem, Ziv J. and Lempel, A. "A Universal Algorithm The lossy type data compression method is 75 for Sequential Data Compression", IEEE Transac- non-reversible coding system. JPEG (Joint Photo- tion on Information Theory, vol. IT-23, No. 3, pp graphic Coding Expert Group), MPEG (Moving Pic- 337-343, September, 1997 and LZ78 system, Ziv, ture Image Coding Expert Group), H.261 for PMS J. and Lempel, A. "Compression of Individual Se- (Picture-phone Meeting Service) or picture tele- quences via Variable Rate Coding" IEEE Transla- phone and so forth are internationally standardized 20 tion on Information Theory, vol IT-24, No. 5, pp 530 systems of this lossy type data compression. The to 536, September, 1978. The former LZ77 system lossy type data compression is advantage for high is also disclosed in U. S. Patent Nos. 5,003,307 compression rate, i.e. approximately 1/50 to 1/1000 and 5,016,009. The later LZ78 system is also dis- while loss of information amount is caused. closed in U. S. Patent Nos. 4,558,302 and On the other hand, the lossless type data com- 25 4,814,746. Algorithms of LZ77 and LZ78 systems pression method is a reversible data coding sys- are in common at the point where the currently tem. This type of data compression system gen- objective string and processed strings are com- erally holds data compression rate approximately pared and the longest matching string is obtained 1/2 and thus cannot achieve high compression rate through the comparison. However, LZ77 stores the as achieved by the lossy type data compression 30 processed strings in a buffer and takes means to method. However, the lossless type data compres- handled the processed data as if the processed sion method is advantageous for capability of en- data in the buffer is slide on the input data stream. coding and decoding without loosing an original On the other hand, the LZ78 system employs data. Run Length coding, Huffman coding, means for assigning dedicated codes for pro- Arithmetic coding, LZ (Lemple-Ziv) system and so 35 cessed strings and registering the codes in dic- forth are typical standardized systems in the los- tionary style. sless type data compression methods. As a result, in comparison of LZ77 and LZ78 The run length coding system is the simplest systems in terms of function, LZ77 system is supe- lossless type coding system. The system utilizes rior than the LZ78 system in terms of compression the fact that probability of appearance is differen- 40 rate, and LZ78 system superior than the LZ77 tiated depending upon the value of the run length. system in terms of data processing speed. Therefore, by assigning shorter code for the run On the other hand, in the field of gate machine, length having higher probability, data compression requirement for high level image expression is pro- is achieved. This coding system has been em- gressively growing. In the commercial gate ma- ployed in CD-I (Compact Disc-Interface), Video for 45 chine, image expression utilizing three-dimensional Windows (Trademark: Microsoft) and so forth. CG (computer graphics) has been employed. The Huffman coding system is a data compression trend is extending to home use television game system primarily used in the field of image pro- machines and multimedia systems. Thus, develop- cessing. MH (Modified Huffman) coding of G3 stan- ment for data processing systems capable of such dard facsimile and so forth are application of the 50 high level image expression are progressed. Huffman coding. Complication of image expression causes in- It should be noted that JPEG, MPEG, or H.261 creasing of data amount. Therefore, a demand for also employs Huffman coding. However, since lossless type data compression method having these method use DCT (Discrete Cosine Transfer) high compression rate and capable of high speed in preparatory process, they are classified as lossy 55 encoding and decoding, is growing. type. Particularly, in case of the game machine, un- Arithmetic coding system is used in JBIG less the display screen reacts to operation of a (Joint Bi-level Image Coding Expert Group) which button on a control pad by a user within 1/60 to

2 3 EP 0 678 986 A1 4

1/30 seconds (corresponding to display period of order in descending order from working string hav- one or two field of color television signal), game ing largest occurrence frequency, and having oc- becomes less interesting. Therefore, it is inherent currence frequency greater than or equal to 3; to achieve both of the high compression rate and performing fourth process step for registering high speed encoding and decoding. 5 compression dictionary data of (S + 1) bits con- In this sense, the above-mentioned the LZ sys- sisted of dictionary number and compression iden- tem (particularly LZ78) and LHA system, in which tifier bit, in the second storage means, correspond- the LZ system and Huffman coding system may be ing to each of combined strings detected by the said as suitable data compression method as data third process step; and compression method. io firth process step for replacing combined string However, in case of the LZ system, since re- among combined strings in the working data spective of the individual strings as object for com- stream matching with one of combined strings reg- pression have variable, and algorithm for compres- istered in the second storage means, with the com- sion and decompression is complicate, the data pression dictionary data corresponding to the processing procedure may contain large number of is matching combined string, steps and the hardware construction may become repeating the third to fifth process steps for R complicate. times with taking data stream replaced through the On the other hand, in the character generator fifth process step as working data stream for out- in the game machine, a relatively small data block putting data stream stored in the first storage is handled. In the conventional data compression 20 means and all combined string and compression method, compression becomes impossible or at dictionary data stored in the second storage least insufficient in compression. means, after R times repletion, as compressed Also, the data block of the character generator data. in the game machine, in has a tendency to have In the alternative, when R = 1, the second high probability of occurrence of the same string, 25 process step is performed without adding the non- and not to cause significant variation of number of compression identifier bit to establish the working the data blocks corresponding to variation mode of data stream take the original strings as the working the characters. strings, and the fourth process step is performed Therefore, it is an object of the present inven- with formulating the compression dictionary data tion to provide a lossless type data compression 30 solely with the S bit dictionary number. method which permits high speed compression In the foregoing construction, the number of and decompression and can realize high compres- repetition cycles of the third to fifth steps is des- sion rate irrespective of the size of data blocks. ignated in the first process step. The designated According to one aspect of the invention, a number of repetition cycles can be designated data compression method, in which a data stream 35 arbitrarily for appropriate for obtaining desired of a digital input signal is stored in first storage compression ratio within the range less than or means and data compression of the data stream is equal to the maximum register number D of the performed with registering strings appearing in the directly generated by the second process step. data stream in a direction of second storage The second process step is a preparatory pro- means, comprises: 40 cess for generating the working data stream con- performing first process step for designating R sisted of working strings of (S + 1) bit length by representing a repetition number of the following adding non-compression identifier bit (1 bit) for third to fifth process steps wherein R is an integer respective of the original strings (S bits) in the smaller than or equal to D, with taking a total data original data stream so that the working strings in capacity of the original data stream being S x N 45 non-compression state can be identified. bits, wherein S is an integer greater than or equal Generation of the dictionary and data compres- to 2 and N is an integer greater than or equal to 3, sion process are performed through the third to and a maximum dictionary register number D fifth process steps. Generation of dictionary is per- where D is 2s; formed per [2x(S + 1) bits] by combining sequential performing second process step for separating 50 two working strings in the working data stream. the original data stream of S bit length stored in the At first, the combined string having occurrence storage means and generating a working data frequency higher than 3 is detected in the third stream by adding a non-conducting identifier bit for process step. Then, in the fourth process step, conversion into a working data stream having corresponding to the detected combined string, the (S + 1 ) bit working strings; 55 compression dictionary data is registered in the performing third process step for detecting second storage means. Here, the compression dic- combined string consisted of two sequential work- tionary data is consisted of the dictionary number ing strings in the working data stream up to D/Rth and the compression identifier bit to have bit length

