TWI323095B - Equalizer training method using re-encoded bits and known training sequences - Google Patents

Description

1323095 IX. Description of the Invention: [Technical Field] The present invention relates to a cellular wireless communication system, and more particularly to a wireless terminal involved in a wireless communication system processing received information to eliminate interference technology. [Prior Art] A cellular wireless communication system provides wireless communication services to many residential areas in the world. The cellular wireless communication system was originally built to serve voice communication, but is now also used to support data communication. The recognition and widespread use of the Internet has spurred demand for data communication services. Historically, data communication has been provided through wired connections, but cellular wireless users now require their wireless devices to support data communications. Many wireless users want to be able to surf the Internet, send and receive emaii, and conduct other data communication activities through their cell phones, wireless personal digital assistants, wireless laptops, and/or other wireless devices. The demand for data communication in such wireless communication systems is growing. As a result, existing wireless communication systems are being expanded/modified to meet these rapidly growing data communication needs. The Beigao wireless network includes a network infrastructure that wirelessly communicates with wireless terminals within the corresponding service coverage area. The lie infrastructure typically includes a plurality of base stations dispersed within the service coverage area. Each base station supports wireless communication within a corresponding cell (wireless cell). The base station is connected to a base station controller (BSC), and each base station controller provides services for a plurality of base stations. Each base station controller is connected to a Mobile Switching Center (MSC). Typically each base station controller is also directly or indirectly connected to the Internet. In operation, each base station communicates with a plurality of wireless terminals operating within its bee/wireless cell. The BSC' connected to the base station provides routing services for voice communication between the MSC and the serving base station (serying). The MSC routes the voice communication to another MSC or PSTN (Public Switched Telephone Network). The BSC provides routing services for data communication between service-based packet data networks, which may include or be connected to the Internet. The transmission from the base station to the wireless terminal is called forward link (downlink) transmission, and the transmission from the wireless terminal to the base station is called reverse link (uplink) transmission. The wireless link between the base station and the wireless terminal it serves typically operates in accordance with one (or more) operational criteria. These operating standards define the way in which wireless keys are assigned, keyed, serviced, and dechained. The Global System for Mobile Communications (GSM) standard is a popular cellular system standard. The GSM standard, or GSM for short, is dominant in Europe and is also widely used worldwide. GSM originally only provided voice communication services, but it has been modified to provide data communication services. General Packet Radio Service based on GSM (GpR Magic and Enhanced Data Rate Evolution (EDGE) can coexist with gsm by sharing GSM channel bandwidth, slot structure and slot timing'. GPRS and EDGE can also be used as migration paths for other standards, such as Is i36 and Pacific Digital Cellular (PDC). EDGE uses higher order modulation, octal phase shifting to increase the data rate on the 200 KHz GSM channel. Keying (8_PSK) modulation and GSM standard Gaussian Minimum Shift Keying (GMSK) modulation ^ EDGE contains (all0Wf0r) with 9 different (automatically and quickly selectable) empty inter-plane formats, ie modulation schemes have Various non-degree error control. For empty hats, according to the immediate demand of the application 'low MCS mode (MCS 1-4) adopts GMSK (low data rate) modulation ancient 1323095 • MCS mode (MCS 5-9) 8-PSK (high data rate) modulation. When the cellular phone is in the receiving mode, the GMSK/8PSK signal appears on the same channel and adjacent channels. (c〇i〇redn〇ise)e for better reception and transmission. • For cell phone information, cellular phones must eliminate these interfering signals as much as possible. Previously, techniques to eliminate these interfering signals included channel equalization of the received signals. However, existing channel equalization techniques cannot effectively eliminate the same channel and adjacent channels. Noise. Therefore, there is a need to improve the interference cancellation technique. [Invention] The present invention relates to an apparatus and method, which will be described in more detail in the following description of the drawings, the specific embodiments, and the claims. According to the present invention, there is provided a multi-branch equalizer processing module for canceling interference in a received RF burst, comprising: a first equalizer processing branch for: based on known Training sequence for training; > equalizing the received radio frequency pulses; extracting data bits from the received radio frequency pulses; and second equalizer processing branches for: based on known training sequences and re-encoding Data bits are trained, the re-formatted bits are generated by processing decoded frames; Equalizing the received radio frequency pulses; and extracting replacement data bits from the received radio frequency pulses. Preferably, in the multi-branch equalizer processing module of the present invention, the extracted data bits are generated in 1323095. The multi-branch equalizer processing module of the present invention further includes: a deinterleaver; and a channel decoding H, the channel decoder and the deinterleaver are connected to the second branch and the second equalized processing branch The combination of the channel solution (4) and the rider interleaver decodes the frame including the extracted data bit and decodes the replacement ugly line including at least the partial replacement data bit. In the multi-branch equalizer processing module of the present invention, in the replacement building, the frame and the replacement medium are preferably, in the multi-branch equalizer processing module of the present invention, a data frame. . Preferably, in the multi-branch equalizer processing module of the present invention, the first equalizer processing branch includes: a 1-component and Q-component interference canceling portion; and a decision feedback equalizer portion; The processor processing branches include: • I component and Q component interference cancellation sections; and a linear equalizer section. Carrier = Ground: The radio frequency pulse of the multi-branch equalizer processing module of the present invention includes a minimum frequency shift keying (_) symbol and an 8PSK interference symbol. The multi-branch equalization reduction module of the present invention further includes: a 1323095 processor; an interleaver, the