TW564621B - Packet header compression using division remainders - Google Patents

Abstract

The present invention is directed to providing a method and apparatus for efficiently compressing and reconstructing a header of a real time communication packet. In a header compressor (28), a header field value is applied to a modulo X operator (101) which can, for example, divide the header field value by a value X, and output the remainder. Optionally, a checksum (38) may be appended to the remainder. The header field value may be scaled (35) prior to being applied to the modulo X operator (101). The compressed header field which is output from the header compressor (28) includes the remainder with or without the checksum appended. A header decompressor (53) includes a field reconstructor (110) which reconstructs the received compressed header field (111) in response to the remainder value (118) and range information (112). The range information (112) represents a range of possible field values which can be reconstructed from the received remainder value (118).

Description

564621 发明 Description of the invention: Cross-reference of related applications This patent application claims the priority of US Provisional Patent Application No. 60 / 164,355 filed on November 9, 1999, which is also in the same application process. This patent application The contents disclosed in the case are hereby incorporated by reference. Scope of the invention The present invention relates generally to packet communication, and more specifically, to header compression in packet communication. BACKGROUND OF THE INVENTION Title Compression (HC)-A term that refers to a technique that minimizes the required bandwidth of information carried in a packet via a point-to-point link and according to each hop. Title compression is usually achieved by sending static information only at the beginning. The semi-static information is then transmitted by sending only the delta (Δ) from the previous title, and 疋 completely random information is transmitted without compression. Therefore, header compression is generally implemented using state machines. The design of the traditional header compression algorithm is basically for narrowband line communication, and the complexity on the receiving side of the decompression is quite small. Similarly, the compression side of the transmission is kept low complexity in order to have efficient subscription in the router, and as much computing energy in the router as possible is needed for path selection. In addition, the existing title algorithms are designed for wired channels, and generally have a very low bit error probability (for example, a bit error rate of 1 () _ 6). Wireless channel (-generally has the characteristics of loss and narrow-band link) _ generally has a higher probability of error, so for header compression in the wireless channel, when designing, you need to imagine that there may be a large probability of bit error (For example, up to i (bit of m = 564621 92, margin). The traditional compression mode for RTP / UDP / IP is often based on a soft machine with a state called context. Solution The compressor context is often updated by each received packet. If the packet is missing on the link, the context will become invalid. When the decompressor context is invalid, all consecutive packets must be discarded until the soft state is removed. A complete (uncompressed) header is updated. When the decompressor finds that the first packet is discarded (missing), an update request is sent from the receiver, and this request is then updated (that is, the header with the uncompressed header). (Packet) before arrival, complete the full roundtrip (ie, from the receiving end to the sending end and then return). If the receiving decompressor fails to successfully decompress the -compressed header, this gyroscope leak may also be sent If the data of the packet with the compressed title contains real-time service = the loss of consecutive packets will have a great impact on the quality of this real-time service; for example, the quality of the real-time voice service will be lost due to continuous loss. The speech towel will decrease sharply with the increase of the sealing & leakage rate. ⑦ 位 / ji 目 士 #, not ^ 大量 The quality of the speech in the mouth is large. The voice quality will be better than that used for the same ^ The rate is worse, but the cut and poor characteristics. ~ The misuse of the field of the pass reduces the invalid context and the reason for this. For example, the method of decompressing the heart = the bit of the title is borrowed. Increasing the probability of "correct guessing" of the correct contextual state and increasing the receiver, even if the "money rate" and "RTp" are known. As an example of a real-time voice service, the value of the RTP time stamp (stamp) is increased in the slave test mode (and therefore can be described between m and 』)-generally predictable (and so on ...) prediction Or guess) 564621

During non-speech periods, time is more random from the receiver's point of view. Existing standards for RTP / UDP / IP headers (see, for example, IETF RFC 2508 in the IETF Network Working Group in February 1999

The paper proposed by Jacobson " compressing Ip / UDp / RTp for low voice serial linking " is called CRTp in this paper. In CRTp, the value of △ is coded according to an indefinite number of bits of this value. The large time change from the last packet causes a large △ value. This unfavorable situation makes the title towel under pressure need to have more? Bits indicate the delta value of the information by the time indicated by the carrier. Λ Whenever D T X (discontinuous transmission) or silence suppression is used for real-time voice service ^ RTP ‘Question day mark field’ will have unpredictable random behavior in the RTP / UDP / IP packet stream carrying voice. Therefore, the time-stamping position is the most difficult position to stop the compression of money. According to CRTP, time, the value of △ is encoded by the right stem TL of the large I that changes according to the time mark from the last packet. Therefore, a long period of silence or non-speech period: f multi-bits are marked with △ modulation time, so that the initial question after the silence period-in general, will be larger than the title in the voice packet corresponding to the voice period . Shrinking technology, and it is not desirable to provide a method for title shrinking / decompression that has the above disadvantages associated with the traditional model. Summary of the invention Compressing and recompressing in an efficient manner, this time stamp value will not be located in the present invention, and a time stamp value for constructing an instant messaging packet is provided in a meaningful manner.