3 5 EP 0 678 986 A1 6 of (S + 1) same as the bit length of the working Namely, with detecting the compression/non- string. compression identifier, the compressed bits are In the fifth process step, data compression present, the working string is replaced with the utilizing the dictionary is performed. Namely, the combined string in the dictionary corresponding to combined strings matching with combined strings 5 the identifier. On the other hand, when the com- registered in the dictionary are replaced with re- bined string holds non-compression signal, com- spectively corresponding compression dictionary bined string is maintained as is and next working data. string is detected and processed. Finally, by re- As a results, with respect to the combined moving the non-compression identifier from the string in the working data stream as objective for io working stream, the original data stream can be replacement, 2x (S + 1) is replaced with (S + 1). obtained. Since the combined strings registered in the dic- Besides, in the third process step, the con- tionary has at least 3 of the occurrence frequency, straint is applied has "occurrence frequency is up 3x(S + 1) or more data can be deleted with respect to a given value (D/R)th" to provide upper limit for to one register in the dictionary. is the number of the combined strings to be detected. On the other hand, with respect to the data This is because that in the third to fifth process compression at the working data stream, data steps in the repeating cycles, it is possible that the amount is increased for 3x (S + 1) bits or more per number of combined strings may exceed the maxi- one register. mum register number. By restriction set forth Accordingly, cancelling increased data and de- 20 above, the designated repetition cycles for des- creased data, when the occurrence frequency of ignated number of cycles R. the combined string is 3, the data amount is held Alternately, it is possible to set R = 1 which is unchanged. The data compression becomes effec- the special setting not to repeat the third to fifth tive when the occurrence of frequency is greater steps. than or equal to 4. 25 In this case, by providing correspondence be- The third to fifth process steps are repeated tween the two original string as the objective for until the designated R timers designated in the first compression and the dictionary number, decom- process step. At every cycle, the data stream re- pression can be enabled. placed in the fifth step is handled as working data Therefore, in the dictionary generation and stream in the third process step in the next data 30 compression process steps, the non-compression compression cycle. identifier bit is not added to the original string in Accordingly, in the second and subsequent the second process step to form the working data dictionary generation and data compression pro- stream with taking the original string as S bits cess steps the non-compressed working strings working string as is. Also, in the fourth process and the compression dictionary data are mixedly 35 step, the compression dictionary data is solely con- presented. In the third to fourth process steps, with sisted of the S bit length of dictionary number. respect to the working strings as objective, the In certain original data stream, higher speed in combined string having the occurrence frequency compression and decompression is required rather greater than 3 can be newly registered in the than the compression ratio. The foregoing alter- compression dictionary. Then, in the fifth process 40 native is suitable in such case. step, among combined strings in the working data According to the second aspect of the inven- stream, the combined strings corresponding newly tion, a data compression system comprises: registered combined register are replaced with the first storage means for storing data; newly generated compression dictionary data. second storage means for storing data; As a result, by repeating the third to fifth pro- 45 first means for receiving an original data cess, higher compression ratio can be obtained. stream and a repetition command indicative of a It should be noted that with respect to the desired repetition number of data compression pro- combined strings having the occurrence frequency cess cycles and separating the original data stream higher than or equal to 3, the compression process into a plurality of sequential data strings having a will not have effect at that time. However, it will 50 fixed unit data length; contribute in the subsequent dictionary generation second means for comparing combined data and compression procedure in the repeated cycle. strings established by given number of sequential On the other hand, in the decompression pro- data strings with a content of the first storage cess of the compressed data, with the data stream means, for registering occurrence data and the and the data registered in the dictionary, the origi- 55 combined data string at the first occurrence of a nal data stream can be reproduced by reversed coincidence between the combined data and the algorithm with utilizing the dictionary. content in said second storage means, and for incrementing the occurrence data at second and

4 7 EP 0 678 986 A1 8 subsequent occurrences; fifth means for generating control signals for third means for generating a dictionary by reg- controlling operations of the first to fourth means istering the combined data strings which has occur- for repeating data compression cycles for the de- rence data satisfying a predetermined register con- sired repetition number indicated in the repetition dition, together with dictionary identifiers thereof, in 5 command. the second storage means; According to the fourth aspect of the invention, fourth means for replacing combined data an operation method of compression system in- strings in the data stream with the corresponding cluding first and second storage, comprising the dictionary identifiers; and steps of: fifth means for controlling operations of the first io receiving an ordinal data stream and a repeti- to fourth means for repeating data compression tion command indicative of desired repetition num- cycles for the desired repetition number indicated ber of data compression process cycles and sepa- in the repetition command. rating the original data stream into a plurality of The first means may add a compression state sequential data strings having fixed unit data identifier bit for each fixed length of data fraction in is length; the data stream to formulate the data string. comparing combined data strings established The third means may register the dictionary by a given number of sequential data strings with a identifier incorporating the compression status content of the first storage for registering occur- identifier indicative of compressed state. rence data and the combined data string at the first The second means may form the combined 20 occurrence of a coincidence between the com- data string with two sequential data strings of fixed bined data strings and the content and for incre- unit data length, and the third means generates the menting the occurrence data at second and subse- dictionary identifier having data length correspond- quent occurrences; ing to the fixed unit data length. generating a dictionary by registering the com- According to the third aspect of the invention, a 25 bined data strings which has occurrence data sat- data compression system comprises: isfying a predetermined register condition, together first storage means for storing data; with dictionary identifiers thereof, in the second second storage means for storing data; storage; first means for receiving an input sequential replacing combined data strings in the data data signal and a repetition command signal indica- 30 stream with the corresponding dictionary identifiers; tive of a desired repetition number of data com- and pression process cycles and sequentially extracting controlling operations of the first to fourth fraction signals of a given fixed bit length from the means for repeating data compression cycles for input sequential data signal; the desired repetition number indicated in the rep- second means for sequentially extracting seg- 35 etition command. ment signals having a bit length plurality times of The present invention will be understood more the given fixed bit length of the fraction signals fully from the detailed description given herebelow from input sequential data signal, comparing the and from the accompanying drawings of the pre- content of each of the segment signal with a con- ferred embodiment of the invention, which, how- tent of the first storage means, for registering an 40 ever, should not be taken to be limitative to the occurrence counter signal value and the segment present invention, but are for explanation and un- signal in said second storage means at the first derstanding only. occurrence of a coincidence between the contents In the drawings: of each of said segment signal and the contents of Fig. 1 is a block diagram showing a data com- said first storage means, and for incrementing the 45 pression system according to the preferred em- occurrence counter signal value at second and bodiment of the present invention; subsequent occurrences; Fig. 2 is a flowchart showing a procedure of third means for generating a dictionary by reg- generation of a working data stream; istering, in said second storage means, the seg- Fig. 3 is a flowchart showing a procedure of ment signals corresponding to the occurrence 50 generation of a dictionary; counter signal satisfying a predetermined register Fig. 4 is a flowchart showing a procedure of condition, together with dictionary identifier signals compression; thereof; Fig. 5 is an illustration showing an example of fourth means for replacing data segment in the working data stream to be generated from an input sequential data signal corresponding to the 55 original data stream; segment signals registered in the second storage Fig. 6 is an illustration showing a register data of means with the corresponding respectively cor- a dictionary generated in the first cycle of dic- responding dictionary identifier signals; and tionary generation procedure;