interleaver and the encoder are combined to: process the decoding block to generate a re-encoded data bit; And providing a training signal to the second equalizer processing branch, wherein the training signal is used to train the second equalizer processing branch based on the known training sequence and the re-encoded data bit. According to an aspect of the present invention, a wireless terminal includes: a radio frequency front end for receiving a radio frequency pulse; a baseband processor connected to the radio frequency front end, the baseband processing is recorded from the radio frequency pulse zone; The multi-branch equalizer and the baseband processing H are connected to the branching equalizer processing module, and the processing module further includes: the first gamma-axis edge is trained based on the known training sequence; Kawasumi performs equalization processing; extracts from the connected (four) riding pro- rush: ##位位; and the second equalizer processing branch for: training based on the rush containing the known training phase and the re-encoded (10) bit And the at least part of the re-encoding and rushing is subjected to equalization processing by the radio frequency pulse received by the 2-coded pulse; and the generated data bit is extracted from the received radio frequency pulse;

Deinterleaving the data block; decoding a frame from the data block; and encoding the data to generate at least partially rewritten data blocks; and arranging to > knife re-encoded data blocks Processing to generate at least partially re-encoded pulses. _ kit纟 The wireless terminal towel of the present invention, the frame is a voice frame or a data file. Preferably, in the wireless terminal of the present invention, the first equalizer processing branch includes: a 1-component and Q-component interference canceling portion; and a decision feedback equalizer portion; the second equalizer processing branch includes: 1 component and Q component interference cancellation section; and # linear equalizer section. Preferably, in the wireless terminal of the present invention, the to-be-divided coded pulses used to train the second equalizer processing branch are fully re-encoded. Preferably, in the wireless terminal of the present invention, the radio frequency pulses include a minimum frequency shift keying (GMSK) symbol and an octal phase shift keying (8PSK) interference symbol carrying data bits. Preferably, the wireless terminal of the present invention further comprises: an encoder; a parent errorer 'the combination of the interleaver and the encoder is used for re-encoding the shell material to generate an at least partially re-encoded data block And interleaving at least partially re-encoded data blocks to generate at least partially re-encoded pulses. A method for homogenizing a receive_radio frequency pulse according to the aspect of the invention includes: receiving a radio frequency pulse; decoding a known training sequence from the received radio frequency pulse; appealing based on the decoded known Naxusan First-etc. (4) training; processing the radio frequency pulse received by the branch equalization with the first equalizer; deinterleaving the radio frequency pulse; encoding the radio frequency pulse to obtain the extracted soft sample; decoding the data bit from the extracted soft sample Re-encoding the data bit to generate at least partially re-encoded soft samples; interleaving the at least partially re-encoded soft samples to generate at least a partial re-matrix; re-reading the received from memory The radio frequency pulse processes the branch to the second equalizer; 77 trains the second equalizer to process the branch using the at least partially recoded pulse; and equalizes the received RF pulse in the memory with the second equalizer Processing; deinterleaving the radio frequency pulses; decoding the radio frequency pulses to obtain replacement soft samples; and 12 1323095 from the replacement soft fetch Replace decoded data bit. Preferably, in the method of the present invention, the method includes a voice dump or a data frame. Preferably, in the method of the present invention, the first equalization cleft branch processes the first set of 4 radio frequency pulses to generate a second set of 4 radio frequencies for processing the second equalized H processing branch processing memory The pulses are generated to generate replacement data bits, wherein the first set of four RF pulses precedes the second set of four RF pulses. Preferably, in the method+ of the present invention, the first equalizer processing branch comprises a 4-tap (4 nal) pre-chopper and a muscle guilt: the second equalizer processing branch comprises 7 拙 (7- Tap) linear equalizer (lE). Preferably, in the method of the invention, the radio frequency pulse comprises an interference symbol carrying a data bit GMSK symbol and 8PSK. The following specific implementations and guidelines will make the other features of this issue more clear. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS The preferred embodiments of the present invention are illustrated in the drawings, and the same reference numerals are used in the drawings. Gaussian minimum _ difficult (GMSK) 镰 _ real area towel single-way round into the two-way output secret. This mode is a virtual single-channel transmission 2-way connection. Multi-antenna interference cancellation technology _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ This bribe is a multi-branch equalizer processing module capable of eliminating interfering signals in the received radio frequency pulses. The multiple = 13 1323095 equalizer processing module includes a plurality of equalizer processing branches. An equalizer processing branch can train based on known training sequences and equalize the received RF pulses. The resulting results are then further processed and used to train the second equalizer to process the branches. The second equalizer processing branch then equalizes the received RF pulses and generates an output based on the age processing of the interference 彳§. Thus, the processing of riding the received radio frequency pulse is improved. 1 is a partial schematic diagram of a cellular wireless communication system 100 supporting wireless terminal communication in accordance with an embodiment of the present invention. The cellular radio communication system 100 includes a mobile switching center (msc) 101'gprs service support node/EDGE service support node (sgsn/sesn) 102 'base station controllers (MSC) 152 and 154, base stations 1〇3, 1〇4, 1〇5 and i(10). The SGSN/SESN 1G2 is connected to the Internet 114 via a GPRS Clearing Support Node (GGSN) 112. The conventional voice terminal 121 is connected to a pstn (Public Switched Telephone Network) 11A. A voice (IP voice) terminal Π3 and a personal computer 125 transmitted via the Internet are connected to the Internet 114. The MSC 101 is connected to the PSTN 110. Each of the base stations 103-106 serves a cellular/wireless cell, and each base station supports wireless communication within the cellular/wireless cell it serves. The wireless link, including the forward link and the reverse link, supports wireless communication between the base station and the wireless terminal it serves. These two wireless links will generate the same