O: \ 67 \ 67386-920808 DOC The sequence of timestamp values expected under normal circumstances. The first part of the time stamp value is selected by the header compressor and then sent. The second part of the time stamp value is estimated by the header decompressor based on the elapsed time between successive packets received. The header decompressor combines the second part with the first part received from the header compression benefit. In addition, the present invention relates in particular to techniques for providing effective compression and reconstruction of the header of an instant messaging packet. Applying a header compression and shrinking to a header applies a header block value (this value can be proportionally processed) to a modulo x operator y. This operator can, for example, divide the header field value by the value x, and output: remainder. -The option is to check and append to the remainder. Therefore, the compressed header block may include the remainder as the output of the modulo x operator, with or without a checksum attached to it. A title decompressor includes a block re-constructor that reconstructs the compressed header block in response to the remainder value and range information. The range information represents the range of possible field values, whose values can be reconstructed from the remaining values received. A more complete understanding of the present invention and its scope can be obtained from the drawings briefly described below, the detailed description of the presently preferred specific examples of the present invention, and the scope of the attached patent application. Brief description of the drawings / a more complete understanding of the method and device of the present invention can be obtained by referring to the detailed description of the following article A and the accompanying drawings. In the drawings ... Figure 1 is a conceptual illustration of the basis Exemplary time stamp compression and decompression techniques of the present invention. FIG. 2 illustrates an exemplary data transmitting station according to the present invention.

O: \ 67 \ 67386-920808 DOC 564621 FIG. 3 illustrates a specific example of the title compressor in FIG. 2. FIG. A illustrates an example of the time mark field in FIG. 3. The example in Figure 4 cannot be performed by the title compression in Figures 2 and 3. Illustrative implementation of m eight-body example Figure 5 illustrates an exemplary packet data connection according to the present invention. FIG. 6 illustrates an exemplary specific example of the title decompressor of FIG. 5. Fig. 7 illustrates a specific example of the irregularity of the time stamp decompression of Fig. 6-7 ·-fixed rail ". Fig. 7A illustrates an example of the decompression of the time stamp of Figs. 6 and 7; Exemplary operations can be performed at the time shown in Figs. 6 to 7. Fig. 9 illustrates the time mark estimation of the exemplary operating principles that can be implemented in Fig. 8. Fig. 10 is explained in proportion. Ways to exemplify the implementation of this part of the compressor. /, The general example of the title-Figure 11 illustrates in a graphical manner the part according to the present invention. Material U-specific examples-one of the header compressor A check FIG. 12 illustrates an exemplary specific example for the title decompression bit reconstructor of FIG. U in a graphical manner. FIG. 13 illustrates an exemplary specific example of the anchor construction 1 in FIG. 2 in a graphical manner. Specific Example Figure 14 illustrates that the title can be decompressed as shown in Figures 丨 丨 to ^

Exemplary operations implemented. V From any number of matching targets Figure 15 illustrates graphically an example of selecting alternative values for the question fields according to the present invention 564621 92 8, 08. A detailed description of one of the correct title field alternatives, an exemplary embodiment, and a presently preferred exemplary embodiment, will now be described in more detail with reference to the accompanying drawings for the present invention in the accompanying drawings. There are preferred specific examples of the present invention. Although the present invention may be embodied in many different ways, it should not be construed as limiting the specific examples set forth herein; these specific examples are provided so that

The contents disclosed are more thorough, and the scope of the present invention is fully expressed to those skilled in the art.

Figure 1 conceptually illustrates an exemplary time stamp compression and decompression technique for instant messaging applications such as instant voice response 1-way instant messaging applications in accordance with the present invention. Basically, the receiver's header decompressor uses a local clock. To estimate the time that elapses between the last speech packet before a segment of speech is still and the first speech packet after a segment of speech. Based on this elapsed time estimation, the header decompression 11 can make an estimate of the difference (or △ value) between the time stamps of the two voice packets of the boundary speech still. The difference between the time stamp values can be used in combination with the known time stamp value of the last voice packet before the voice is still to make an educated guess about the time stamp value of the first voice packet after the voice is still. As shown in Figure 1, at the title compressor at the transmitting end, after the voice is still at rest, the "smallest effective bit" of the packet on the day of the packet is less than 8 (18 is called [selected by Channel 13 is transmitted. Channel 13 may be a wireless channel, such as a UMTS air interface or other cellular radio interface. The estimated time stamp of the received packet at the receiving end 15 may be 564621.

