WO1997000558A1 - Apparatus and method for lossless waveform data compression - Google Patents
Apparatus and method for lossless waveform data compression Download PDFInfo
- Publication number
- WO1997000558A1 WO1997000558A1 PCT/US1996/007740 US9607740W WO9700558A1 WO 1997000558 A1 WO1997000558 A1 WO 1997000558A1 US 9607740 W US9607740 W US 9607740W WO 9700558 A1 WO9700558 A1 WO 9700558A1
- Authority
- WO
- WIPO (PCT)
- Prior art keywords
- data
- vector
- compressed
- difference
- processor
- Prior art date
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T9/00—Image coding
- G06T9/005—Statistical coding, e.g. Huffman, run length coding
-
- H—ELECTRICITY
- H03—ELECTRONIC CIRCUITRY
- H03M—CODING; DECODING; CODE CONVERSION IN GENERAL
- H03M7/00—Conversion of a code where information is represented by a given sequence or number of digits to a code where the same, similar or subset of information is represented by a different sequence or number of digits
- H03M7/30—Compression; Expansion; Suppression of unnecessary data, e.g. redundancy reduction
- H03M7/3053—Block-companding PCM systems
Definitions
- the field of the invention relates to data compression and more particularly to apparatus and method of lossless data compression particularly suitable for compression of EEG waveforms.
- Lossless data compression is known. Lossless data compression may be used where ever a data file must be passed through some sort of process where the process itself limits the volume of the data yet still requires that the data be accurate. Examples of such a process include data storage and retrieval or data transmission systems.
- the object is to be able to store large amounts of data in a memory device that would otherwise be too small for the data.
- a processor associated with the memory device first compresses the data under an appropriate compression process for storage within the memory device. The same or another processor may decompress the data using a similar process.
- the data inputs to a voice channel may include a real-time voice signal and control information necessary to maintain the voice channel.
- the communication system may require a certain information transfer rate for each voice signal on each channel to maintain acceptable voice quality in the presence of a variable rate of control information.
- Data compression e.g., Huffman coding
- Huffman coding may be used in such a case to pack the control information and voice signal into an allocated spectrum without loss of voice quality.
- the process is a data transmission system and includes data and control information, exclusively such as for interconnections between computers, then data compression may become even more important. In such cases data compression becomes important not only to reduce the cost of maintaining such connections but also to optimize the value and utility of such connection between geographically diverse data processing and storage facilities.
- Data comprised for purposes of communication may also be stored in the compressed state and vice versa. Improved communication efficiency and storage makes remote storage and retrieval practical at geographically diverse research or manufacturing facilities.
- Data compression has also had an impact on small portable data processing devices.
- data processors e.g., laptop, notebook, or palmtop computers
- data processors have become smaller and more portable and product features more dependent upon stored software, it often becomes necessary to more efficiently use the higher capacity hard disk drives and floppy disks that have also become available.
- Lossless data compression is generally implemented using one of two types of modeling.
- the first of the two types of modeling is referred to as statistical modeling and the other is referred to as dictionary based modeling.
- Statistical modeling is a compression technique based upon statistical probability of use of a particular symbol in any given time period. The more likely that a particular symbol may be used results in fewer bits being used to describe the symbol.
- Statistical modeling is typically based upon use of a look-up table (static model) of compression codes.
- Compression codes are often modeled to the type of data that is to be compressed and is often adapted during use.
- each symbol is received by a compressing processor, the symbol is compared with the look-up table and the compression code retrieved and substituted into an output stream in place of the received symbol.
- the compression code may then be transmitted or stored depending on the process for which compression is being used.
- Decompression of data under statistical modeling is also based upon a look-up table.
- a compression code is received (i.e., from memory or from a communication receiver) the code is compared with a look-up table. A corresponding symbol is identified from the look-up table and used as an output of the decompression process.
- Dictionary based compression schemes are also based upon look-up tables.
- Dictionary based compression schemes look for groups of symbols within the look-up table. As input symbols are read, a compressing processor looks for groups of symbols that appear within the dictionary. If a match for the symbol-string is found, a pointer or index identifying the string is output. The longer the match, the better the compression ratio.
- Recent improvements in dictionary compression schemes rely on an adaptive dictionary. Under LZ77 currently received data is compared with a dictionary of previously transmitted data. Where symbol-string matches are found, the matches are encoded as dictionary pointers into the output stream.
- Prior compression techniques i.e., statistical and dictionary compression methods
- Prior compression techniques are based upon an assumption that the transmitted data has repetitive elements that may be used in compressing the transmitted signal based upon an appropriate algorithm for identifying the repetitive elements. Where the transmitted data does not contain repetitive elements then the prior art compression techniques break down.