channel (co-chgj^ei) and adjacent channel (a (jjacent cjlannei) signals, which will appear as colored or white noise. As mentioned above, these noises may interfere with the expected interest. Signals. Accordingly, the present invention provides a technique for eliminating interference in such poor signal-to-noise ratio (SNR) or low signal-to-interference ratio (SIR) environments. These wireless links can support digital data communication, !p voice communication, and others. Digitally more than 14 1323095 'Media communication. The cellular wireless communication system 100 is backward compatible in supporting analog communication. Therefore, the cellular wireless communication system can support the Global System for Mobile Communications (GSM) standard and its extension. Enhanced Data Rate Evolution (EDGE). Cellular•Wireless Communication System 1(8) can also support GSM extended General Packet Radio Service (GPRS). The invention is also applicable to other standards such as TDMA standard, CDMA standard, etc. Usually, The present invention can be applied to digital communication technology to solve the problem of communication interference and elimination β Φ wireless terminals 116, 118, 122, 124, 126, 128 and 130 are coupled to the cellular wireless communication system 100 via a wireless link and base stations 103-106. As shown, the wireless terminal can include cellular mobile telephones 116 and 118, laptops 12 and 122, and a desktop Computers 124 and 126, data terminals 128 and 130. However, the cellular wireless communication system also supports communication with other types of wireless terminals. As is well known, laptop computers, desktop computers, la and 126, data terminals Devices such as 128 and 13 蜂, bee-write mobile phones 116 and 118 are capable of "surfing, transmitting, and receiving data communications such as emails" on the Internet 114 to send and receive files, as well as performing other data operations. Many of the billing operations require a high download data transfer rate, and the upload data transfer rate requirement is λ. Therefore, some or all of the wireless terminals can support the EDGE operating standard. These wireless terminals 116·13 also support The gsm standard may also support the GPRS standard. Figure 2 is a schematic block diagram of the wireless terminal 200. The wireless terminal 2 in Figure 2 includes a radio frequency transceiver H 202, a digital processing unit 204, and various other components within the chassis. The digital processing component 204 includes two main functional components: physical layer processing, speech codec 15 1323095 (CC) DEC), baseband encoding/decomposing device (CODEC) Function block 206; protocol processing, human interface function block 208. Digital signal processor (DSp) is physical layer processing, speech codec (CODEC), baseband encoder/decoder (c〇DEC) function block 2〇6 The main components of the microprocessor, such as the RJSC processor, are the main components of the protocol processing, human interface function block 208. The DSP may also be referred to as a wireless interface processor. The *rjsc processing state may be referred to as a system processor. However, these naming conventions should not be considered as limitations on the functionality of these components. The RF transceiver 202 is coupled to an antenna 203, a digital processing component 204, and a battery 224, wherein the battery 224 provides power to all components of the wireless terminal. The physical layer processing, voice codec/decoder (CODEC), baseband encoder/decoder (CODEC) function block 2〇6 is connected to the protocol processing, the human interface function block 208, the microphone 226, and the speaker 228. The protocol processing, human interface function block 208 is connected to various components including, but not limited to, a personal computer/data terminal device interface 210, a keyboard 212, a user identification card (SIM card) 213, a camera 214, and a flash memory. 216. Static memory (SRAM) 218, liquid crystal display # (LCD) 220, and light emitting diode (LED) 222. When there are cameras 214 and LCD 220, these components support still images and/or moving images. Thus, the wireless terminal 2 shown in FIG. 2 can support video and audio services through the Bee Heights network. Figure 3 is a pictorial diagram of the general structure of a GSM frame and the manner in which the GSM frame carries the data block. A GSM frame with a duration of 20 milliseconds (ms) is divided into four quarters. Each - quarter frame includes 8 time slots (time slots 0-7). Each time slot lasts approximately 625 microseconds (# s), including the left, right, and middle yards. The left and right RF pulses on the time slot carry the data, while the intermediate code is the training sequence. 1323095 According to the supported modulation and coding scheme mode, the radio frequency pulses on the 4 time slots of the GSM frame carry a segmented RLC (wireless key control) block, a complete block or two RLC blocks. For example, the data block A is subdivided by the time slot 四 of the quarter frame i, the time slot 0 of the quarter frame 2, the time slot 四 of the quarter frame 3, and the time slot of the quarter frame 4

The time slot 1 of quarter frame 3 and the time slot 1 of quarter frame 4 are carried. Each group of ♦ time slots, that is, the Mcs mode of each quarter of the time slot η, for the GSM building,

Between them, the MCS mode is different, such as his 8 mode for each quarter of the time slot, and every - quarter red time slot 丨 _7 #Mcs mode may be different. The RLC block can carry voice data or other materials. Figure 4 depicts the various steps of mapping data into radio frequency pulses. The data is initially unencoded: it may have a block header. The block coding operation performs external coding of the data block and supports error detection/correction of the data block. The external encoding operation usually uses the loop redundancy code check (CRC) or the Farr code (FireC() de) e diagram to show the outer touch of the added data bit and / Wei code lai (BCS). It is attached to the capital. Under the (3) coding scheme, the header and the data are compiled together using block coding and convolutional coding; under the non-(3) coding scheme, the header and the data information are usually separately coded. , '·, and the poor material bit of the touch bin. The odds that the Falcon's error was passed were only 2,40 times. The legal code is to add the pure error detection ability of the riding header to the header code to be undetected. After the block coding adds the spare time for the error detection, the redundancy is caused by the transmission error caused by the wireless channel. (10) The error or coding scheme is based on convolutional coding. Some redundant bits generated by the β-convolution encoder can be used for puncture operations prior to transmission. The k-hole operation increases the rate of convolutional coding and reduces the amount of spare material for each transmission. The hole also reduces the bandwidth requirement to make the convolutionally encoded signal suitable for the available channel bitstream. The convolutional coding bits are passed to the interleaver, which interleaves the various bit streams into four pulses. • Figure 5 is a schematic