8. 〇B η described in the exemplary manner. Let packet n] be the packet received during the voice quiescence period, and let the packet be the second consecutive voice packet, that is, the-voice packet after the period. If the compressor located at the receiving end notices that the packet η-1 arrives, it is called -υ, and it also pays attention to the packet arrival time τ⑻, then the absolute time difference between the two packet arrivals can be estimated by subtracting ㈣. . This time difference represents the elapsed time between the arrival of the packet_. This elapsed time can be multiplied by the indicated value unit by how much time there is in each unit of%. H such as time 4 is converted into time. Now △-T is the time warp represented by the difference between the loading time, and TS- is changed to the estimated value of how much the time value changes per unit time. TS _Change the value, and use the estimated unit of production time and the elapsed time to focus on the estimation, and change: two ':: packet n] and the difference between the time stamp of packet n The value, that is, the TS-estimated value, is expressed by taking the time-marked value + the one-value (TS-change multiplied by Δ-T) plus the known two values of the packet ηβ1. —Don’t Ts—The estimated value is executed at β ^ The most significant bit of the estimated value is the smallest value attached to the real time label TS. It belongs to the version L received, so the time label of the packet n is generated. Guess it. At 17th, the title de-duster tried out a trial decision: T,: know is the correct guess for the original time mark TS. If it is not, then :::, another-item guess, this processing step can be repeated to β, until a Zhiliu can be produced or a suspension condition can be satisfied. Fig. 2 is an exemplary time stamp compression technique shown in Fig. Lt. 564621 I " -I—I • I ~ I》 一一 ^ Ί: Hugh, 1: | U —_____ ㈣ One exemplary packet data transmitting station . This transmitting station may be, for example, a transmitter that operates in a fixed position or mobile in a communication network. In the specific example of FIG. 2, a packet data communication application generates data block glycosides at its location and header information at 26 locations. The data column information can be used in a conventional manner by the data block 23 processor 20, and the header information 26 is added to a header compression σσ. The core information is reduced to M, and the header information M is reduced to generate the resized header 22. The compressed header 22 and data block 23 constitute a packet 21. A conventional radio transmitter 29 may use well known techniques to transmit packets 2J over a radio link such as a cellular radio link. The communication application program 24 also includes a continue execution signal 27. This signal indicates that the current column and title information of 2 5 and 26 corresponds to an RT p voice packet. This packet is to be transmitted after a period of speech is still. The first voice packet (corresponding to the packet n described above with reference to FIG. I). The title compressor M responds to the activation of the signal 27 to implement, for example, an innovative time stamp compression technique including the time stamp compression technique illustrated in the figure. FIG. 3 illustrates an exemplary specific example of the title compressor 28 in FIG. 2. In the specific example of the title compressor of FIG. 3, a separator 33 receives the title information 26 from the communication application 24. The separator 33 separates the time stamp stop information from the other title information at%, so that the time stamp information can be compressed separately from the rest of the title information. A divider 35 divides the time scale value by a scale value (TS increment) and processes the time scale value proportionally. Taking an instant speech service that carries speech information generated by a speech codec with a constant bit rate as an example, the time stamp can be expected to increase by a constant increment per continuous packet during a period of speech activity. The value TS—increment represents O: \ 67 \ 67386-920808 DOC -12 · 564621 i v_ y Years and months supplementation = Estimated value over several days. This estimated value can be determined based on empirical observation. Therefore, the divider The operation of 35 is to reduce the time-marked value proportionally, and f is decreased to represent the number of bits required for the time-marked value. In other concrete cases, the removed state 35 is shown as a broken line in the figure, and 彳 is omitted or used in a manner. Tan-The Least Significant Bit Extractor 36 receives the scaled time stamp value from the divider 35 and extracts the least significant bits (LSBs) from the proportionally processed value. At 37, an additional device is added to the continuation code generated by the -compiler in response to the enable of the continuation of the execution of the message wear # in FIG. 2 ^ = 2 may be added by a selective check and generator 38 and The checksum (for example, the CRC checksum) generated by the time stamp does not reflect (think of it) the other poor news (see the broken line in Figure 3). Output of additional equipment

Add to a selector 3〇_ 于 入 1. + The input of L is 39, the other input of this selector is connected to the output of a conventional time stamp compressor 301, and this compressor also receives the time stamp value from the splitter 33. The operation is controlled by the continue execution signal 27. Therefore, if the execution is continued: for effective use, the LSBs' continue to execute the execution, and check and supply to the shore via the selector 30. 夕 2 Ή 茂 Ή " a picture 2 ^^ The factory side of the time mark stop of title 22 'If the signal 27 of continued execution has no effect, on the traditional day, the output of the indentation section 301 is provided to the time mark standstill. : Sample ::: The one in Figure 3 is another title of the separator 33. The output can be used as shown in 30_ 'results. Other Blocks 32 in Title 22. Menyi Chuan, Xian Fang 564621 Figure 3A illustrates that when the n-successors μ-L in Figures 2 and 3 continue to function, 27 is a non-blocking 31. As shown in Figure 3, Basuo mm Μ ^ ^ ^ L Tau knows whether to block 31. It includes taking advantage of only the code, and after the proportional processing, it is spelled out%-_ Dingmao worthless LSBs, and as shown in Figure A, No, you can choose the way to include the checksum generated at 38. Figure 4 illustrates an exemplary time stamp pressure ^ ρ This preparation can be implemented by a specific example of the dry question compressor shown in Figure 3. Before 41 Decide whether or not to continue to execute the h signal as an effective signal. If it is, you will then implement the traditional method without compressing it at 42 places, and then send it to ^ ^ at 48 places. If you continue to implement The signal at 41 is an effective signal ~ time is not worth (see TS in Figure 1) ρ is used to generate-check sum. Then, the time label value is incremented by ts Proportional processing is performed. Later, the 44 least significant bits are extracted from the proportionally processed time ^ value towel, and the execution code and checksum (or sexual) secret are added to the least significant bits. Line graphs indicate the generation and scaling operations of checksums at 46 and 43, which can be omitted or selectively applied to other specific examples. In the least significant bit and continue execution code (and, in some cases, there are also After checking and adding together at 45 places, then the time mark stop can be combined into a compressed header at 47 places, after which a packet is waiting at 48. Figure 5 Example of a packet receiving station An exemplary embodiment, the receiving station yoke the non-uniform time stamp decompression technique as illustrated in Fig. 1. The receiving station may be, for example, a fixed station or a mobile receiver operating a cellular communication network. In the specific example of FIG. 5, a conventional radio receiver 54 may use a well-known technology to receive a transmitted signal from a radio communication link, such as a cellular radio link, such as one of the packets 2 illustrated in FIG. O: \ 67 \ 67386-920808 DOC -14- 564621 Private 8⑽ 丨! One of the received versions 21, as shown in Fig. 5. Such received version 21 may include the compressed header in Fig. 2 The received version 22 of 22, and the received version 23 of the data column 23 of FIG. 2. This received data block version 23 can be supplied to the data column processor 58, which is traditionally generated for The input received block information of 5 1 is then sent to the packet communication application 52. The received compressed header version 22 is supplied to the header decompressor 53, which will decompress the received version 22 and decompress it. Compress to generate the received header information for input at 50 and send it to the communication application 52. FIG. 6 illustrates an exemplary specific example of the header decompressor in FIG. The received version 22 of the compressor header is input to the size detector 61. This detection can be performed using conventional techniques to detect whether the received packet is an RTp packet. In response to detecting that this packet is not an RTp packet, this packet indicates that a speech quiescence period is occurring, and the detector 61 starts an output signal%. This controls the selectors 68 and 69 appropriately so that the compressed header can be changed by Conventional header decompressor 64 processes. If the detector 61 decides that the RTp packet has been received, the control # 66 controls the selectors 68 and 69 so that the compressed header can be processed via a processing path 600, which completes the time period τ according to the present invention. Unzip the shed. Process yttrium 600 includes a separator 65 which separates the time stamp field from the other stops of the received version 22 of the compressed header. Except for the received version except the time mark field (see 32 in Figure 3), it can be added to 6 ± 7 ^ traditional title decompressor. The time mark at 63 places was entered into the guard, so it was not decompressed ^ benefit 60. Time stamp decompressor also receives RTp detection