- the apparatus includes a memory for storing the compressed data vector and a first comparator for determining a largest and smallest data point of the data vector.
- the apparatus further includes a second comparator for comparing the largest and smallest data point and when they are equal, causing a first data point of the data vector to be stored in the memory as the compressed data vector.
- the invention further includes novel method and apparatus for compressing a data vector of a predetermined number of data points under an alternate format.
- the apparatus includes a memory for storing the compressed data vector and a first processor for determining a data field width necessary to uniquely describe a largest relative magnitude data point.
- the apparatus also includes a second processor for storing the data points of the data vector as the compressed vector in data fields of memory of the data field width.
- the invention further includes a novel apparatus for compressing a data vector of a predetermined number of data points under a third alternate format.
- the apparatus includes a memory for storing the compressed data vector and a comparator and register for creating a difference vector of difference values of adjacent data points of the data vector.
- a first processor is also included for determining a data field width necessary to uniquely describe a largest magnitude difference value of the difference vector.
- a second processor is provided for storing the difference points of the difference vector as the compressed vector in data fields of the memory of the determined second field width along with the first data point of the data vector.
- FIG. 1 is a block diagram of an apparatus for performing data compression in accordance with an embodiment of the invention
- FIG. 2 is a flow chart depicting the steps of data compression under a first format using the apparatus of FIG. 1;
- FIG. 3 provides further detail of the apparatus of FIG.1 in accomplishing the process steps of FIG. 2;
- FIG. 4 is a flow chart depicting the steps of data compression under a second format using the apparatus of FIG.l;
- FIG. 5 provides further detail of the apparatus of FIG.1 in accomplishing the process steps of FIG. 4;
- FIG. 6 is a flow chart depicting the steps of data compression under a third format using the apparatus of FIG.1;
- FIG. 7 provides further detail of the apparatus of FIG.1 in accomplishing the process steps of FIG. 6;
- FIG. 8 provides further detail for the apparatus of FIG.1 in accomplishing data decompression
- FIG. 9 provides further detail of channel inputs of the apparatus of FIG. 1 when used for data compression of data from a human subject.
- Appendix I provides pseudo-code of data compression using each of three possible compression formats. Description of a Preferred Embodiment
- FIG. 1 is a block diagram of a data compression system 10, generally, in accordance with an embodiment of the invention. Included within the data compression system 10 is a memory 12, a controller 14, an analog to digital (A/D) converter 16, and an optional post processor 18.
- the memory 12 may be any appropriate memory mass storage device (e.g., random access memory (RAM), floppy disk drive, a hard disk drive, etc.).
- the controller 14 may be any processor operating under any suitable operating system, for example, DOS or UNIX.
- the A/D converter 16 may be any 32 channel device (e.g., 4 Linear Tech model LTC 1290 connected in parallel) having an appropriate resolution (e.g., 12 bits) and an output buffer (not shown) addressable by the processor 14.
- the data compression system 10 may be used for biomedical (e.g., EEG, EKG, etc.) monitoring.
- the data compression system 10 may also be used for any other data activity consistent with the input capabilities of the A/D converter 16 (e.g., seismographic logging, etc.).
- an appropriate buffer amplifier 15 e.g., Analog Devices model AD621
- An amplifier 15 in such case is interconnected between the A/D converter 16 and human subject for each input channel (channels #1-n, FIG. 9) used for EEG monitoring.
- the controller 14 periodically reads the outputs of the A/D converter 16 at a fixed time interval consistent with the conversion time of the A/D converter 16.
- the readings of each channel of the 32 channels are stored in a separate buffer area within RAM 20 of the controller 14 for subsequent compression and storage within the memory 12.
- the controller 14 accumulates readings (data points) from each channel of the A/D converter 16 in a separate channel buffer for a predetermined time period (epoch). Each data point is assumed to be a signed, 2s complement integer, with the least significant bit of the point residing in the least significant bit of the word used to store it. The number of samples in the epoch must be a multiple of the number of bits in a word used to store a sample (e.g., 16). During the epoch the controller 16 may read each channel output of the A/D converter 16 a number of times (e.g., 256) and store the sequentially collected data points in adjacent positions within the channel buffer for that channel. The sequentially collected data points together form a data vector. For purposes of simplicity, the compression process will be described in terms of a single channel. It is to be understood that the compression process may be applied to any number of channels (i.e., 1 to 32) for any number of A/D converters 16.
- compression may occur under a number of different formats.
- an input data vector temporarily stored in a channel buffer 20a in RAM 20, is converted into a compressed data vector which may then be stored in the mass storage device (i.e., memory 12) or passed to the optional post processor 18.
- Post processor 18 may be a data manipulation operation or some other volume limited device requiring a compressed stream of input data (e.g., a communication channel).