5 9 EP 0 678 986 A1 10

Fig. 7 is an illustration showing a working data Next, the data compression process of the stream generated in the first cycle of compres- preferred embodiment of the invention will be dis- sion procedure; cussed with reference to the flowcharts of Figs. 2 Fig. 8 is an illustration showing a register data of to 4 and data tables illustrated in Figs. 5 to 12 a dictionary generated in the second cycle of 5 showing the contents of actual data stream and dictionary generation procedure; dictionary register data. Fig. 9 is an illustration showing a working data At first, when the repetition number command stream generated in the second cycle of com- R is input from the host system 200, CPU 4 stores pression procedure; the value thereof in an internal buffer. Thereafter, Fig. 10 is an illustration showing a register data io when the original data stream is received, CPU 4 of a dictionary generated in the third cycle of stores the received original data stream in RAM 1 dictionary generation procedure; (steps P1 and P2 in Fig. 2). Fig. 11 is an illustration showing a working data Assuming that the original data stream is con- stream generated in the third cycle of compres- sisted of N in number of original strings respec- sion procedure; 15 tively having data length of S bits, CPU 4 separates Fig. 12 is an illustration showing a register data the original data stream stored in RAMI at every S of a dictionary generated in the fourth cycle of bits to define respective original strings. Then, a dictionary generation procedure; and non-compression identifier bit "0" is added at the Fig. 13 is an illustration showing a working data leading end of each original string to convert into a stream generated in the fourth cycle of com- 20 working string. Thus, a working data stream con- pression procedure. sisted of N in number of working strings having the The data compression method according to the data length of (S + 1) bits, is formed (step P3 in preferred embodiment of the present invention will Fig. 2). be discussed hereinafter in detail with reference to The shown embodiment of working data stream the accompanying drawings. In the following de- 25 will be discussed with reference to Fig.5. scription, numerous specific details are set forth in The working data stream of Fig. 5 is formed order to provide a thorough understanding of the from the original data stream consisted of 512 in present invention. It will be obvious, however, to number of the original string having the data length those skilled in the art that the present invention of 7 bits. Therefore, in Fig. 5, there is illustrated may be practiced without these specific details. In 30 512 in number of 8 bit working string in a form of other instance, well-known structures are not shown hexadecimal number. in detail in order to unnecessary obscure the Each working string is expressed in a range of present invention. values 00 to 3d in hexadecimal number and has Fig. 1 is a block diagram showing system the most significant bit (MSB) set at "0" as non- construction of a data compression system 100 for 35 compression identifier bit. It should be noted that, implementing data compression. In Fig. 1, the ref- in the shown embodiment, all of the original string erence numeral 1 denotes RAM for storing data of the original data stream have data contents stream, 2 denotes RAM for storing dictionary regis- which can be expressed by 6 bits in binary number ter data, 3 denotes ROM storing a data compres- "000000" to "111101". Therefore, each working sion program, 4 denotes CPU. As can be seen, the 40 string of the working data stream can be expressed data compression system is connected to a host in a range of 00 to 3d. However, even when the system 200. The host system 200 is designed to original string contains data to be expressed by 7 supply an original data stream and a designation bits, no problem will be arisen. for number of cycles to repeat data compression to Next, CPU 1 reads out two sequent working the data compression system, which designation 45 strings (combined string) from the working data will be hereinafter referred to as "repetition number stream stored in RAM 1 in order from the leading command", and to receive compressed data from address. At every occurrence of reading out of the the data compression system. The original data combined string, CPU 1 checks if the read out stream transmitted to the data compression system combined string is the strings read out in the past, 100 from the host system 200 is received by CPU 50 on the basis of register content of RAM 2. If the 4. Then, CPU 4 performs data compression by read out current combined string is new, an occur- software according to a data compression program rence frequency data "1 " and the current com- stored in ROM 3 with utilizing RAMs 1 and 2. It bined string in RAM 2. On the other hand, when should be noted that while RAMs 1 and 2 are the current combined string is the string which is illustrated in Fig. 1, it is, of course, possible to 55 already registered in RAM 2, the occurrence fre- define a data stream storage region and a dic- quency data corresponding thereto is incremented tionary register data storage region in a single by 1 (steps P4 to P10 in Fig. 3). RAM.