O: \ 67 \ 67386-920808 DOC -15- "Control of 1 output ^ Say 66 'as its input. Responsive to the control signal and mark the block received at 63, time mark decompressor at A time stamp is output at 62. This time stamp is added to the other decompressed header information generated by the compressor 67 by means of additional device labor. Therefore, the formation and selection method I is coupled to the communication of FIG. 5 by the selection 69 The received header information required by the application 52 (see 50 in FIGS. 5 and 6). FIG. 7 illustrates an exemplary specific example of the time stamp decompressor 6 in FIG. 6. The exemplary in FIG. In a specific example, the time stamp received at 0 is input to a code detector 70, which is used to detect the continuation code in Fig. 3. If the continuation code is not detected, the received RTp field is not a paragraph. The first voice packet after Lu Jiu period, so the code detector 70 outputs the control k 702. This control signal appropriately controls the selectors 703 and 700, so that the traditional time mark decompressor 73 can change the time. Mark the time stamp required for decompression and generation at M (See also Figure 6.) If the code detector 70 detects that the code execution is continued, the control signal 702 controls the selectors 703 and 700, so that the time mark stop can be based on the exemplary time mark field of the present invention described above. Decompression technology to decompress. In this case, the received time stamp stop 63 is input to the picker 72 via the selector 703, which extracts and checks from the time stamp field. The received version (see Figure 3A). It should be noted that the execution of the code is only one example of the decompression operation technique required for triggering. The inter-standard estimator 75 can generate the general description about the above-mentioned figure i. TS-Estimated Time Stamp Estimation. The time stamp estimator has an input time stamp of 705 for receiving packet n-1, that is, a period of speech inactivity time O: \ 67 \ 67386-920808 DOC -16- The time stamp of the last RTP packet received. The time stamp value TSh-i) generated by the decompressor 73 is stored in the storage unit 77, and the unit is then lightly closed to the estimator input 705. The self-decompressor is One labeled output can be stored in The storage unit 77 (this unit can be a single register), so that when the packet n arrives, the time stamp 2 of the packet = TS (nl) can be used by the time stamp estimator 75. The time stamp estimator 75 It can also receive information indicating the time (T⑻ and T㈤) of the envelope ^ and packet ^ already received. This time information can be obtained from the storage list W6, and the storage unit 76 is lightly closed to receive the partial 2 obtained from the local clock μ. For each RTp packet detected by the detector 61 in FIG. 6, the storage unit 76 stores the arrival time of the packet measured by the local clock 74. The storage unit 76 only needs to be stacked in depth to capture the above packet n and n]. The% time & non-estimator 75 can also reach the above-mentioned time mark change value «, and the above-mentioned time mark increase value by-increment. This time stamp estimation can be operated in response to the local time information received from the storage unit 76, the time stamp value received from the storage unit γ, and ㈤), and the time stamp change in time to set the value of a to 1 in order to generate Estimated roughly as described above. The TS-estimated value is added to the maximum significant bit extractor 78, which extracts from the estimated value the most significant bits (MSBs) that constitute one of the truncated estimated values of the time labeling value. An additional device 702 adds the least significant bits (LSBs) taken from the extractor 72 to the most significant bits (⑽bs) output from the extractor 78, and the result is multiplied by TS in the multiplier 71 -Increment, thus producing a guess of ts as described above. The time-stamped estimator uses an increment of D s, by which

O: \ 67 \ 67386-92O808 DOC -17- The same as shown in Figure 3, the time stamp estimates are scaled down to facilitate the accurate combination of MSBs and LSBs at 702, so multiplication State 71 can be used to rescale this result to produce a guess value. The / checker 79 receives as input the TS-guess value and the collated version from the picker 72. The checker 79 is operable to generate a checksum from the received TS-guessed value and (if optional) the self-defining title ^ (not shown in the polyline in the figure), and then check the generated check Compare with the checksum received. If the two checksums match, the checker wheel ㈣ 704 activates a connection unit 701, which connects the ts-guess value to the selector 700. If the check received by the checker 79 and the mosquito do not match the generated check, the control signal 704 keeps the connection unit 7G1 in its open circuit position (as shown in figure :), and then informs the time mark estimator 75 Need another-time indicates the estimated value. The time stamp estimate II can therefore continue to generate time stamp estimates until it is checked and matched or until, for example, the time stamp estimates are checked and matched, or until, for example, the time stamp estimator is completed or the checker 79 is satisfied. Until the suspension. TS—The number of bits in the estimated value may be equal to, for example, the number of bits received by the lsb extractor in FIG. 3; ^ The number of bits in the value is not 'The number of deleted 8 extracted by the extractor M in FIG. 7 may be It is equal to the maximum number of significant bits retained (and abolished) after the LSBs are extracted at% in Figure 3. The number of LSBs picked at 36 and the number of centimeters picked at 78 can be determined by empirical observation, for example, to determine which combination of Lsb / msb wire can produce kilograms in various cases. 'Different combination of L-Fighter SB'