- the data vector may contain a number of data points that are identical. Where the data points of the vector are all equal, a single data point may be stored as the compressed data vector along with indicia of format.
- FIG. 2 is a flow chart of data compression under the first format. Under the first format, a data vector of the epoch is read 100 from the input buffer 20a of RAM 20 to determine 102 a range of values of the epoch.
- the controller 14 stores one of the values (e.g., the first value) of the data vector, along with the indicia of format (e.g., the number "0").
- FIG. 3 is a block diagram of a specific embodiment of apparatus used for data compression under the first format.
- a first comparator 24 reads data points of the data vector one at a time while comparing the magnitudes of the data points with the contents of register 22 holding the largest data point and the contents of register 26 holding the smallest data point.
- a second comparator 28 tests whether the smallest point is equal to the largest point. If they are equal, one of the data points is stored with the indicia of the first format.
- Table 1 shows an example of a data vector having sixteen data points (each equal to a value of 32) .
- the comparator 24 reads the data points from the buffer 20a of RAM 20 one at a time. Upon reading the first data point 32, the comparator 24 stores the magnitude of the first data point in the largest data point register 22 and the smallest data point in register 26. The comparator 24 then proceeds to read and compare each subsequent data point of the data register 20a with the largest and smallest data point registers 22, 26. If a subsequent data point is larger than the value stored in the largest data point register 22, then the larger value is substituted into the larger data point register 22. Likewise, if a smaller data point is found that is smaller than the value in the smallest data point register 26, then that value is substituted into the smaller data point register 26.
- Appendix I contains a summary of data compression under the invention. Reference will be made to Appendix I as appropriate to an understanding of the invention.
- Table II and FIGS. 4 and 5 will be used to describe data compression under the second format along with the decision points of how the second format is selected over the third format.
- Table III and FIGS. 6 and 7 will be used to describe data compression under the third format along with how the third format is selected over the second format.
- the 16 input data points of the data vector range from 12 to -1. Since the largest point (12) is not equal to the smallest point (-1), the first format cannot be used.
- a difference vector is determined for adjacent data points.
- the difference vector is determined by subtracting an adjacent data point on the left from a data point on the right.
- a maximum and minimum value of the data vector were already utilized to determine that the first format could not be used. From these determinations, the number of bits necessary to describe each data point is determined from the magnitude of each. A minimum and maximum value of the difference vector are also determined, and from these the number of bits necessary to describe each difference value. In the example shown in Table II the number of data bits is 5 (with a fifth bit provided for sign). The number of difference bits is also 5. Since the bits of the data vector is the same as the bits of the difference vector the system selects the second format for compressing the data vector.
- FIG. 5 is a simplified block diagram depicting apparatus for compressing data under the second format.
- data points of the data vector are retrieved from the buffer 20a of RAM 20 one at a time for determining a proper format.
- the processor 30 stores the data point in a data point register of buffer 20c.
- the processor 30 next retrieves the next data point. if the second data point is larger than the first data point, then the data point stored in the data point register in buffer 20c is replaced. The processor 30 also subtracts the first data point from the second data point and stores the difference as a difference point in a first storage location of a difference vector buffer of buffer 20c and also in a largest difference point data register 20c.
- a third data point is retrieved from the data vector stored in buffer 20a. Again, if the third data point is larger than that stored in the data point register in the buffer 20c the third data point is stored in the data point register.
- the second data point is subtracted from the third data point.
- the difference is stored as the second difference point in the difference vector stored in the buffer 20c.
- the second difference point is compared with the value in the largest difference point and, if larger, is stored in the largest difference point register in buffer 20c.
- the process repeats for data points 4-16 each time determining a maximum value for the data vector and the difference vector.
- the processor 30 determines a data bit size for the largest point of the data vector and a difference bit size for the largest difference point through use of a first series of inequalities (e.g.. Appendix I, Algorithms 3 and 4) or by reference to a look-up table.
- the processor 30 compares the data bit size with the difference bit size. If the data bit size is less than, or equal to, the difference bit size, then the second format is chosen for compression, otherwise the third format is chosen.
- the data bit size is 5 and the difference bit size is also 5. Since the data bit size is the same as the difference bit size, the second format of compression is chosen.
- the processor 32 stores 116 an indicia of format in a first word of the compressed data vector.
- the indicia of the second format may be a positive numerical value equal to the data bit size. In the case of the example of Table II, a value of 5 is stored in the first word of the compressed data vector.
- the steps used to compress the data points of the data vector into the compressed data vector are dependent upon the data bit size.