block diagram of the steps associated with recovering a data block from a radio frequency pulse. Usually 1 data block consists of 4 RF pulses. These pulses are received and processed. When all four radio bursts are received, the four radio frequency pulses are combined to form an encoded data block. Subsequently, the coded data block is decoded by a depunc (four) (if needed) according to an internal decoding scheme, and then decoded according to an external programming scheme. The decoded data block includes ==: the root material and the _coded manner, It is possible to perform partial decoding. Figure 6 is a recovery from the ship's voice t-smoke: a splicing _ schematic block diagram. The diligent process is similar to that of Figure 5. Typically, a 2 〇 millisecond voice pad is transmitted. In the basin, ^ _ _ part of the shirt, fine delivery, 彳 _ minutes in L = rush to transmit. Figure 6 shows - group of 4 RF pulses, this __ paste n ^ sword coffee. Its towel, The property of the pot η, the first half of the voice _, is encoded and interleaved to the four RF pulses. “4 After the touch processing, the code block _ the general process, the material flow contains the outline butterfly _ (9) (four) The first half of the speech frame η in memory 18 1323095 can be combined with the second half of the speech frame n to generate an effective speech frame η related material. Figure 7 shows the speech frame η Re-encoding of the data, which will generate at least partially re-encoded data pulses, the re-encoded data The rush can be used to train the second equalizer to process the branch. As mentioned earlier, the half of the speech age recovered from the treasury-group blasting is combined with the voice halved half recovered from the current set of RF pulses. The speech speech is generated by using a loop of several codes to verify and correct the speech frame to generate an effective speech ♦ Φ 贞. The effective speech is re-encoded. However, only the re-encoded speech 贞The latter half is used to partially reconstruct the RF pulse. The latter half of the recoded speech block η can be knife-cut and interleaved to generate a partially encoded RF pulse. Since the second half of the voiced duck state has not been processed, these The surface pulse is only partially re-encoded. Since the speech Φ贞η1 is not confirmed, the first half of the re-encoded speech rectification cannot and is not used to recreate the radio frequency pulse. According to an implementation of the present invention For example, based on the partially re-encoded RF pulse of the speech structure η, combined with the known training sequence, the second equalizer can be trained to process the branch. Figure 8Α and Figure 8Β疋A flowchart of the radio terminal 2 receiving and processing radio frequency pulses. The operation of the figure and FIG. 8 corresponds to a single radio frequency pulse on the corresponding time slot of the gsm block, and the m-band baseband processor and the equalizer processing branch module are executed. These operations are performed when the operation of the above-mentioned components is performed. However, the processing functions between these components may be different without departing from the scope of the present invention. The processing flow starts from the RF front end receiving the radio frequency pulse on the corresponding time slot of the GSM frame (step, , and thereafter, the RF front end converts the RF pulse into a baseband signal

丄J 丄J t

No. (step 8〇4). After the conversion is complete, the RF front end sends an interrupt signal to the baseband processor (vi. 806). Thus, as shown in the figure, the RF front end performs step 8 〇 fiber. 'Next, the basics of the (4) base with money (step is called - a baseband base city touch. Wei posts with the number of materials, the baseband processor ^ 810 (blind detecti〇n) 〇 ^ New Zealand blind The check is determined by the record 崎 _ _ _ _ _, according to the deletion criteria, the modulation mode can be either "Gaussian minimum frequency GMSK modulation" or octal phase shift keying (8 psK) modulation. After the baseband processor determines the modulation assignment, it selects the appropriate processing branch for processing based on the shot_axis equation (step 812). - For GMSK modulation, in step 8M, the baseband processor derotates and frequency the baseband signal; kJE. In step 816, the baseband processor performs a pulse=rate evaluation of the baseband signal. In step 820 (see Figure 8B, page break arrow A), the baseband processor is then evaluated for timing, channel, noise, and (SNR). Subsequently, the base π processor performs an automatic gain control (AGC) loop calculation (10)c as discussed in (4) ^step 822). The baseband processor then performs a soft decision on the baseband signal to determine the number of smokes (/step 824). After step 824, at step jam, the baseband processor performs a matched filtering operation of the baseband signal. v steps 808-826 are referred to as pre-equalization processing operations. After the baseband processor performs some pre-equalization processing operations on the baseband signal, the processed silk signal is generated. After completing these pre-equalization processes, the baseband awards the equalizer to send a command. 20 1323095 • The equalization II module running on the multi-branch equalizer will be discussed further in Figure 9. After receiving the command, the equalizer module prepares to equalize the processed baseband signal based on the difficult mode (GMSK or κ). In step 828, the equalizer module receives the processed signals, settings, and/or parameters from the baseband processor and performs maximum likelihood sequence estimation (MLSE) equalization on the left side of the baseband signal. As shown in Figure 3 above, each RF pulse consists of the left side of the data, the middle code, and the right side of the data. Typically, in step (10), the equalizer module equalizes the left side of the radio frequency pulse to generate the left soft decision. Then in step ♦ 830, the equalizer module equalizes the right side of the processed baseband signal. This equalization operation produces multiple soft decisions associated with the right side. Typically, the equalization of the pulses is based on a known training sequence in the pulse. However, in embodiments of the present invention, re-encoding or partial re-encoding data may be utilized to improve equalization processing. This can be in the form of iterative processing, where the 'first branch performs pulse equalization on the RF burst and the second module performs quadratic equalization based on the result of the first branch equalization process. The 'equalizer module' then sends an interrupt signal to the baseband processing indicating that the radio frequency pulse has been completed. The baseband processor then receives soft decisions from the equalizer module. At step 832, the baseband processor determines the average phase on both the left and right sides based on soft decisions from the equalizer module. In step 8; 36, the baseband processor is based on a soft evaluation and frequency tracking from the equalizer module. Here, the operations of the step surface and the step 836 are referred to as "equalization processing". After step 836, processing of the radio frequency