O: \ 67 \ 67386-920808 DOC -18-564621 M ;, R Οδ can be used according to factors such as transmission delay variation and clock accuracy in the compressor and decompressor. The desired combination of LSB / MSB extraction can therefore be determined by empirical observations under various transmission delay variations and different clock accuracy. For example, the number of MSBs extracted from the premises may be determined by the accuracy of the clock 74. The more accurate the clock 74 is, the more MSBs are picked at 78, and vice versa. The number of LSBs picked at 36 can then be determined based on the number of 5 picked at 78. The compressor and decompressor can be pre-programmed to complete the required combination of LSB / Msb extraction, or in the packet The combination is changed dynamically during the flow. For example, the compressor may choose the number of LSBs to be extracted based on the real change of the time stamp value, and send this message to the decompressor, for example, as a part of the continuation code illustrated in FIG. 38. FIG. 7A illustrates an alternative specific example of the decompressor in FIG. 7 with a broken line, in which the connection unit 701 (and the checker 79) in FIG. 7 is the same as the check and use or omission in FIG. Use in a selective manner; and / or the multiplier 71 is the same as the use or omission of the divider 35 in FIG. 3 and is omitted or used in a selective manner. The estimator 75 is consistent with the division and multiplication processes, and the TSj value is processed proportionally or the proportional processing of this estimated value is omitted. FIG. 8 illustrates an exemplary time stamp decompression operation, which can be implemented by a specific example of a time stamp decompressor in layer ⑴A. Yu Jiao first set the time to indicate whether the block includes the execution code. If it is not, then use the traditional solution of the agricultural department to shrink the time mark to block the time mark, then-seal the spoon and wait at 89. If code execution is detected at 8G, it will be

O: \ 67 \ 67386-920808 DOC -19- 564621 92, 8. 08 indicates the estimated value ⑽—estimated value, which is proportionally processed at 82), with a maximum of $ ^ (depending on which) is a significant significant digit That is to say, at the point μ has been said by Yura at 84, and the smallest received in the compressed title is:. On the basis of the estimated effective bit after proportional processing, that is, ^.,. ^ TT is added to the larger significant bit, and 7 is re-scaled (for time ―: howling is then called at 85, time ㈣ 1Γ) 'The resulting check is compared with the check in the time stamp stop at 86. If the check generated and the sum match the received sum', the time stamp guess is accepted at 87, missing The next time it waits at 89. If the check generated and received is not at 86 — the decision whether to abandon the estimated time mark, for example, based on a predetermined elapsed time value or the number of guesses in advance. Then calculate the scaled time-marked estimated value at 82, and repeat the operation at 83. At the time of making another time-marked estimate, the day guard 'estimator 7 5 can, for example, change one or more self-estimated values. The least significant bit of the bile s. In the example, if changing a specific bit (or a digital bit) yields a re-evaluation of the success of the -set packet time labeling, then: when -estimate -the subsequent packet time labeling, But first Do the same: Change. If you decide to give up at 88, the next packet is waiting immediately. The broken line in Figure 8 corresponds to the specific example in Figure 7A, where the check of the checksum is omitted or implemented in an optional manner. Fig. 9 illustrates an exemplary operation that can be implemented at 82 in Fig. 8 to calculate an estimated time scale. At 91, the time (T) nT (nl) that has elapsed since the last rule!) Packet is determined. At 92, the elapsed time is converted to time O: \ 67 \ 673E6-920808 DOC -20- 564621 Inter-standard: unit (using TS-change). At 93, the percentage between the number of passes passed at 92 and not only 70 is added to the last RTP packet (the time stamp value (TS⑹) of the packet is called) to generate a time stamp estimate. At the material point, a J The number of points (TS-addition) is added to the estimated value of the time mark generated at 93, thereby generating the estimated value of the time mark that is required to be proportionally processed. 1 In the specific example of FIG. 7A, the proportional processing is omitted or implemented in a selective manner. According to the 7F dry operation mode, the execution code in FIG. 3A is not required. According to this method, the time of FIG. 1 is always used to indicate compression and decompression Compression technology 1 can thus always control the selectors 30, 703, and 700 (see Figures 3 and 7) to select "Υ". In the corresponding case, the operations at 41 and 42 in Figure 4 and in Figure 8 at The operations at 80 and 81 can be omitted in this mode. Obviously, above and included in this article for reference, U.S. Patent Application No. 09 / 335,55 One of the clarifications in Advantages: The number of bits required to encode the time-stamped value can be reduced. The number of bits required to encode the time-stamped value can be maintained regardless of the size of the time-stamp change. Because the absolute time stamp value is encoded in the compressor, rather than the number of time stamp changes, the health is increased. In addition, in light of the above progress, the applicant has discovered improvements in the title compression / decompression technology According to the concrete example of the innovative pole title compressor / decompressor illustrated in Figs. 10 to 14, 'title compression is performed by sending only the title booth value modulus X' instead of all title title values. Title title values The modulus χ is the remainder obtained by dividing the value of the title field by X. If, for example, 6 is corrected, the standard