- Appendix I, Algorithm 5 indicates compressed data point position within the compressed vector as well as compressed data vector size for a number of bit sizes. For example, if the data bit size were one bit, then the first data point would occupy position "0" on the far left side of the compressed vector as indicated in Appendix I, Algorithm 5. Likewise, the information of the second data point would occupy the second bit position (labeled "1"). The remaining data points of the data vector would successively occupy other positions to the right up to the final positions (i.e., "F"), which is occupied by the information of the sixteenth data point.
- bit size is shown to be 5 bits. From Appendix I, Algorithm 5, a bit size of 5 is shown to require 5 data words for storage of the data points of the data vector in a compressed vector. Since the significant bits of a number of data points must be packed 112 into each word of the compressed data vector, a number of bit shifting operations are required for creation of the compressed data vector.
- a processor 30 (as noted above) first reads 100 and determines 110 the bit size (i.e., 5) of the largest data point of the data vector. After determining that the second format of data compression is to be used, the processor 30 may store the indicia of format (i.e., 5) directly into a first word of an allocated compressed data vector location in memory 12 or may piece together the entire compressed data vector within a compressed data vector register of the buffer 20c, first, and later store the entire compressed data vector into memory 12.
- the processor 30 determines a data bit size and allocates a field length (i.e., several words of memory) to the completed compressed data vector based upon the determined data bit size.
- the processor then stores the indicia of format and compressed data into the allocated field length.
- the processor 32 loads a first data point and masks off the 11 bits to the left of the bits designated by the data bit size.
- the processor 32 then shifts the remaining (unmasked) bits to the far left (i.e., by 11 bits, to the positions labeled "0" in Appendix I, Algorithm 5) and stores the result in a first word of a data buffer 35 of the processor 32.
- the processor 32 then loads a second data point, masks off the 11 bits to left and shifts the remaining bits 6 bits to the left before storing in the buffer 35.
- the third data point is masked and shifted one bit to the left before storing in the buffer 35.
- the fourth data point is loaded, masked and shifted four bits to the right, resulting in three bits (i.e., the right three data bits of the fourth data point) being shifted out of the shift register.
- the remaining bit of the fourth masked data point is then stored in the buffer 35 in the right-most data position.
- the masking and shifting described to this point has resulting in the completion of a second word of the compressed data vector (the first word is occupied by the indicia of format) .
- the completed second word of the compressed data vector may now be stored in memory 12 or in the buffer 20c.
- the second word of compressed data of the compressed data vector begins by again loading the fourth data point. Again, the 11 left-most bits are deleted by masking. The remaining 5 masked bits are now shifted 12 bits to the left (resulting in one bit being shifted out of the register) and stored in the buffer 35 to begin constructing the third word of the compressed data vector.
- the fifth data point is loaded, masked, shifted 7 bits to the left, and stored.
- the sixth data point is loaded, masked, and shifted 2 bits to the left.
- the final data point of the second word of the compressed data i.e., the seventh data point
- the processor 32 then writes the second word of the compressed data into the third word location of the compressed data vector.
- the eighth through sixteenth data points are similarly masked and shifted and stored.
- the bottom row of Table II (labeled "OUTPUT") summarizes the stored number for the example of Table II. Shown in Table II is the INPUT DATA before and after reduction to 5 bit binary, in accordance with the process. At the bottom of Table II the five bits of each data point of the reduced data vector is shown packed into binary words.
- the first word (i.e., the indicia of format) of the compressed data vector i.e., 5
- Table III offers a further example of data compression under the third format.
- the bit size necessary to express the difference vector as a reduced difference vector is smaller than the number of bits necessary to express the data vector as a reduced data vector.
- the third format is used and the reduced difference vector is packed into binary words and stored as shown, along with indicia of compression.
- the indicia of compression in this case is -1.
- the digit 1 is indicative of the number of bits and the negative sign is indicative of the third format.
- the compressed data vector now requires one 16 bit memory location.
- the processor accordingly, allocates 3 words of memory (1 for the compressed data, 1 for the first data value and 1 for the indicia of format).
- the processor 36 (FIG. 7) determines a bit size for the largest data point of the data vector shown in the top line of Table III.
- the controller also creates a difference vector (by subtracting the first data point from the second, the second from the third, etc.) and stores the difference vector in the buffer 20d.
- the controller determines a bit size for the difference vector and compares the difference bit size with the data bit size. Since the difference bit size is less than the data bit size, the processor 36 determines that the third format is to be used in data compression for the data vector shown at the top of Table III.
- the processor 38 may begin constructing the compressed data by storing the indicia of format (i.e. -1) in the first word of the compressed data vector.
- the processor 38 also stores a value (e.g., the magnitude of the first data point) upon which the data vector can be reconstructed from the compressed data vector (which in this case is the stored, compressed difference vector).