6 11 EP 0 678 986 A1 12

Accordingly, when the foregoing procedure is pression identifier bit. With subsequent 7 bits, 128 taken place for overall working data stream stored ( = D: maximum registering number of the dictio- in RAM 1, data content and occurrence frequency nary) of dictionary numbers are expressed. data of all of the combined strings in the working In the shown example, designated repetition data stream are registered with establishing cor- 5 number is set at "4" (R = 4), and thus D/R = 32. respondence. Therefore, all combined string satisfies the fore- Then, CPU 4 checks respective of the regis- going term (1). tered combined strings in terms of the following Next, CPU 14 moves to the compression pro- two conditions: cess for the working data stream utilizing the dic- (1) if the order of the combined string in ques- io tionary register data. tion has the occurrence frequency data ordered In the compression process, at first, the com- at an order higher than D/Rth (wherein D is 2s) bined strings of the working data stream stored in as ordered in descending order from one having RAM 1 are sequentially read out from the leading the largest occurrence frequency data; and address. Then, the read out combined strings are (2) if the occurrence frequency data is greater is compared with respective combined strings regis- than or equal to 3. tered in RAM 2 (steps P12 and P13 in Fig. 4). When the combined string in question satisfies When the matching combined string is present both terms, such combined string is determined to in RAM 2, the corresponding combined string in be maintained, and otherwise the combined string RAM 1 is replaced with the compression dictionary is removed. Then, for each of the remained com- 20 data stored in RAM 2 and corresponding to the bined data, compression dictionary data consisted registered combined string matching with the read of compression identifier bit "1 " and a dictionary out combined string (steps P14 and P15). Namely, number, is assigned and fixedly registered. at this stage, 2x(S + 1) bits of combined string is It should be noted that the foregoing term (1) is replaced with (S + 1) bits of compression dictionary set at D/R as an average value of an upper limit of 25 data. registering number in respective cycles of com- Also, when the foregoing replacement is per- pression procedure under the condition where the formed, CPU 4 shifts the data of the working data maximum registering number of the dictionary is D stream following the combined string which is the and designated number of repetition cycles as dis- objective for replacement in order so that the lead- cussed later is R. This is because that while it 30 ing end of the shifted part of the working data depends on the content of the working data stream stream is located at the address immediately fol- but in most case, number of combined strings lowing the trailing end of the replaced data (step satisfying the term (2) becomes quite large in the P16 in Fig. 4). Namely, since data length is re- initial repetition cycles of the compression proce- duced by compression for (S + 1) bits, the subse- dure to cause overflow of the predetermined regis- 35 quent part of data is shifted for the correspondingly tering capacity of the dictionary to make it impos- to maintain continuity of the working data stream. sible to perform compression procedure for the On the other hand, when the matching com- demanded number of cycles. Therefore, by thin- bined string is not present in RAM 2 (steps P13, ning combined strings having lesser occurrence P14 in Fig. 4), the read out RAM 1 is maintained as frequency, capability of repetition of the compres- 40 is, and process moved to comparison for next sion procedure for the demanded number of cycles combined string. is assured. Accordingly, when the combined string of RAM The necessity of the second term (2) will be 1 is replaced with the compression dictionary data, discussed later in the discussion for the data com- the next combined string is read out by advancing pression procedure. 45 the read out address for (S + 1), and when the Here, the dictionary register data obtained combined string of RAM 1 is not replaced, the next through the foregoing dictionary generating proce- combined string is read out by advancing the read dure with respect to the working data stream of out address for 2x(S + 1) (step P18 in Fig. 4). Fig. 5, is shown in Fig. 6. The foregoing compression process for each In Fig. 6, the data described by hexadecimal 50 individual combined string (steps P13 to P18 in number of 80 to 9a are the compression dictionary Fig. 4) is performed for all of the combined strings data. Data associated with respective compression in the working data stream in RAM 1 (step P19 in dictionary data correspond to data satisfying the Fig. 4). Then, the first cycle of dictionary genera- foregoing terms (1) and (2). In the shown example, tion and compression procedure is completed. 27 combined strings are registered. 55 Here, in the foregoing compression procedure, Since the compression dictionary data is start- since the combined string in RAM 1 as objective ed from hexadecimal number 80, MSB of all data for replacement should satisfy the foregoing term become "1 ". This bit (MSB) serves as the com- (2). Therefore, at least three same combined

7 13 EP 0 678 986 A1 14 strings are present in the working data stream. As can be clear from Fig. 8, as the compres- Therefore, observing the working data stream, with sion dictionary data 9b to 9f are additionally regis- one dictionary register data, at least 3x(S + 1) bits tered corresponding to each of the working strings. data compression can be achieved. The working data stream obtained through data On the other hand, one dictionary register data 5 compression utilizing the additionally registered is consisted of (S + 1) bits of compression dic- dictionary register data is illustrated in Fig. 9. tionary data and 2x(S + 1) of combined string. In the second cycle of dictionary generation Therefore, one dictionary register data occupies and compression procedure, since the total capac- 3x(S + 1) bits. ity of the dictionary register data of Fig. 8 and the Therefore, in view of total data capacity of the io working data stream of Fig. 9 is 362 bites (= 96 dictionary register data and the working data bites + 266 bites), the data compression ratio of stream, the combined string having occurrence fre- about 80.8% can be achieved. quency F = 3 does not contribute for data com- Similarly, in the third cycle of dictionary gen- pression in the first cycle of dictionary generation eration and compression procedure, the dictionary and compression procedure. However, the com- is register data of Fig. 10 and the working data bined strings having occurrence frequency F great- stream of Fig. 11 are obtained. The total capacity er than or equal to 4, data compression for (F-3)x- of the dictionary register data and the working data (S + 1) bits can be realized. stream of Figs. 10 and 11 becomes 353 bites. It should be appreciated that even the com- Therefore, the compression ratio becomes about bined string having the occurrence frequency F = 20 78.8%. Also, in the fourth cycle of dictionary gen- 3, it serves for replacing the 2x(S + 1) of combined eration and compression procedure, the dictionary string with (S + 1) bits of compression dictionary register data of Fig. 12 and the working data data. Therefore, such combined string contributes stream of Fig. 13 are obtained. The total capacity for improvising data compression efficiency in the of the dictionary register data and the working data second and subsequent cycles of dictionary gen- 25 stream of Figs. 12 and 13 becomes 352 bites. eration and compression procedures. Therefore, the compression ratio becomes about When the first cycle of dictionary generation 78.6%. and compression procedure (steps P12 to P19 in When designated number of cycles of dic- Figs. 3 and 4) is performed for the working data tionary generation and compression procedure is stream of Fig. 5 with respective dictionary register 30 completed, CPU 4 reads out all dictionary register data of Fig. 6, a working data stream (hereinafter data in RAM 2 and the compressed working data referred to as once compressed data stream) as stream in RAM 1 and transmits to the host system illustrated in Fig. 7 can be obtained. 200 (step P21 in Fig. 4). In the first cycle of dictionary generation and On the other hand, the transferred compressed compression procedure, the total capacity of the 35 data is decompressed and reproduced in the fol- original data stream is 448 bites. On the other lowing manner. hand, the total capacity of the dictionary register At first, the working string is detected per data of Fig. 6 and the once compressed working (S + 1) from the leading end of the compressed data stream of Fig. 7 is 408 bites (= 81 bites + working data stream. When the MSB of the read 327 bites). Therefore, approximately 91.1% of data 40 out (S + 1) bits is "1", judgement can be made that compression can be obtained. the objective working string has been replaced with Subsequently, with taking the working data the compression dictionary data. Therefore, utiliz- stream obtained through the first cycle of data ing the dictionary register data, the combined compression procedure as new objective for com- string corresponding to the relevant compression pression, the dictionary generation and compres- 45 dictionary data is replaced. On the other hand, sion procedure (steps P4 to P19) illustrated in Figs. when the MSB of the read out (S + 1) bits is "0", 3 and 4 is repeated for the demanded repetition the combined string has been maintained without number R (step P20 in Fig. 4 - step P4 in Fig. 3). being replaced. Therefore, process is moved to the It should be noted that, in the repeatedly per- next combined string. Then, by executing the fore- formed dictionary generation and compression pro- 50 going procedure for entire working data stream, cedure, the working data stream to be taken as the first cycle decompressed working data stream is objective for compression is the result of the imme- obtained. The same procedure is repeated for R diately preceding data compression procedure. times to finally reproduce the working data stream Naturally, in the working data stream, there are before compression. mixedly present the non-compressed working 55 It should be noted that since the non-compres- string and compression register data, as shown in sion identifier bit of "0" is added at MSB of each Fig. 8. working string of the finally obtained working data stream, the original data stream is obtained by