O: \ 67 \ 67386-920808 DOC -21-2 The 4 least significant bits of the block represent the remainder mentioned above, which directly constitutes the modulo X of the block value of question T. It can be clearly seen that as long as X is a power of 2, the remainder can be directly formed by a certain number of the least significant bits of the title block. Figure 丨 0 illustrates graphically the completion of a part of the modulus x header compressor. For example, any designated title stop serving the separator 33 of FIG. 3 can be compressed in an advantageous manner by the specific example of the title compressor of FIG. As shown in Fig. 10, a divider 35 may be used to process the title field value in a proportional manner, or as shown by the polyline, it may be omitted in a selective manner. The value of the title block (scaled or unscaled) is divided by X, and the remainder is output. Figure 10 also exemplifies a checksum (such as the one generated at 38 in Figure 3 above) to the remainder. Therefore, the compacted title field includes the modulo X operator 101, with or without the checkup attached to it. And the remainder. As can be clearly seen from the following, the compression / decompression techniques described with reference to Figs. 10 to 14 can be applied not only to the above-mentioned time stamp fields, but also to various title booths. FIG. 11 graphically illustrates a portion of a title decompressor according to the present invention, which includes a field re-constructor 110 that reconstructs the title in response to the residual value (compressed title field value) received at 118 Field. The rebuild 110 uses the range information at 112 places along with the remaining values received to generate a reconstruction field at 111 places. The range information at 112 indicates that the stop can be reconstructed from the remaining value received. The range information at 112 represents the range of possible field values that can be reconstructed from the remaining values received. The size of this range corresponds to the value of the modulus X operation performed in the header compressor of FIG. The position O: \ 67 \ 67386-920808 DOC -22- relative to the range of the most recently reconstructed field value generated by the field reconstructor 110, for example, based on the characteristics of known or expected packet communication applications and (Or) The characteristics of the packet communication path between the known compressor and decompressor are defined. According to an example, this range is set from to X + M. Each value in the range represents a possible difference between the most recently reconstructed stop value and the new stop value reconstructed by the decompressor. For example, if and X 16 ′, the range is i to 丨 6, this range represents 16 possible barrier values that exceed the latest reconstruction barrier value 1 to 16, respectively. As another example, if χ = ΐ6 and M-1, the range is set from-; 1 to 14. This range represents 16 possible stop values where the value (corresponding to -1 in the range) is a value less than the value of the most recently reconstructed field 1 and the other value (corresponding to "0" in the range) is equal to The most recent reconstruction barrier value, the remaining 14 possible barrier values exceeded the last reconstruction barrier value by 1 to 14, respectively. A positive value of M ', for example, can be effectively used to apply a sequence number field (for example, RTP sequence number block) to a packet arriving out of order. A positive value of μ also has its advantages, such as its negative delta time indication field, which is suitable for use in packets that include so-called B-pictures (bidirectional pictures predicted in MPEG applications). In MPEG (moving day file format), the 'B-picture' is traditionally transmitted after it surrounds the anchor picture temporarily. Therefore, compared with the transmission order, it will cause the front and back to jump on the time stamp field value. Therefore, the time mark △ is sometimes negative and sometimes jumps forward. This is a well-known phenomenon for technical workers in the industry. Selecting the M value so that the range includes the Π0Π value can be applied to the image divided into many packets with the same time mark field. FIG. 12 illustrates an example of the exemplary O: \ 67 \ 67386-920808 DOC -23-body of the column reconstructor 11 of FIG. 11 in a graphical manner. The range information and the newly reconstructed fields are input to an alternative field generator 125 at 112 and 113, respectively. The generator 125 responds to these inputs to provide an equal range size at 126, such as equal to the modulus and Several alternative blocking values for the X value of the operator 101. The alternative field values at 120 are entered into one of the modulo X operators 101 as shown in FIG. 10, and are also buffered in. The modulo X operator 101 outputs the remainders related to the buffered alternative field values, respectively. A comparator 123 compares the received remainder at 118 with the remainder of each alternative block value. Since there are alternative field values adjacent to 乂 in the range, because the received remainder at 118 represents the remainder of the division by χ, the remainder of one of the alternative field values of X will be Matching the received remainder 118, and the corresponding alternative stop value is rotated out of the buffer at 122. As shown in Fig. 12, the output of the stop value at 122 can be enlarged according to this example as needed, so as to be applicable to any scaling down possible in the title compressor in Fig. 10. This scaling up can include adding the remainder due to the scaling down division operation in the title compressor. The & remainder is generally a constant that only needs to be emitted one at a time, for example, by sending the complete field at the beginning of the packet stream The value 'informs the header to decompress the small remainder in an inclusive manner. As shown in Figure 12, the actual value at 121 (can be equal to ㈣ or not scaled) can be used as the output of the restructured value ⑴ of Figure u. This reconstructed stop value is also stored in the buffer, using it as the most recently reconstructed shelf in the sub-title reconstruction operation. In one specific example of x = 12, 4 bits are required to represent the operator

O: \ 67 \ 67386-920808 DOC -24- 564621 M y0 \ The block value of the source (see Figure 10) is modulo 12, but all possible bit patterns other than 4-bit t are required. Specifically, when the value of a title is divided by the operator D101 and the order 12 is divided by 12, it only requires 0_ money combinations to express the 12 possible remainders, so it can leave 4 bit patterns ⑽ ... ⑴ ) For other purposes such as signaling special events. Examples of such events include when the packet is normally shrinking, and the context (Contes request, context update, or a situation that includes static or full stop values) indicates the packet type. For example, RTP time The complete title block value indicating the block value and the RTp serial number block value, which is generally an unsigned integer of 16 or 32 bits. When the block value is incremented from 216-1 or 232_1, such integer is returned to ，， 〇 ”. Therefore, if X is not a power of 2, when the stop value is returned to“ 〇 ”, the remainder generated by the modulo X operator 101 in FIG. 10 will not complete the other The cycle of waiting. For example, for the modulo 12 (X == 12), when the stop value is facing, 〇 ”increases and returns via“ 〇 ”, the following sequence will be generated: • .. (65531,1 1) (65532,0) (65533,1) (65534? 2) (65535,3) (0,0) (1,1) ... The first number of each group in the brackets in the sequence is All 16-bit block values, the first number is the block value modulus 12. If the most recently reconstructed field value is 6553 1, and the remainder received is 丨, and assuming that for this example, M = 1 , Two of the last 12 alternative stop values, namely 65533 and 1, will have a remainder that matches the remainder received in the comparator 丨 23 of Fig. 12. One example of this problem is exemplary The answer is to use the received checksum and to verify that the two possible alternative values are correct.