- the processor 38 provides the value by storing the first data point (i.e., 32) in the second word of the compressed data vector.
- Packing 124 the difference points of the difference vector into word 3 of the compressed data vector occurs in a manner substantially identical to the process described above.
- the processor 38 loads a difference point of the difference vector from the buffer 20d, masks off the bits outside the determined bit size, and shifts to an appropriate position, as indicated in Appendix I, Algorithm 5.
- the shifted masked bits are then stored in the word buffer 42.
- the vector is stored in memory 12.
- Decompression under the embodiment is accomplished by a reversal of the process described above.
- the processor 50 (FIG. 8) reads a compressed data vector and may store the vector in the buffer 54. The processor then reads the first data word (indicia of format) to determine the process to be used in decompression and proceeds accordingly.
- the processor 50 recognizes the vector by reference to the look-up table 20e as a compressed data vector stored under the first format.
- the processor 50 accordingly sets an internal counter 56 equal to 16 the number of words in the uncompressed vector.
- the processor reads the second word of the compressed data vector into an internal buffer 54 and begins outputting a decompressed data stream to the post processor 18 based upon the contents of the buffer 54.
- the processor 50 decrements the counter 56.
- the processor reads another compressed data vector.
- the processor Upon reading the next compressed data word, the processor again inputs the first data word providing the indicia of format (e.g., the value of 5 as in the case of Table II). The processor by reference to the look-up table 20e determines that the compressed data vector is stored under the second format and proceeds accordingly.
- format e.g., the value of 5 as in the case of Table II.
- the processor 50 reads the second word of the compressed data vector (first data word) into the buffer 54. The contents are then transferred to the shift register 52 for unpacking. To recover the first data point the processor shifts the contents of the shift register to the right by 11 bits. Note that in the above, shifting to the right is done, with sign extension, so that signed values are produced. (Sign extension means that after a signed number is shifted to the right, the most significant bit of the original word is copied to the most significant bit of the shifted word. This function is an integral part of a 2s compliment processor.) The contents of the shift register 52 is then transferred to the post processor 18 as a decompressed first data point of the first data vector.
- the processor 50 again transfers the contents of the buffer 54 to the shift register 52. This time, the processor 50 causes the shift register to shift left by 5 bits and then right by 11 bits. The contents of the shift register 52 is then provided to the post processor 18 as the second data point of the data vector.
- the processor 50 again loads the contents of the buffer (second word of the compressed data vector) into the shift register and shifts left 15 bits.
- the shifted contents of the shift register 52 may then be temporarily stored in the buffer 20f as a partial data point while the remainder of the third data word is recovered.
- the processor 50 then loads the third word of the compressed data vector into the buffer 54 and into the shift register 52.
- the shift register may then be shifted to the right by one bit to produce an interim result and to make room for the remaining bit of the previous word (i.e., the partial data point) stored in the buffer 20f.
- the remaining bit may be appended to the contents of the shift register 52 by logically "OR”ing the contents of the shift register 52 with the partial data point stored in the buffer 20f.
- the contents of the shift register is then shifted 11 bits to the right to complete the unpacking of the third data point.
- the third data point of the data vector is then output to the post processor 18.
- the fourth data point is unpacked by loading the third word of the compressed data vector into the shift register, shifting to the left by 4 bits, followed by shifting to the right by 11 bits.
- the fifth data point is unpacked by loading the third word of the compressed data vector into the shift register, shifting to the left by 9 bits, followed by shifting to the right by 11 bits.
- the sixth data point is again divided between two compressed data words and a partial data point must again be stored in buffer 20f and again logically "OR”ed with the next interim result before finally shifting the entire contents to produce the final result of the sixth data point.
- the seventh through sixteenth data points are similarly unpacked.
- the unpacked data points are sequentially transferred to the post processor 18.
- the processor 50 may retrieve another compressed data vector. Again, the processor 50 first retrieves the first word as the indicia of format. The indicia is determined by comparison of the first word with the contents of an indicia look-up table 20e. Where the indicia is negative, the processor 50 determines that the compressed data vector is stored under the third format and uses the number associated with the indicia as a bit length of the difference vector stored after the second data word of the compressed data vector. After determining that the compressed data vector is stored under the third format, the processor 50 reads the second data word. Since the second data word stored under the third format is a magnitude of the first data point, it can be output to the post processor 18 without unpacking. A copy of the first data word, however, must be stored in the buffer 20f as a reference data word for a determination of the second data word.
- the difference vector must be unpacked from the third and subsequent words of the compressed data vector.
- the processor 50 unpacks the difference vector through use of the shift register 52 and the bit size recovered from the indicia of format. Following the reconstruction of the difference vector, the original data vector can be recreated by adding the first data point to the first difference point and the result to subsequent difference points.