pulse has been completed. Returning to Fig. 8A, when the result of the blind detection in step 810 is 8pSK modulation, the baseband processor and the equalizer module select the processing branch on the right. First, in step 818, the baseband 21 1323095' processor performs derotation and frequency correction on the baseband signal. In a subsequent step 820, the baseband processor performs a pulse power evaluation of the radio frequency pulse "following the page-by-page connection arrow B with reference to FIG. 8B." In step 840, the baseband processor performs timing, channel, noise, and signal-to-noise ratio. (SNR) assessment. Next, in step 842, the baseband processor performs an AGC loop calculation of the baseband signal. Next, in step 844, the baseband processor calculates a decision feedback equalizer (DFE) coefficient, which is used by the intermediateizer module in step 844. These processes for generating these coefficients will be explained in more detail later. Figure 9 and subsequent diagrams discuss these decisions using a multi-branch equalizer. Next, in step 846, the baseband processor performs a pre-equalization operation on the radio frequency pulses. Finally, in step 848, the baseband processor breaks the soft decision scale factor for the radio frequency pulse. Here, steps 818-848 performed by the baseband processor 3 are referred to as "pre-equalizer processing, operation of the 8PSK modulated baseband signal. After the step is completed, the baseband processor sends commands to the equalizer module for equalization processing. Subsequent baseband signal. After receiving the command from the baseband processor, the equalizer module receives the pre-equalized baseband signal, setting, sum, or parameter from the baseband processor, and starts the pre-equalization process. The baseband is equalized. The equalization n-module is used to determine the baseband signal of the 8PSK modulated pre-equalization process in step 850. In the illustrated embodiment, the state value is used. The equalizer module uses the maximum posterior probability (viewing)/equalization method. Then, in step 852, the equalizer module uses the touch equalization method to equalize the left and right sides of the baseband signal of the pre-equalizer to generate The soft decision of the baseband signal after the processing. After the step 8M is completed, the 'equalizer module sends an interrupt signal to the baseband processor, indicating that the equalization processing of the baseband signal has been completed. 22 丄323095, and then 'baseband processing is received Miscellaneous (four) soft turn. In the next step, the base f determines the left and right rate of the processed baseband signal by the step m: t. Finally, in step 836, the baseband processor performs the frequency of the baseband signal. H. The operation of the step paste and the 836 is called the equalization post-processing operation. The step secret frequency Wei Wei shooting ship rushing (four) thief has been completed. The above processing is difficult to describe the steps of recovering the data block from the RF pulse. Although Figure 8A And the operations in Figure 8B can be performed with specific elements of the wireless terminal, == partitioning can be performed with different components, as in other embodiments, the system can be implemented with a baseband processor or system processor. In addition, in another embodiment, the 'decoding operation can be performed by the baseband processor processor. The vertical diagram 9 is the structure of the multi-branch equalizer processing module _ according to an embodiment of the present invention. Block diagram, according to an embodiment of the present invention, the processing module 9 (10) can be used to perform single day 2 interference cancellation (SAIC). There are two types of equalization methods: node detection (10) and blind interference cancellation (10). Inventive aspects, _ shame method 19 The read 'can be' can be a hardware component, or a software component executed by the processor as shown in Fig. 2, 2〇6 and 2〇8, or a hardware component and a software combination. The branch equalizer processing mode The group_ includes a first-equalizer processing branch 9()2 and a second equalizer processing branch 9 the heart-reversing module 906 receives the in-phase component (1) and the quadrature component (9) of the baseband pulse. The baseband impulse corresponds to _3.7 The anti-rotation module samples the received I and Q pulses in a reversed 'generating chirp' and Q-pulsing samples. In one embodiment, the first equalizer processing branch 902 includes a pulse equalizer. According to an embodiment of the invention, these pulse samples are then equalized, and then combined with other samples to form a data group, such as the rainbow 23 1323095 * group. In some operating situations, in addition to the pulse level equalization, a second class can be performed. Theizer handles the iterative processing of branches. The pulse equalizer 'includes the I and Q finite impulse response (FIR) filters 908 and 910 and the Minimum Least Squares Estimation (MLSE) 912, for each of the slave anti-rotation modules 9〇6 The received pulse is processed. The training module 913 trains the modules using known training sequences (TS) in the intermediate code of each received pulse. Optionally, these components are capable of training on multiple pulses. The first equalizer processing branch 9〇2 generates a soft decision, and the multi-record decision represents each negative bit before decoding. The mother soft samples are provided to a deinterleaver 914 which deinterleaves the soft samples and provides the deinterleaved soft samples to the channel decoder 916. Channel decoder 916 decodes the data frame from soft samples (i.e., multiple soft samples representing each data bit are decoded by the channel decoder to generate hard bits after decoding). The re-encoder 918 acknowledges and re-encodes the data duck decoded by the channel decoder 916 to generate re-encoded data bits. The interleaver 92 receives the re-encoded data bits • το to generate a re-encoded data pulse. The re-encoded data pulse and the known training sequence can then be used to train the second equalizer processing branch 920. The second equalizer processing branch 904 includes a buffer 922, a finite pulse chopper (FIR) 924 and 926. The buffer 922 is capable of storing a plurality of pulses into the memory. The training mode 928 can be trained with known training sequences and at least partially re-encoded pulse pairs and q-choppers 24 and 926. Thus, the second equalizer processing branch trains the (9) q radio frequency chopper with at least partially encoded = batting and known training sequences. This improves the pulse of the buffer milk treatment. After the chopper is trained, 24 1323095 is used to process the stored pro-pulse. Coronation (4) combines the results obtained. This produces a soft-sampling _te) which is provided to the de-interlacing (4) 4 and channel decoder 916 to generate replacement data bits. The first processing of the multi-branch equalizer shown in Figure 9 is described in detail in Figure 1 because the first processing branch can train the front weft 908 with _ 4 _ and _ because there are only 26 training symbols. Train a 4-tap anti-carrier wave.