8 15 EP 0 678 986 A1 16 removing MSB of respective working string. permitted. It should be noted that the shown embodiment Also, since the present invention can be re- registers the compression dictionary data consisted alized solely by the software process, it can be of the compression identifier bit "1 " and the dic- realized by simple system constituted of RAM, tionary number, it is possible to store a stream of 5 CPU and so forth. the compression/non-compression identifier bit in In particularly, assuming that the working string separate memory area as required for hardware or is 1 bite and the combined string as a unit for software construction. process in the working data stream and dictionary It should be further noted that while the shown register data is 2 bites, it becomes possible to embodiment relates the dictionary generation and io effectively compress even small data block, such compression procedure for a plurality of times de- as 64 bite which cannot be compressed in the pending upon the designated repetition number R, conventional method. performing dictionary generation and compression Also, while high compression ratio cannot be for only once (R = 1) can be more convenient in expected, data transmission period and period re- certain case. 15 quired for compression or decompression can be For instance, in certain type of data stream shortened. Therefore, it becomes possible to pro- content, one time of compression procedure may vide data compression method which can be adap- achieve relatively high compression ratio, and may tively employed in the case where high speed be required high speed data decompression and image display is required, such as for game ma- reproduction. 20 chine and so forth. Therefore, when CPU 4 detects that designated Although the invention has been illustrated and repetition number R is "1 ", the non-compression described with respect to exemplary embodiment identifier bit of "0" is not added to the original thereof, it should be understood by those skilled in string in the working data stream generating proce- the art that the foregoing and various other dure. In such case, only S bits dictionary number is 25 changes, omissions and additions may be made stored as the compression dictionary data in the therein and thereto, without departing from the spir- dictionary generation procedure of Fig. 3. Then, in it and scope of the present invention. Therefore, the compression procedure of Fig. 4, the working the present invention should not be understood as string is handled as S bits data. limited to the specific embodiment set out above In this case, the total capacity of the finally 30 but to include all possible embodiments which can obtained dictionary register data and the com- be embodies within a scope encompassed and pressed data stream can be reduced. Also, since equivalents thereof with respect to the feature set the compression procedure is performed only out in the appended claims. once, compression and non-compression can be discriminated from the combined string per se reg- 35 Claims istered in RAM 2 upon decompression and reproduction. 1. A data compression method, in which a data Accordingly, data transmission period and de- stream of a digital input signal is stored in first compression and reproduction period can be sig- storage means (1) and data compression of nificantly shortened so as to permit high speed 40 the data stream is performed with registering image display in the case of image data or so forth. strings appearing in said data stream in a It should be noted that since compression pro- dictionary of second storage means (2), char- cedure is performed only once, high compression acterized by comprising: ratio cannot be expected. However, by combining performing first process step for designat- the foregoing method with known algorithm, such 45 ing R representing a repetition number of the as run length method and so forth, this drawback following third to fifth process steps wherein R may be solved. is an integer smaller than or equal to D, with The data compression method according to the taking a total data capacity of the original data present invention constructed as set forth above stream being S x N bits, wherein S is an may achieve the following advantages. 50 integer greater than or equal to 2 and N is an The present invention employs a method for integer greater than or equal to 3, and a maxi- performing data compression by replacing fixed mum dictionary register number D where D is length of sequentially or discretely presenting com- 2s; bined string into a fixed length of string with half performing second process step for sepa- length with adding the compression/non-compres- 55 rating the original data stream of S bit length sion identifier bit. Therefore, by simplicity of al- stored in said storage means and generating a gorithm of compression and decompression, high working data stream by adding a non-conduct- speed compression and decompression process is ing identifier bit for conversion into a working

9 17 EP 0 678 986 A1 18 data stream having (S + 1) bit working strings; storage means, and for incrementing said oc- performing third process step for detecting currence data at second and subsequent oc- combined string consisted of two sequential currences; working strings in said working data stream up third means for generating a dictionary by to D/Rth order in descending order from work- 5 registering said combined data strings which ing string having largest occurrence frequency, has occurrence data satisfying a predeter- and having occurrence frequency greater than mined register condition, together with dictio- or equal to 3; nary identifiers thereof, in said second storage performing fourth process step for register- means; ing compression dictionary data of (S + 1) bits io fourth means for replacing combined data consisted of dictionary number and compres- strings in said data stream with the corre- sion identifier bit, in said second storage sponding dictionary identifiers; and means, corresponding to each of combined fifth means for controlling operations of strings detected by said third process step; said first to fourth means for repeating data and is compression cycles for the desired repetition firth process step for replacing combined number indicated in said repetition command. string among combined strings in said working data stream matching with one of combined 4. A data compression system as set forth in strings registered in said second storage claim 3, characterized in that said first means means, with the compression dictionary data 20 adds a compression state identifier bit for each corresponding to the matching combined fixed length of data fraction in said data stream string, to formulate said data string. repeating said third to fifth process steps for R times with taking data stream replaced 5. A data compression system as set forth in through said fifth process step as working data 25 claim 4, characterized in that said third means stream for outputting data stream stored in registers said dictionary identifier incorporating said first storage means and all combined said compression status identifier indicative of string and compression dictionary data stored compressed state. in said second storage means, after R times repletion, as compressed data. 30 6. A data compression system as set forth in claim 3, characterized in that said second A data compression method as set forth in means forms said combined data string with claim 1, characterized in that when R = 1, two sequential data strings of a fixed unit data said second process step is performed without length, and said third means generates said adding said non-compression identifier bit to 35 dictionary identifier having a data length cor- establish said working data stream take the responding to said fixed unit data length. original strings as said working strings, and said fourth process step is performed with 7. A data compression system characterized by formulating said compression dictionary data comprising: solely with said S bit dictionary number. 40 first storage means for storing data; second storage means for storing data; A data compression system characterized by first means for receiving an input sequen- comprising: tial data signal and a repetition command sig- first storage means for storing data; nal indicative of a desired repetition number of second storage means for storing data; 45 data compression process cycles and sequen- first means for receiving an ordinal data tially extracting fraction signals of a given fixed stream and a repetition command indicative of bit length from said input sequential data sig- desired repetition number of data compression nal; process cycles and separating said original second means for sequentially extracting data stream into a plurality of sequential data 50 segment signals having a bit length plurality strings having a fixed unit data length; times of said given fixed bit length of said second means for comparing combined fraction signals from input sequential data sig- data strings established by given number of nal, comparing the content of each of said sequential data strings with a content of said segment signal with a content of said first first storage means, for registering occurrence 55 storage means, for registering an occurrence data and the combined data string at the first counter signal value and said segment signal occurrence of a coincidence between the com- in said second storage means at the first oc- bined data and the content in said second currence of a coincidence between the con-