O: \ 67 \ 67386-920808 DOC -25-564621

No, a checker 136 may couple it to (scaled or unscaled) buffer wheel out 121 to receive alternative field values that match the output of buffer 128 bis. The checker 136 may calculate a checksum for each candidate value, compare it with the received checksum, and select a candidate value whose checksum matches the received checksum. This alternative value is then provided at 111 as a reconstructed stop value. Another solution to avoid getting a short cycle of the remainder is to keep the cycle of the values sent out and adjust the received values to rebuild the remainder .... ••• (6553 1,1 1) (65532,0) (65533,1) ( 65534,2) (65535? 3) (〇? 4) (1,5) ... 5 In this case, this means that the value issued corresponds to + N, where N is 0, 4 or 8 and the experience '^ Circulate and return. This situation is exemplified in the exemplary embodiment of FIG. 15, where the title block received from the separator 33 in the specific example of the exemplary title compressor, or the alternative block in the specific example of the exemplary title decompressor The header block received by the generator 丄 25 is input to a surround detector 丨 5 丨 and the modulo X operator Π0. The surround detector 151 detects when the title stop value is near a turn. When an impending return is detected, the surround detector 5 is shifted by a circular shift register 152 which is initially loaded as shown in FIG. 15. The shifted temporary state position 154 is coupled to an adder 156 and is also fed back to a shifted register position 15 5 ′ which is then coupled to an adder 5 7. The surround detector 151 is also coupled to a discriminator 153, which has an input to its header to receive the title field value and an output to its selector 158. The discriminator 153 is distinguished by "0", the value of the high title field O: \ 67 \ 67386-920808 DOC -26- 564621 92. 8 08 before the return and the value of the low title field after the "0" return. As long as the surround detector 151 detects that the title stop value is near the return, the detector 151 keeps the discriminator 153 in the starting state. At startup, the discriminator 153 controls the selector 158 to be near the return (ie, "〇" ,, Value and "〇, and subsequent values" select the adder 156 for the low title block value, and add the adder 157 for the high title block value near the return (the value before the "0" value). When the title When the stop value is not near the return, the surround detector 151 disables the discriminator 153, in which case the discriminator 153 controls the selector 158 to select only the adder 156. Therefore, when the first return occurs, the detection The register 151 shifts the register 152, whereby the adder 156 adds 4 (from the position 154) to the remainder related to the low-title stop value starting with "0", and the adder 157 adds, ,, (from position 155) to the remainder related to the high title block value before the turn. At a turn, the detector 151 shifts the register 152 again, whereby the adder 156 adds 8 to the remainder associated with the value of the low header field starting with "0 " and the adder 157 adds 4 to and The remainder associated with the high title block value before rewinding. -For a set value of X, the fe range of possible field values at the decompressor can be increased according to the desired performance. For example, this range can be defined as -M to 2X-1-M. Therefore, for example, when M = · αχ = 16, the possible stop values may be 1 to 32. The received remainder value U8 will match the remainder of 2 alternative values in this range. The correct candidate value can be identified by the checking and inspection operations related to FIG. 13. The specific example # of FIG. 13 may, of course, be the correct candidate value selected from any number of matching candidate values generated by a range of _M to kx_UM of any size (where k = 2, 3 ...). FIG. 14 illustrates a specific example of a title decompressor that can be exemplified in FIGS.

〇: \ 67 \ 67386-920808 DOC -27- 564621 S2. Δ, οδ Exemplary operations performed. At 14 i, the compressed header block is received as a modulus X remainder value. At i 42, the range information and the recently reconstructed stop values are used to generate alternative field values. At 148, the modulus X remainder of the candidate value is determined. The received remaining values at 4 and 3 are compared with the respective alternative stop values to determine the matching alternative value. It is then determined at 44 if there are multiple matching alternatives. If so, the candidate value is subjected to a check and inspection process at 145 to determine the correct candidate value, and the backup remote value is then loaded at 149 as a reconstructed stop value. If there is only one matching candidate value at 144, this candidate value can be determined by selection and verification at M? As indicated by the virtual Λ, and the candidate value at 149 can be regarded as a new one. Constructed field values are loaded directly. --L. Lu Yue, the title compression / decompression technology of our bank's technology, 'Generally applicable to various types of standard tags' and can be compiled to suit specific applications and / or communication paths. Obviously, the above-mentioned specific examples can be easily implemented by modifying the title compressor and traditional software 'hardware' in the packet data transmitting and receiving stations, or: Erming is a preferred specific example for implementing the present invention, and should be limited by this description. The scope of the invention is limited by the scope of the patent application. Component symbol comparison table 20, 58 21, 21 '22, 22,

Column Processor Packet Compressed Header

O: \ 67 \ 67386-920808 DOC -98. 564621 23, 23, Field 24, 52 Communication application 25 Field data 26 Header information 27 Continue signal 28 Header compressor 29 Transmitter 30, 68, 69, 158 Selector 700, 703 Selector 31 No time slot 32 Other fields 33 Separator 35 Divider 36 LSB extractor 37 Add-on device 38 Check and benefit 39 Input 50 Title information 51 Column data 53 Title decompressor 54 receiver 60 time stamp decompressor 61 RTP detector 62 output