Figure 11 depicts the second processing branch of the multi-branch equalizer shown in Figure 9 in more detail. After the channel is decoded, the data is re-encoded and used to train the 7-tap le924*926. The second processing branch selects the linear equalizer (4) because (4) interleaving (int_me and speech_off re-encoded bits can only provide half a pulse (even if it is a data bit). DFEs require defamation to provide a coherent turn. The pie, le is fresher than DFE (MLSE). Other embodiments using fully re-encoded bits can use DFE for the second processing branch instead of LE. The following discussion will describe the indirect training method in more detail, The training method is based on the least square channel estimation (LS_CE), similar to that used in EDGE. First, the channel is estimated using the training sequence, and then the pre-filtering and MLSE parameters are calculated as if they were feedforward or feedback filtering of the DFE. One problem with indirect training is that CE (channel estimation) is poor 'because SAIC (single antenna interference cancellation) usually operates at low signal-to-interference ratio (sir). CE errors are expanded when calculating filter coefficients (pr〇pagates) In Figure 10, the signal model at the MLSE input can be thought of as ISI channel plus noise. Rice 5 and DFE feedback filtering, wave β impulse response is {b(0), b(1),...,b(Lb-l)} The goal of training is The pre-filtering coefficient 25 丄 milk ϋ 95 ' {f 丨 (0), ... fi (Lrl), f2 (0), ... f2 (Lrl)} and MLSE shall be obtained from the given training symbols and the received signals. Parameter b. Based on the above mode, the noise at the MLSE input is: ring =|/ga...Η +1/2 (04 ice 0, tree, ·) where 'Xl and & respectively are the inverse rotation outputs I and Q, s Is the training symbol, d is the system delay. In vector form: n(k) .«(A: + l) w two n{k + ri) .x2(k + d-~i^ + i^ ~ • x2 (k + d + \^L/+l) , x2(k + d + N - L^+1) xx (k + d)...xx(k + d-Lf +1) χ2 (Jc + . (A : + ί/ +1) .·· xx{k + d + \-Lf -\Λ) x2{k + d ^{k^d + Ny-x^k + d^N-Lj+^x^k + d + ^y Λ(〇) · — 1) Λ(9) S(k) ·.. i(A:-Xi+l) s(k + \) ··· i(A: + l-Z4 +1) H 〇) .Λ -l) _s(k + N)--s(k + N-Lb+l) KLb-ll For convenience, the vector is represented in bold lowercase letters, and the uppercase letters in bold indicate the matrix, above equation Expressed as: n^Xf-Sb The discriminant of the thief is to find £ and b to minimize the input noise of the hall, ie min 丨| n || 2 〇 because the number of symbols is _, so 丨And b Optimization of noise sensitive. The following discussion discusses the use of suboptimal aspects to reduce the estimate to the pre-chopped f. 26 2. The interaction between the round 4 (Xf) and the training symbol can be entered with the fox π. It is represented by the 1 magic channel of the second. Therefore, b can be represented by f. The pre-filtered output uses LS CE ' b as the channel estimate: b = s + xf where 0 is not pseudo-inverse (pSecd〇-inverse). Substituting the above formula will get the maximum value of the function, and get: 匪丨1 Xf-SSlf 丨丨2=min 丨丨(I_ss'Xf 丨丨2=minf, Af calls its t A = X'(I-SS^)X , 〇, is a transposition operation. To avoid trivial solutions, constraints are applied. Two commonly used constrained green small integer paradigms (touch (4)) constraints and linear constraints. When a paradigm i constraint is used, the optimal solution corresponds to The eigenvector A of the smallest eigenvalue: f = eigvec(A) The linear constraint can be chosen for f. For example, we can fix the ith element of _ to 1. In other words, the ith of the MLSE channel b When the tap is the i-th row vector of (s+x), the linear constraint is: cf=l The corresponding optimal solution is:

f=AV Usually linear constraints are better than minimum integer paradigm constraints. In the linear constraint, if the first tap is selected as 1 ′, the minimum discriminant is equal to the DFE discriminant. Diagonal loading also contributes to the high SIR range. Figure 11 depicts the second processing branch of the multi-branch equalizer shown in Figure 9 in more detail. 27 ^23095 The 'after channel decoding' data is re-encoded and used to train 7 fine l and 926. The line feeder (10) is selected for the second processing branch because of the interleaving. The re-encoded bit with voice can provide only half a pulse (even if it is a poor bit). DFEs need to provide a consistent sampling of the feedback chopper. In addition, le is more than DFE (MLSE) ffi. Other implementations of the lion's fully re-encoded bits may be used to devise the DFE of the second processing branch without the use of the LEe diagram shown in Figure 12 for the equalization process of the _frequency_pulse for equalization processing. Including: Step 12: Receive a plurality of RF pulses. This ship is reversed in step 1202. In step 12〇4, the radio frequency pulse is processed by the first equalizer processing branch as shown in FIG. In step 12-6, the first branch processing is trained with a known training sequence. The received pulses can be provided to the first processing branch and the second processing branch. In the second equalization H processing branch, there is another memory that buffers the thief, which is used to store the received RF pulse and wait for further processing. In step 12-8, the first equalizer processing branch equalizes the received pulses by filters trained based on known training sequences. • The equalized RF pulse produces a series of sampling or soft decisions. These sampling or soft decisions are deinterlaced in the step coffee. In step 1212, these samples or soft decisions are decoded to produce extracted data bits. In step 1214, the data frame is decoded from the extracted