- the indicia of format (i.e., -1) indicated a difference vector bit length of 1 bits.
- the controller reads the third word of the compressed data vector.
- Algorithm 5 it may be seen that to recover the first difference value of the difference vector the processor 50 simply causes the shift register 52 to shift right by 15 bits. The result is the transferred to a first difference location within a difference vector register in buffer 20f.
- the processor 50 reads the third data word into the shift register 52.
- the controller 50 then causes the shift register 50 shift left by 1 bit and then right by 15 bits.
- the result is then stored in a second difference location within the difference vector register in buffer 20f.
- the processor 50 again reads the third data word into the shift register 52. This time, the controller 50 causes the shift register to shift left by 2 bits and then right by 15 bits. The result is then stored in a third difference location within the difference vector register in buffer 20f. Difference points 4-16 are recovered in a similar manner.
- the first data point is read from the buffer 20f and added to the first difference point of the difference vector.
- the result is output to the post processor 18 and stored as a new reference data point.
- the third data point is evaluated by adding the new reference point to the second difference point of the difference vector.
- the result is output as the third data point and, also, stored as the new reference data point. Subsequent data points are evaluated in an identical manner.
- vector A sequence of words.
- Bw The number of bits in a word (here assumed to be 16).
- Bd The number of significant bits needed to represent either min_diff or max_diff.
- B The number of bits (Ba or Bd) used to compress a vector.
- n The number of words in the original vector, a multiple of Bw.
- P The number of words produced by a low level packing algorithm.
- m The number of words in the compressed vector.
- each vector contains data from one of the channels.
- C1. .Cm Compressed vector containing compression type, P1..Pp,
- INPUT A1..An or D1..Dn, B, n
- Each of these low level compression algorithms processes vectors 16 words at a time, because the patterns of bit fields repeat every 16 fields.
- Each algorithm consists of the shifts, ANDs and ORs needed to extract the significant bits from the input vector and pack them into the output vector using one of the patterns below.
- the sequence of 0's is the bit field for A1, A17, ..
- the sequence of, 1's- is- the bit field for A2, A18, .. the sequence of 2's is the bit field for A3, A19, ..
- the sequence of 9's is the bit field for A10, A26, ..
- the sequence of A's is the bit field for A11, A27, ..
- INPUT P1..Pp, B, n
- Each of these low level decompression algorithms consists of the shifts, ANDs and ORs needed to extract the original significant bits from the compressed data. Sign extension is performed to produce Bw bit values. Each algorithm expects the input vector to be in the pattern produced by the corresponding compression algorithm.
- the compressed vector is organized as follows:
- the compressed vector is organized as follows, where
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Multimedia (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Compression, Expansion, Code Conversion, And Decoders (AREA)
Abstract
Description
Claims
Priority Applications (3)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
JP9503076A JPH11514165A (en) | 1995-06-19 | 1996-05-29 | Apparatus and method for waveform data compression without loss |
AU59335/96A AU5933596A (en) | 1995-06-19 | 1996-05-29 | Apparatus and method for lossless waveform data compression |
EP96916653A EP0834221A4 (en) | 1995-06-19 | 1996-05-29 | Apparatus and method for lossless waveform data compression |
Applications Claiming Priority (2)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
US08/492,369 | 1995-06-19 | ||
US08/492,369 US5602550A (en) | 1995-06-19 | 1995-06-19 | Apparatus and method for lossless waveform data compression |
Publications (1)
Publication Number | Publication Date |
---|---|
WO1997000558A1 true WO1997000558A1 (en) | 1997-01-03 |
Family
ID=23955988
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
PCT/US1996/007740 WO1997000558A1 (en) | 1995-06-19 | 1996-05-29 | Apparatus and method for lossless waveform data compression |
Country Status (5)
Country | Link |
---|---|
US (1) | US5602550A (en) |
EP (1) | EP0834221A4 (en) |
JP (1) | JPH11514165A (en) |
AU (1) | AU5933596A (en) |
WO (1) | WO1997000558A1 (en) |