10 19 EP 0 678 986 A1 20

tents of each of said segment signal and the and the content and for incrementing said oc- contents of said first storage means, and for currence data at second and subsequent oc- incrementing said occurrence counter signal currences; value at second and subsequent occurrences; generating a dictionary by registering said third means for generating a dictionary by 5 combined data strings which has occurrence registering, in said second storage means, said data satisfying a predetermined register con- segment signals corresponding to said occur- dition, together with dictionary identifiers there- rence counter signal satisfying a predeter- of, in said second storage; mined register condition, together with dictio- replacing combined data strings in said nary identifier signals thereof; 10 data stream with the corresponding dictionary fourth means for replacing data segment in identifiers; and said input sequential data signal corresponding controlling operations of said first to fourth to said segment signals registered in said sec- means for repeating data compression cycles ond storage means with the corresponding re- for the desired repetition number indicated in spectively corresponding dictionary identifier is said repetition command. signals; and fifth means for generating control signals for controlling operations of said first to fourth means for repeating data compression cycles for the desired repetition number indicated in 20 said repetition command.

8. A data compression system as set forth in claim 7, characterized in that said first means adds a compression state identifier bit for each 25 fixed length of data fraction in said input sequential data signal to formulate said fraction signal.

9. A data compression system as set forth in 30 claim 8, characterized in that said third means registers said dictionary identifier signal incorporating said compression status identifier bit indicative of compressed state. 35 10. A data compression system as set forth in claim 7, characterized in that said second means forms said segment signal with two sequential fraction signals, and said third means generates said dictionary identifier sig- 40 nal having a bit length corresponding to said given fixed bit length.

11. An operation method of compression system including first and second storage, character- 45 ized by comprising the steps of: receiving an ordinal data stream and a repetition command indicative of desired repetition number of data compression process cycles and separating said original data stream so into a plurality of sequential data strings having fixed unit data length; comparing combined data strings established by a given number of sequential data strings with a content of said first storage for 55 registering occurrence data and the combined data string at the first occurrence of a coincidence between the combined data strings

11 EP 0 678 986 A1

FIG. 1

100

RAM _ (ORIGINAL DATA (DATA STREAM) STREAM) —(NUMBER OF CYCLES TO- REPEAT DATA COMPRESSION) CPU

RAM COMPRESSED (DICTIONARY DATA REGISTER DATA)

ROM (DATA COMPRESSION PROGRAM)

FIG. 2

( START )

INPUT OF REPETITION P1 (Ex R = 4) COMMAND NUMBER R :

STORING THE ORIGINAL DATA P2 STREAM IN RAM 1

ORIGINAL DATA STREAM = S BITSxn (Ex : 7 X 512 = 448 BYTES) SEPARATING THE ORIGINAL DATA STREAM AT EVERY S BITS TO DEFINE RESPECTIVE ORIGINAL STRINGS AND ADDING A NON-COMPRESSION IDENTIFIER BIT " 0" AT THE LEADING END OF EACH ORIGINAL STRING TO CONVERT INTO A WORKING STRING

WORKING DATA STREAM = (S + 1) BITSxn (Ex : 8x512 = 512 BYTES)

12 EP 0 678 986 A1

FIG. 3

P4 i = 0

P5 READING OUT TWO SEQUENT WORKING STRINGS (COMBINDED STRINGS) STORED IN RAM 1

P8 REGISTRATION OF AN INCREMENT OF OCCURRENCE P10 OCCURRENCE FREQUENCY FREQUENCY DATA i^i + S + 1 DATA " 1" AND THE CORRESPONDING THERETO CURRENT COMBINED BY 1 STRING IN RAM2

P9 — " - IS THE N REGISTRATION COMPLf TED FOR OVERALL WORKING DATA STREAM ?

CHECKING THE REGISTERED COMBINED STRINGS AND FIXEDLY REGISTERING THE SELECTED COMBINED STRINGS WITH A COMPRESSION DICTIONARY DATA

COMPRESSION DICTIONARY DATA ^COMPRESSION IDENTIFIER BIT "1" + DICTIONARY NUMBER

13 EP 0 678 986 A1

FIG. 4

P13 READING OUT TWO WORKING STRINGS (COMBINED STRINGS) STORED IN RAM 1 AND COMPARING THEM WITH COMBINED STRINGS REGISTERED IN RAM 2

P14 N ^ IS THE MATCHING ^ COMBINED STRING PRESENT

P15 REPLACING THE CORRESPONDING COMBINED STRING IN RAM 1 WITH THE COMPRESSION DICTIONARY DATA STORED IN RAM 2

P16 SHIFTING THE DATA OF THE WORKING DATA STREAM IN RAM 1 SO THAT THE LEADING END OF THE WORKING DATA STREAM IS LOCATED AT THE ADDRESS OF A (i + S + 1)

P18 P17 i + 2S + 2 i + S+1

( END )