564621 ~ r 令 δ — ί a ·! 丨 i i

||. I 63 64 65 66, 702, 704 67 70 71 72 73 74 75 76, 77 78 79, 136 101 110 111 112 118 121 123 125 128 151 Traditional title decompression Extraction of time indicator Controller control signal traditional header decompression continues to execute code detector multiplier extractor time label decompression clock time label estimator store MSB extractor detector modulus X operator field reconstructor field reconstructed Range information reception remainder buffer output comparator alternative field generator buffer detector O: \ 67 \ 67386-920808 DOC -30- 153 discriminator 154, 155 shift register 156, 157 adder 301 traditional Time stamp compression 302 Traditional header compression 600 Processing path 601 Add-on device 701 Connection unit

Claims (1)

Scope of patent application: A method for generating a compressed title shelf from the original title shelf, including: obtaining the remainder generated when the value of the original title field is divided by a first predetermined number (丨 〇i); and The obtained remainder determines the value of the compressed title field. For example, if the method of applying for the first item of patent scope is applied, the step of determining includes adding the remainder to a second predetermined number by selection (38). For example, the method of applying for the second item of the patent scope includes implementing the additional step if the first predetermined number is not one of the powers of two. A method of generating a reconstructed title block (111) at the title decompressor (53), this field is intentionally related to one of the original compressed header fields (11 8) represented by the received The header field matches. The header field (118) is received at the header decompressor (53) via a packet communication path, including: according to a second number used by a header compressor (28) together with the original Title field, and identify the reconstructed title block alternatives (number of 苐 20 to generate a compressed header block corresponding to the received header (118) for transmission to the title decompressor (53) Use; and k select one of the reconstructed title field alternatives as the title rest (1 1 1). ~ Μ As in the method of applying for the fourth paragraph of the patent application park, the identification step includes The newly reconstructed title shelf (Π3) generated at the title decompressor (53) identifies the alternative value. For example, the method of claiming item 5 of the patent dry garden is based on the recently reconstructed 7 title shelf (11 3 ) And the step of identifying alternatives includes responding to the first Number and the value of the recently reconstructed title field (113), and the range of possible values for the restructured title block (111) is identified. • For the method of applying for the scope of patent application item 6, the steps of identifying a range include Decide on the position of the range related to the recently restructured title field (113). 8. If the method of applying for item 7 of the patent scope, the range includes the value of the recently restructured title field (113). The method of item 7 of the patent scope, wherein the range is an offset from the value of the most recently reconstructed title field (113). 10. The method of item 7 of the patent scope, wherein the title is determined from the most recently reconstructed title The steps of the position of the range related to the field (113) include the range of characteristics of the communication application in response to the generation of the original title field. 11 · The method of item 7 of the scope of the patent application, in which the title and the recently reconstructed title are determined The step of bit (113) related to the position of the range includes determining the position of the range in response to the characteristics of the packet communication path. The selection step includes comparing the received compressed title field (11 8) with the compressed title field associated with the reconstructed title field alternative (126), respectively, and selecting a reconstructed one The title field alternative value is used as the reconstructed title field (111), and the related compressed title field of this alternative value matches the received compressed title field (118). 13 · The method according to item 12 of the patent application, wherein the last-mentioned selection step includes using a checksum received at the header decompressor (5 3) 564621 via the packet communication path to re-value from multiple Π261Φ, Qiangli, K ^ Tong block alternative ^ The reconstructed title field alternative value, the value of the alternative value " " "Λ is the same as the received compression < I ~ shelving (118) match. U. As for the method of applying for the fourth item of the patent application park, the title fields of ^ ^ m ^ ^ ^ ^ and 2 compressed represent the original seven; the value of the higher digit of b is the second. A surplus is generated by yong 15. 15. The number of methods such as the fourth item of the patent application park. Where the first number is equal to the second Among them, the number-two is twice the number of the second method 16 as in item 4 of the scope of patent application. 17. A marker (28) for generating compression from an original title field, comprising: title compression of the title field in response to the original title, a first predetermined number divided by An input is used to receive the original title block. A device (101) is coupled to the input, and the title field is provided as a remainder when the original title field is generated; 'To output the compressed title slot with a value based on the remainder. 18.-A title decompressor (53) that generates a reconstructed title field (111) intended to match an original title bit, including-input for receiving the original title field-Jing Chen The shortened title field (118); the device () is used to identify the O: \ 67 \ 67386-920808 DOC 564621 used by a title compressor (28) / the second number together with the original title field, One of the first number of reconstructed title slot alternatives (126) to generate a compressed title field corresponding to one of the far-compressed title slots (11 8) received at the input for transmission To the title decompressor (53); and a second device (123, 128) coupled to the first device (I25) and the input (118) 'to select the reconstructed title bar One of the alternatives (126) is used as the reconstructed title field 11). I9. The title decompressor according to item 18 of the scope of patent application, wherein the first number is equal to the second number. 20. The title decompressor according to item 18 of the scope of patent application, wherein the first number is twice the second number. 564621 Patent Application No. 089122975 Chinese Graphical Replacement Page (August 1992) Time Stamp Compression 41 46 43 44 45 47 No Tradition curled up next packet? 564621 Patent Application (^ 912 ^ 975) Chinese Schematic Replacement Page (August 1992) 564621 Department B G8 repair, 丨 month and month — _ !! § | ^ 2975 patent application — Chinese graphic replacement page (August 1992) Figure 14