Families Citing this family (11)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US6307489B1 (en) * | 2000-03-15 | 2001-10-23 | Robert Allen Freking | Fast and small serial huffman decoder for decoding at an optimally high rate |
US7401180B1 (en) | 2001-12-27 | 2008-07-15 | Netlogic Microsystems, Inc. | Content addressable memory (CAM) device having selectable access and method therefor |
US7301961B1 (en) | 2001-12-27 | 2007-11-27 | Cypress Semiconductor Corportion | Method and apparatus for configuring signal lines according to idle codes |
US7298298B1 (en) * | 2003-04-15 | 2007-11-20 | Cisco Technology, Inc. | Multi-epoch codebook learning in a lossy packet network |
US7342902B2 (en) * | 2003-04-15 | 2008-03-11 | Cisco Technology, Inc. | Two stage loss-less compressor for a clear channel over a packet network |
US7466254B2 (en) | 2006-02-03 | 2008-12-16 | L&L Engineering Llc | Systems and methods for digital control utilizing oversampling |
WO2009061814A2 (en) * | 2007-11-05 | 2009-05-14 | University Of Florida Research Foundation, Inc. | Lossless data compression and real-time decompression |
US9165403B2 (en) * | 2011-01-14 | 2015-10-20 | Intel Corporation | Planetary scale object rendering |
US9648144B2 (en) * | 2014-01-24 | 2017-05-09 | Mediatek Inc. | Data compression method and decompression method |
US10031949B2 (en) * | 2016-03-03 | 2018-07-24 | Tic Talking Holdings Inc. | Interest based content distribution |
US10176623B2 (en) | 2016-05-02 | 2019-01-08 | Tic Talking Holdings Inc. | Facilitation of depiction of geographic relationships via a user interface |
Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US4633884A (en) * | 1983-05-24 | 1987-01-06 | Kabushiki Kaisya Advance Kaihatsu Kenkyujo | Physiological signals processing system |
US4777620A (en) * | 1985-02-20 | 1988-10-11 | Elscint Ltd. | Data compression system |
US5049880A (en) * | 1990-04-02 | 1991-09-17 | Tektronix, Inc. | Bit-wise run-length encoding for data compression |
US5119092A (en) * | 1988-11-22 | 1992-06-02 | Sharp Kabushiki Kaisha | Apparatus for encoding, decoding, and storing waveforms |
US5260693A (en) * | 1991-10-11 | 1993-11-09 | Spacelabs Medical, Inc. | Method and system for lossless and adaptive data compression and decompression |
Family Cites Families (17)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US4449536A (en) * | 1980-10-31 | 1984-05-22 | Sri International | Method and apparatus for digital data compression |
US4802222A (en) * | 1983-12-12 | 1989-01-31 | Sri International | Data compression system and method for audio signals |
US4579125A (en) * | 1984-01-23 | 1986-04-01 | Cns, Inc. | Real-time EEG spectral analyzer |
JPH0644712B2 (en) * | 1984-05-22 | 1994-06-08 | 株式会社アドバンス | Signal processing method |
US4684923A (en) * | 1984-09-17 | 1987-08-04 | Nec Corporation | Encoder with selective indication of compression encoding and decoder therefor |
GB8708010D0 (en) * | 1987-04-03 | 1987-05-07 | Crosfield Electronics Ltd | Image reprocessing |
US4920489A (en) * | 1987-08-14 | 1990-04-24 | Cardiodata Inc. | Apparatus and method for solid state storage of episodic signals |
US4882754A (en) * | 1987-08-25 | 1989-11-21 | Digideck, Inc. | Data compression system and method with buffer control |
JP2841362B2 (en) * | 1987-12-25 | 1998-12-24 | 松下電器産業株式会社 | High efficiency coding device |
CU22179A1 (en) * | 1990-11-09 | 1994-01-31 | Neurociencias Centro | Method and system for evaluating abnormal electro-magnetic physiological activity of the heart and brain and plotting it in graph form. |
JPH04185172A (en) * | 1990-11-20 | 1992-07-02 | Matsushita Electric Ind Co Ltd | High-efficiency coding device for digital image signal |
US5233629A (en) * | 1991-07-26 | 1993-08-03 | General Instrument Corporation | Method and apparatus for communicating digital data using trellis coded qam |
US5255186A (en) * | 1991-08-06 | 1993-10-19 | Telectronics Pacing Systems, Inc. | Signal averaging of cardiac electrical signals using temporal data compression and scanning correlation |
US5263486A (en) * | 1991-11-01 | 1993-11-23 | Telectronics Pacing Systems, Inc. | Apparatus and method for electrocardiogram data compression |
US5333615A (en) * | 1992-06-22 | 1994-08-02 | William Craelius | Apparatus for digitally recording and analyzing electrocardial and other bioelectric signals |
US5321645A (en) * | 1993-07-15 | 1994-06-14 | Hewlett-Packard Company | Simplified method and apparatus for loss signal compression |
JP3227931B2 (en) * | 1993-09-10 | 2001-11-12 | ソニー株式会社 | Encoding device and decoding device |
-
1995
- 1995-06-19 US US08/492,369 patent/US5602550A/en not_active Expired - Lifetime