14 EP 0 678 986 A1

FIG. 5

0000000000000000000000000000 36 39 oooooooooooooooooooooooooooo3c39 0000000000000000000000000038 3d38 oooooooooooooooooooooooooo 33 3c3b OOOOOOOOOOOOOOOOOOOOOO 00 00 36 3c 36 00000000000000 00 000000000038 37 33 00000000000000000000000000 36 37 33 (X)(X)0O(X)0O0O0O0O0O0O0O0O0c 33 36 33 00(X)0000(X)(X)000000000000 15 15 0c32 000000000000000000000000 0c 08 05 05 00000000000000000000000d1a09 1a17 00 00 00 00 00 00 00 00 00 00 00 Od 1a 1a Ob 15 00 00 CO 00 00 00 00 00 00 00 00 Oc 1a 17 17 15 00 00 00 00 00 00 00 00 00 00 00 1a Oc Ob Ob 20 00 00 00 00 00 00 00 00 00 00 00 Oc 17 20 Oa Oa 00 00 00 00 00 00 00 00 00 00 17 Oc Ob Ob 20 Oa 39 35 34 31 31 35 31 31 32 32 Ob Oa Oa 20 20 07 35 35 33 31 2f 31 35 34 2f 31 Oa 05 25 Oa Oa Oa 3b 36 32 32 31 2f 37 30 31 2e 05 04 04 25 25 05 35 17 17 15 32 31 36 35 2f 2f Ob 04 04 04 04 04 Od Ob Ob 20 17 33 33 35 32 30 Ob 04 04 04 04 04 Oc Ob Oa 20 17 36 32 30 35 30 Ob 04 25 25 05 Oa 17 17 20 17 33 33 32 30 35 34 27 Ob 05 20 20 14 32 Oc 17 32 36 32 30 32 33 27 14 20 20 14 27 35 33 32 32 36 34 30 30 33 30 32 00 00 Od 35 32 34 Oa 17 Od 34 30 32 30 30 31 31 34 35 34 28 14 20 15 Oa Oa 04 32 35 30 31 30 35 34 00 00 00 00 20 07 Ob 20 Oa 04 Oa Ob 17 00 00 00 00 00 00 00 00 0b0b0c17 0b 15 00000000000000000000 Oa Oa Ob Oc 17 00 00 00 00 00 00 00 00 00 00 00 20 0b0a 00000000000000000000000000 0b 17 0000000000000000000000000000

I5 EP 0 678 986 A1

FIG. 6

FIG. 7

80 80 80 80 80 80 80 36 39 80 80 80 80 80 80 80 3c 39 80 80 80 80 80 80 00 38 3d 38 80 80 80 80 80 80 00 33 3c 3b 80 80 80 80 80 80 83 3c 36 80 80 80 80 80 80 00 38 37 98 80 80 80 80 80 80 36 37 98 80 80 80 80 80 81 33 36 98 80 80 80 80 80 00 15 15 Oc 32 80 80 80 80 80 80 Oc 08 05 05 80 80 80 80 80 82 1a 09 1a 8d 80 80 80 80 80 Od 1a 1a Ob 8c 80 80 80 80 80 Oc 1a 8e 8c 80 80 80 80 80 1a 8a 89 80 80 80 80 80 81 17 8f Oa 80 80 80 80 80 17 8a 89 Oa 39 99 95 35 95 97 87 Oa 91 07 35 35 33 31 2f 31 99 2f 31 Oa 05 25 85 Oa 3b 9a 32 31 2f 37 92 2e 05 84 25 25 05 35 8e 15 32 31 36 35 2f 2f 86 84 84 Od 88 90 33 33 35 96 86 84 84 8a Oa 90 9a 94 30 86 25 25 05 Oa 8e 90 33 33 96 99 27 Ob 05 91 14 32 8b 32 9a 93 33 27 14 91 14 27 35 33 97 36 34 30 30 33 93 80 Od 35 32 34 Oa 17 Od 34 93 30 92 31 34 99 28 14 20 15 85 04 32 35 92 94 34 80 80 20 07 89 Oa 04 Oa Ob 8d 80 80 80 00 88 8b Ob 8c 80 80 80 80 00 85 Ob 8b 80 80 80 80 80 00 20 87 80 80 80 80 80 80 00 Ob 8d 80 80 80 80 80 80 00

16 EP 0 678 986 A1

FIG. 8

80 81 82 83 84 85 86 87 0000 000c OOOd 0036 0404 OaOa Ob04 ObOa 88 89 8a 8b 8c 8d 8e 8f ObOb 0b20 OcOb 0c17 1500 1700 1717 200a 90 91 92 93 94 95 96 97 2017 2020 3031 3032 3035 3131 3230 3232 98 99 9a 9b 9c 9d 9e 9f 3300 3534 3632 8000 8080 8c80 8d80 9880

FIG. 9

9c 9c 9c 80 36 39 9c 9c 9c 80 3c 39 9c 9c 9c 00 38 3d 38 9c 9c 9c 00 33 3c 3b 9c 9c 9c 83 3c 36 9c 9c 9c 00 38 37 9f 9c 9c 80 36 37 9f 9c 9c 81 33 36 9f 9c 9c 00 15 15 Oc 32 9c 9c 9c Oc 08 05 05 9c 9c 80 82 1a 09 1a 9e 9c 9c Od 1a 1a Ob 9d 9c 9c Oc 1a 8e 9d 9c 9c 1a 8a 89 9c 9c 80 81 17 8f Oa 9c 9c 80 17 8a 89 Oa 39 99 95 35 95 97 87 Oa 91 07 35 35 33 31 2f 31 99 2f 31 Oa 05 25 85 Oa 3b 9a 32 31 2f 37 92 2e 05 84 25 25 05 35 8e 15 32 31 36 35 2f 2f 86 84 84 Od 88 90 33 33 35 96 86 84 84 8a Oa 90 9a 94 30 86 25 25 05 Oa 8e 90 33 33 96 99 27 Ob 05 91 14 32 8b 32 9a 93 33 27 14 91 14 27 35 33 97 36 34 30 30 33 93 80 Od 35 32 34 Oa 17 Od 34 93 30 92 31 34 99 28 14 20 15 85 04 32 35 92 94 34 9c 20 07 89 Oa 04 Oa Ob 9e 9c 00 88 8b Ob 9d 9c 9b 85 Ob 8b 9c 9c 9b 20 87 9c 9c 9c 00 Ob 9e 9c 9c 9b

17 EP 0 678 986 A1

FIG. 10

80 81 82 83 84 85 86 87 0000 000c OOOd 0036 0404 OaOa 0b04 ObOa 88 89 8a 8b 8c 8d 8e 8f ObOb 0b20 OcOb Oc17 1500 1700 1717 200a 90 91 92 93 94 95 96 97 2017 2020 3031 3032 3035 3131 3230 3232 98 99 9a 9b 9c 9d 9e 9f 3300 3534 3632 8000 8080 8c80 8d80 9880 aO a1 a2 a3 a4 a5 a6 9c00 9c80 9c9b 9c9c 9d9c 9e9c 9f9c