data bits. In step 1216, the data frame is re-encoded to generate a re-encoded data bit. For speech frames, the data of the current RF burst group needs to be combined with the last radio burst group to generate a valid speech frame. These speech frames are then re-encoded to generate re-encoded data bits. In step 1218, the re-encoded data bits are interleaved to generate a re-encoded data pulse. These re-encoded data pulses can include 28 1323095 • Partially recoded bits when applied to speech frames. In step 1220, the radio frequency pulse is recovered from the memory with the second processing branch. This may include restoring one or more pulses processed by the second equalizer processing branch. In step, these re-encoded data bits are provided as signals to train the second equalizer to process the branches. Step 1224, processing the branch equalization of the radio frequency pulse stored in the register with the second equalizer. The equalization H processing branch is not only trained by the known training sequence, but also to: > The partially re-encoded data bit generated by the original output of the channel decoder is trained. Not only the known training sequences but also the re-encoded data bits are used in order to better train the second equalizer to process the branches and better balance, so that the second equalizer processing branches provide better than the first one. The processor handles the branch's better output. The second equalizer processes the branch to generate an alternate soft decision, and at step 1226, the record decision is deinterleaved; in step 1228, the soft decisions are decoded to generate a replacement data frame in step 123. In the noise limiting scheme, the interference cancellation for a single antenna is worse than that of a conventional receiver. In addition, due to the short pre-filter length, long-delay channels (such as mountainous terrain) will also result in large performance degradation. To solve this problem, a switching function has been added to enable interactive single antenna interference processing. The money can be based on any combination of SNR, colored noise acquisition and channel attribute detectors. In summary, in summary, the present invention provides a multi-branch equalizer processing module that eliminates interference from received RF pulses. The multi-branch equalizer processing module includes a first equalizer processing knife and a second equalizer processing branch. The processing branch can train and equalize the received radio frequency pulses based on the known training sequence. This results in soft or soft decisions, after which these soft or soft decisions are converted into data bits. These soft samples are processed by deinterleaving and channel 2 code processing, where the deinterleaver and channel decoding bins are able to generate decoded bits of poor bits from the soft samples. The re-encoder re-encodes the decoded duck to generate a re-handled or to-part re-encoded data bit. The interleaver then processes the at least partially re-encoded bucket 7G to generate an at least partially encoded pulse. The second equalizer processing branch can store the received RF pulse by using the at least partially re-encoded data bit to the linear equalizer in the second equalization processor branch, after the linear equalizer is trained The second equalization H processing branch recovers and equalizes these RF pulses. This will result in a soft-sampling or soft-decision of the two alternatives, which are then converted to the j-bits. These alternate soft samples are processed by a deinterleaver and a channel decoder, where the deinterleaver and channel decoder combination is capable of generating an alternative solution of data bits from the replacement soft samples. At the job, it is possible to eliminate interference and process the received RF pulse more accurately. A person of ordinary skill in the art would like to do so, the term "substantially, or "about", as used in k, provides an industry-acceptable tolerance for the corresponding term. This industry can accept The tolerances are from less than 1% to 20%, and correspond to, but not limited to, component values, • integrated circuit processing fluctuations, temperature fluctuations, rise and fall times, and/or thermal noise. It is recognized that the term "operably connected", as used herein, includes "directly connected and indirectly connected through another element, element, circuit, or module," "indirectly connected, intervening element, element, circuit or The module does not change the information of the signal, but can adjust its current level, voltage level and/or power level. As will be appreciated by those of ordinary skill in the art, inferred connections (ie, one element is connected to another according to inference) - Component) includes direct and indirect connections between two components using the same method as "operably connected." As will be appreciated by those of ordinary skill in the art, the term "compared to the results of 30 1323095" as in this shot energy (four), refers to the comparison of two or more components, providing a desired relationship. For example, 'when the desired relationship number ^ is greater than the amplitude of the signal 2>, an advantageous comparison result can be made when the amplitude of the signal i is larger than the amplitude of the signal 2嶋_number: is smaller than the amplitude of the signal 1. The description of the preferred embodiment of the present invention is for the purpose of illustration and description. These real complements are not poor ones, which means that the precise shape of the publication is open.