-
1996
- 1996-05-29 EP EP96916653A patent/EP0834221A4/en not_active Withdrawn
- 1996-05-29 AU AU59335/96A patent/AU5933596A/en not_active Abandoned
- 1996-05-29 JP JP9503076A patent/JPH11514165A/en active Pending
- 1996-05-29 WO PCT/US1996/007740 patent/WO1997000558A1/en not_active Application Discontinuation
Patent Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US4633884A (en) * | 1983-05-24 | 1987-01-06 | Kabushiki Kaisya Advance Kaihatsu Kenkyujo | Physiological signals processing system |
US4777620A (en) * | 1985-02-20 | 1988-10-11 | Elscint Ltd. | Data compression system |
US5119092A (en) * | 1988-11-22 | 1992-06-02 | Sharp Kabushiki Kaisha | Apparatus for encoding, decoding, and storing waveforms |
US5049880A (en) * | 1990-04-02 | 1991-09-17 | Tektronix, Inc. | Bit-wise run-length encoding for data compression |
US5260693A (en) * | 1991-10-11 | 1993-11-09 | Spacelabs Medical, Inc. | Method and system for lossless and adaptive data compression and decompression |
Non-Patent Citations (1)
Title |
---|
See also references of EP0834221A4 * |
Also Published As
Publication number | Publication date |
---|---|
US5602550A (en) | 1997-02-11 |
EP0834221A1 (en) | 1998-04-08 |
JPH11514165A (en) | 1999-11-30 |
AU5933596A (en) | 1997-01-15 |
EP0834221A4 (en) | 1999-12-29 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
US5293164A (en) | Data compression with pipeline processor having separate memories | |
US5839100A (en) | Lossless and loss-limited compression of sampled data signals | |
US6006232A (en) | System and method for multirecord compression in a relational database | |
US5260693A (en) | Method and system for lossless and adaptive data compression and decompression | |
US5602550A (en) | Apparatus and method for lossless waveform data compression | |
CA2139095C (en) | Real-time digital audio compression/decompression system | |
JPH0568893B2 (en) | ||
US5857036A (en) | System and method for the fractal encoding of datastreams | |
CA2117720C (en) | Data compression method and apparatus for waveforms having recurring features | |
US5049880A (en) | Bit-wise run-length encoding for data compression | |
US20130018856A1 (en) | Compression of bitmaps and values | |
US6292115B1 (en) | Data compression for use with a communications channel | |
EP0597126A1 (en) | Method for compressing sound information and apparatus for reproducing sound information | |
US5710919A (en) | Record compression | |
US5739778A (en) | Digital data formatting/deformatting circuits | |
CA2280623C (en) | Real-time digital audio compression/decompression system | |
JP3258526B2 (en) | Compressed audio decompression device | |
US7218263B2 (en) | Generating bit-streams with higher compression gains | |
JPWO2005053163A1 (en) | Signal processing device | |
Andrii et al. | UDC 004.627 LZW COMPRESSION ALGORITHM | |
JPH0778065A (en) | Data array processing method | |
Adams et al. | Error-free storage compression of binary-coded infrared spectra | |
Soraghan et al. | ECG signal compression using classified gain-shape vector quantisation in the wavelet transform domain | |
Honea et al. | Lossless waveform compression: A case study | |
JPS6356020A (en) | Compression decoding device |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
AK | Designated states |
Kind code of ref document: A1 Designated state(s): AM AT AU BB BG BR BY CA CH CN CZ DE DK EE ES FI GB GE HU IS JP KE KG KP KR KZ LK LR LT LU LV MD MG MN MW MX NO NZ PL PT RO RU SD SE SG SI SK TJ TM TT UA UG UZ VN |
|
AL | Designated countries for regional patents |
Kind code of ref document: A1 Designated state(s): KE LS MW SD SZ UG AT BE CH DE DK ES FI FR GB GR IE IT LU MC NL PT SE BF BJ CF CG CI CM GA GN ML MR NE SN TD TG |
|
DFPE | Request for preliminary examination filed prior to expiration of 19th month from priority date (pct application filed before 20040101) | ||
121 | Ep: the epo has been informed by wipo that ep was designated in this application | ||
ENP | Entry into the national phase |
Ref country code: JP Ref document number: 1997 503076 Kind code of ref document: A Format of ref document f/p: F |
|
WWE | Wipo information: entry into national phase |
Ref document number: 1996916653 Country of ref document: EP |
|
WWP | Wipo information: published in national office |
Ref document number: 1996916653 Country of ref document: EP |
|
REG | Reference to national code |
Ref country code: DE Ref legal event code: 8642 |
|
NENP | Non-entry into the national phase |
Ref country code: CA |
|
WWW | Wipo information: withdrawn in national office |
Ref document number: 1996916653 Country of ref document: EP |