FIG. 11

a3 a1 36 39 a3 a1 3c 39 a3 aO 38 3d 38 a3 aO 33 3c 3b a3 9c 83 3c 36 a3 aO 38 37 a6 a1 36 37 a6 9c 81 33 36 a6 aO 15 15 Oc 32 a3 9c Oc 08 05 05 a3 80 82 1a 09 1a a5 9c Od 1a 1a Ob a4 9c Oc 1a 8e a4 9c 1a 8a 89 a3 80 81 17 8f Oa a3 80 17 8a 89 Oa 39 99 95 35 95 97 87 Oa 91 07 35 35 33 31 2f 31 99 2f 31 Oa 05 25 85 Oa 3b 9a 32 31 2f 37 92 2e 05 84 25 25 05 35 8e 15 32 31 36 35 2f 2f 86 84 84 Od 88 90 33 33 35 96 86 84 84 8a Oa 90 9a 94 30 86 25 25 05 Oa 8e 90 33 33 96 99 27 Ob 05 91 14 32 8b 32 9a 93 33 27 14 91 14 27 35 33 97 36 34 30 30 33 93 80 Od 35 32 34 Oa 17 Od 34 93 30 92 31 34 99 28 14 20 15 85 04 32 35 92 94 34 9c 20 07 89 Oa 04 Oa Ob a5 00 88 8b Ob a4 9b 85 Ob 8b a3 9b 20 87 a3 aO Ob a5 a2

18 EP 0 678 986 A1

FIG. 12

80 81 82 83 84 85 86 87 0000 000c OOOd 0036 0404 OaOa 0b04 ObOa 88 89 8a 8b 8c 8d 8e 8f ObOb 0b20 OcOb 0c17 1500 1700 1717 200a 90 91 92 93 94 95 96 97 2017 2020 3031 3032 3035 3131 3230 3232 98 99 9a 9b 9c 9d 9e 9f 3300 3534 3632 8000 8080 8c80 8d80 9880 aO a1 a2 a3 a4 a5 a6 a7 9c00 9c80 9c9b 9c9c 9d9c 9e9c 9f9c a380 a8 a3a0

FIG. 13

a3 a1 36 39 a3 a1 3c 39 a8 38 3d 38 a8 33 3c 3b a3 9c 83 3c 36 a8 38 37 a6 a1 36 37 a6 9c 81 33 36 a6 aO 15 15 Oc 32 a3 9c Oc 08 05 05 a7 82 1a 09 1a a5 9c Od 1a 1a Ob a4 9c Oc 1a 8e a4 9c 1a 8a 89 a7 81 17 8f Oa a7 17 8a 89 Oa 39 99 95 35 95 97 87 Oa 91 07 35 35 33 31 2f 31 99 2f 31 Oa 05 25 85 Oa 3b 9a 32 31 2f 37 92 2e 05 84 25 25 05 35 8e 15 32 31 36 35 2f 2f 86 84 84 Od 88 90 33 33 35 96 86 84 84 8a Oa 90 9a 94 30 86 25 25 05 Oa 8e 90 33 33 96 99 27 Ob 05 91 14 32 8b 32 9a 93 33 27 14 91 14 27 35 33 97 36 34 30 30 33 93 80 Od 35 32 34 Oa 17 Od 34 93 30 92 31 34 99 28 14 20 15 85 04 32 35 92 94 34 9c 20 07 89 Oa 04 Oa Ob a5 00 88 8b Ob a4 9b 85 Ob 8b a3 9b 20 87 a8 Ob a5 a2

19 Patent European EUROPEAN SEARCH REPORT APPliCMl0n ""^ OITicc

DOCUMENTS CONSIDERED TO BE RELEVANT EP 95106020.1 r„,.„._, Citation of document with indication, where appropriate, Relevant CLASSIFICATION OF THE t-"e80ry of relevant passages to daim APPLICATION Out. CI. 6)

A EP - A - 0 573 208 1.3.7, H 03 M 7/42 ( HEWLETT- PCKARD) 11 G 06 F 5/00 * Totality *

D,A US - A - 4 558 302 1,3.7, (WELCH) 11 * Abstract *

D , A US - A - 4 814 746 1,3,7, (MILLER et al. ) 11 * Totality *

A US - A - 4 876 541 1,3,7, (STORER) 11 * Abstract; claims 1,17,18 *

D,A US - A - 5 003 307 1,3,7, (WHITING et al. ) 11 * Abstract *

D A US-A-5016 009 1,3,7, technical fields , searched ao«. a. 6), (WHITING et al.) 11 . * Abstract; claims 1,13 * H 03 M 7/00 A US - A - 5 049 881 1,3.7, (GIBSON et al. ) 11 * Claims 1,9,10-14 *

A US - A - 5 140 321 1,3.7. (JUNG) 11 * Column 3, line 49 - column 4, line 34 *

A US - A - 5 142 282 1.3.7. (TOBIN et al. ) 11 * Totality *

A US - A - 5 150 119 1.3.7. (YOSHIDA et al. ) 11 * Claims *

The present search report has been drawn up for all claims Place of search Date of completion of the tearch Examiner VIENNA 09-06-1995 BAUMANN CATEGORY OF CITED DOCUMENTS T : theory or principle underlying the invention E : earlier patent document, but published on, or X : particularly relevant if taken alone after the filing date Y : particularly relevant if combined with another D : document cited in the application document of the same category L : document cited for other reasons O : non-written disclosure & : member of the same patent family, corresponding P : intermediate document document Patent Application Number 3 European EUROPEAN SEARCH REPORT Office

-2- DOCUMENTS CONSIDERED TO BE RELEVANT EP 95106020.1 Citation of document with indication, where appropriate, Relevant CLASSIFICATION OF THE Category of relevant passages to claim APPLICATION (Int CI 6)

US 5 175 543 1,3,7, (LANTZ) 11 * Claims *

US - A - 5 243 341 1,3,7, (SEROUSSI et al. ) 11 * Totality *

US 5 254 990 1,3,7, (YOSHIDA et al. ) 11 * Claims; abstract *

US 5 253 325 1,3,7, (CLARK) 11 * Totality *

TECHNICAL FIELDS SEARCHED (Int. CL 6)

The present search report has been drawn up for all claims Place of search Dale of completion of (he March Lxuntner VIENNA 09-06-1995 BAUMANN

CATEGORY OF CITED DOCUMENTS T : theory or principle underlying the invention E : earlier patent document, but published on, or X : particularly relevant if taken alone after the filing date Y : particularly relevant if combined with another D : document cited in the application document of the same category L : document cited for other reasons A : technological background O : non-written disclosure & : member of the same patent family, corresponding P : intermediate document document