Various modifications and variations of these embodiments are possible in light of the teachings of the invention. The embodiment was chosen and described in order to clarify the principles of the invention and the practice of the present invention so that the present invention can be utilized in various embodiments and various modifications can be made to the invention. The range is defined by the paste requirements of the present invention and its contents. In addition, it is to be understood that various changes, substitutions and substitutions can be made without departing from the spirit and scope of the invention. BRIEF DESCRIPTION OF THE DRAWINGS FIG. 1 is a partial schematic diagram of a bee-writing communication terminal supporting a secret communication according to the present invention; FIG. 2 is a schematic block diagram of a wireless terminal constructed according to the present invention; FIG. 3 is a GSM frame. Figure 4 is a schematic block diagram showing the structure of a downlink transmission; Figure 5 is a schematic block diagram showing the steps of recovering a data block from a series of radio frequency pulses; Figure 6 is a schematic block diagram of the steps of recovering data blocks from a series of radio frequency pulses; A schematic block diagram of the steps associated with recovering speech data in a series of radio frequency pulses; Figure 7 is a schematic block diagram of the steps associated with recovering pulses from a data or speech frame; 31 1323095 Figures 8A and 8B: Wireless terminal receiving and processing radio frequency pulses FIG. 9 is a schematic structural diagram of a multi-branch pulse equalization element according to an embodiment of the present invention; FIG. 1 is a schematic block diagram of a pulse equalization element according to an embodiment of the present invention; FIG. 11 is a flowchart of the present invention. A schematic block diagram of a pulse equalization element of the example; Figure 12 is a flow chart of the operation of an embodiment of the present invention. [Main component symbol description] 100 cellular radio communication system 101 mobile switching center (MSC) 102 GPRS service support node / edge service support node (sgsn/sesn) 103 104, 105, 1〇6 base station u〇PSTN (Public Switched Telephone Network) 112GPRS Gateway Support Node (GGSN) 114 Internet 120, 122 Laptop 123 Voice (IP Voice) Terminal 125 Personal Computer 152, 154 Base Station Controller (MSC) 202 RF Transceiver 204 Digital Processing Element 208 Human Machine Interface Function Block 212 Keyboard 214 Camera 218 Static Memory (SRAM) 222 Light Emitting Diode (LED) 226 Microphone 116, 118 Cellular Mobile Phone 121 Voice Terminal 124, 126 Desktop Computer 128, 130 Data Terminal 200 Wireless Terminal 203 Antenna 206 Baseband Codec/decoder (CODEC) function block 210 PC/data terminal device interface 213 user identification card (SIM card) 埠 216 flash memory 220 liquid crystal display (LCD) 224 battery 32 1323095 • 228 speaker H _ multi-branch Equalizer processing module 902 first equalizer processing branch 904 second equalizer processing branch 906 anti-rotation module • 908, 910 I and Q finite impulse response (FIR) filter 912 minimum Square Least Squares Estimation (MLSE) Equalizer 913 Training Module 914 Deinterleaver Spring 916 Channel Decoder 918 Recoder 920 Interleaver 922 Buffer 924, 926 I and Q Finite Pulse Filter (FIR 928 training module 930 adder

33

Claims (1)

1323095 • X. Patent application scope: 1. A multi-branch equalizer processing module 'is used to eliminate interference in the received RF pulses, including: • First equalizer processing branch for: Knowing training sequences for training; equalizing processing of the received radio frequency pulses; and extracting data bits from the received radio frequency pulses; ^ second equalizer processing branch for: based on known Training sequences and re-encoded data bits are generated, the re-encoded data bits are generated by processing decoded frames; equalizing processing of the received radio frequency pulses; and extracting replacement data from the received radio frequency pulses Bit. The multi-branch equalizer processing module of claim 1, wherein the decoded frame is generated by the extracted data bit. #3. The multi-branch equalizer processing module of claim i, further comprising: a deinterleaver; and a channel decoder, the channel decoder and the deinterleaver connected to the first equalization a branch and a first equalization n processing branch, the combination of the channel solver and the deinterleaver for: decoding a row comprising the extracted data bits; and including at least partial replacement The replacement duck of the bit is decoded. 4. The multi-branch equalizer processing module of claim 1, wherein the 34 1323095 'frame and replacement frame are voice frames. 5. A wireless terminal comprising: a radio frequency front end for receiving a radio frequency pulse; a baseband processor coupled to the radio frequency terminal, the baseband processor and the radio frequency front end for generating a baseband signal from the radio frequency pulse; and a baseband processor a connected multi-branch equalizer processing module, the multi-branch equalizer processing module further comprising: a ##化器面' for receiving a baseband signal and an output soft decision from the baseband processor; a branch, configured to: perform training based on a known training sequence; perform equalization processing on the received radio frequency pulse; extract a data bit from the received radio frequency pulse; and process the branch by the first equalizer And: performing training based on a pulse age including a known phase and a re-encoded at least partially re-encoded pulse, the at least partially re-encoded pulse being generated by processing decoding _; equalizing the received RF pulse Processing; and extracting replacement data bits from the received radio frequency pulses; wherein the wire processing H and the multi-branch equalization H are used in combination Generating a data block from a soft decision or a replacement soft decision; deinterleaving the data block; decoding a frame from the data block, · 35 1323095 ' To: #料魅New coding to generate at least partially recoded data Blocking; and interleaving the at least partially re-encoded data block to generate an at least partially re-encoded pulse. 6. The wireless terminal of claim 5, wherein the frame is a voice thief data frame. 7. The wireless terminal of claim 5, wherein: the first equalizer processing branch comprises: • a 1 component and a Q component interference cancellation portion; and a decision feedback equalizer portion; The equalizer processing branch includes: an I component and a Q component interference canceling portion; and a linear equalizer portion. 8. A method of equalizing received radio frequency pulses, comprising: receiving radio frequency pulses; • decoding a known training sequence from the received radio frequency pulses; training based on the decoded known scale relative to the equalizer Processing the radio frequency pulse received by the branch equalization with the first equalizer; deinterleaving the radio frequency pulse; decoding the radio frequency pulse to obtain the extracted soft sample; decoding the data bit from the extracted soft sample; re-encoding the data Bits are generated by at least partially re-encoding soft samples, interleaving the at least partially re-encoded soft samples to generate at least partially re-encoded pulses 36 1323095, re-reading the received RF pulses from the memory to a second An equalizer eight; the time division uses the at least partially re-encoded pulse to train the second equalizer to process the branch. The second equalizer equalizes the received RF pulse in the memory. Radio frequency pulses are deinterleaved; the radio frequency pulses are decoded to obtain replacement soft samples; and decoded from the replacement soft samples Replace the data bit. 9. For example, the material building mentioned in item 8 of the patent application scope. In the above, the method includes the method described in claim 8, wherein the replacement data bit is generated. First and 4 RF pulses to 37