WO2006045226A1

WO2006045226A1 - Method and equipment adapted to 8psk equalization demodulation in edge system

Info

Publication number: WO2006045226A1
Application number: PCT/CN2004/001241
Authority: WO
Inventors: Gang Xiong
Original assignee: Zte Corporation
Priority date: 2004-10-29
Filing date: 2004-10-29
Publication date: 2006-05-04
Also published as: CN100452890C; CN1994003A

Abstract

A method and equipment adapted to 8 PSK equalization demodulation in EDGE system, said method includes: reversing the received I,Q signal with e-j3hl I8 ; relating the reversed signal to obtain the evaluation value of channel parameter; according to the evaluation value of channel parameter, performing delay synchronization and determining the time advance and energy , as well as the biggest channel parameter evaluation value; matching-filtering the evaluation value of channel parameter with the reversed and synchronized signal; by using the channel parameter evaluation value and matching-filtered signal, searching the maximum likelihood sequence and outputting the maximum likelihood sequence as demodulation result; converting the output maximum likelihood sequence to bit value. The present invention also discloses an equipment adapted to 8PSK equalization demodulation in EDGE system. By reducing the state number, selecting the maximum likelihood sequence from all the possible sequence as the output sequence, under the premise of meeting the EDGE protocol specification, the present invention reduces the calculation complexity, and guarantees the performance of base band demodulation.

Description

Method and device for 8PSK equalization demodulation for EDGE system

Technical field

The present invention relates to a method and apparatus for equalization demodulation in the field of mobile communications, and more particularly to a method and apparatus for 8PSK equalization demodulation for an EDGE (Enhanced Da ta tes for GSM Evoluting) system. Background technique

The GSM system, which is the second generation mobile cellular communication system, has been widely used worldwide. However, with the development of mobile communication technologies and the diversification of services, the demand for data services is increasing. In order to meet people's needs, the GSM system supporting voice services has proposed two types of high-speed data services in its PHASE2 and PHASE2+ specifications, namely HSCSD (High Speed Circuit Switched Data) based on high-speed data bit rate and circuit switching. And GPRS (General Packet Radio Service) based on packet switched data. Although HSCSD and GPRS adopt a multi-slot operation mode, the data transmission rate has been improved to some extent, but it still uses the modulation method of GMSK (Gaussian Minimum Shift Keying), and the third generation mobile The wide-area coverage of the 384 kb it/s data rate of the letter system and the local coverage of the data rate of approximately 2 Mb it/s are still far-reaching, so it is necessary to adopt more advanced communication and signal processing techniques to further expand the GSM system. Capacity. ETSI (European Telecommunications Standards Institute) has decided to develop a GSM implementation for enhanced data speed EDGE as an evolutionary direction for GSM in the future. So EDGE came into being.

The figure is a basic schematic diagram of a channel model of a mobile communication system. The baseband receiver receives the data transmitted through the wireless channel port. The data is first demodulated by the demodulation module for the received baseband I and Q signals, and the demodulated result is sent to the channel decoding module for channel decoding. For the control channel, the transmission information of the system can be obtained directly; and for the traffic channel, the source decoding is required to obtain the voice and data sent by the system. In the channel model of the signal system, the demodulation module is located at the front end of the receiver, and it can be seen that the demodulation performance is good or not. The connection determines the performance of the entire mobile communication system.

In order to provide higher data communication rates in existing cellular systems, EDGE introduces a multi-level digital modulation scheme, 8PSK modulation. Since 8PSK modulation is a linear modulation, three consecutive bits are mapped to one symbol of the I/Q coordinates, thereby providing higher bit rate and spectral efficiency, and the implementation complexity is moderate. The modulation of GMSK used in the GSM system is also part of the EDGE modulation scheme. The symbol rates of both modulation modes are 271kbi t/s, and the net bit rate per time slot is 22. 8kbi t/s (GMSK) and 69. 2kb i t /s (8PSI ). 8PSK modulation is used for the user's data channel, and GMSK is used for all control channels on the GPRS 200kHz carrier.

In mobile communications, the channel characteristics of the wireless channel are very poor, mainly manifested by multipath fading and Doppler fading. Multipath fading can cause inter-symbol interference between signals, and the receiving end must use equalization techniques to eliminate the effects of the channel. Equalization compensates for amplitude and delay in the channel by an equalizer in the receiver to eliminate inter-symbol interference. The demodulator must estimate the original modulated data to the greatest extent possible in the received interfered signal. In order for the demodulator to perform this work, each burst sequence contains a predetermined sequence that the receiver can recognize. That is, the training sequence is such that the receiver can estimate the signal distortion caused by the propagation.

Equalization techniques are generally classified into linear equalization methods, nonlinear equalization methods, and maximum likelihood sequence equalization (MLSE). Compared with some suboptimal equalization techniques, the most likelihood sequence equalization method is used as an optimal sequence estimation method in a channel with intersymbol interference, and is often used in an equalizer of a mobile radio channel. In the GSM system, the equalizer based on the Vi ter bi algorithm is commonly used to implement the MLSE. When the optimal MLSE algorithm is adopted, the total number of states of the channel is M", where is the size of the modulation symbol table of the signal, and Z is the channel. In other words, the computational complexity of the MLSE algorithm based on the Vi terbi algorithm depends on M and L. When M is large, the computational complexity is still very high even when L is small.

In the GSM system, due to the use of GMSK binary modulation, the optimal Vi terbi algorithm can generally be used for demodulation, and the dispersion length in the channel is generally taken as ⁶ in the system, so the computational complexity is not very high. In the EDGE system, as shown in Figure 2, Figure 2 is in the EDGE protocol. The specified 8PSK constellation. In the EDGE system, the size of the modulation symbol table of the signal is = ⁸ . It can be seen that if the Vi terbi-based MLSE method implemented in GMSK modulation is used, the total state number of ⁸ PSK demodulation is ⁸⁵ - 32768, which is complicated. The degree will increase greatly. For baseband demodulation systems, this is not possible with software radio. Therefore, it is necessary to find a suitable method to achieve equalization demodulation, which not only has low computational complexity, but also ensures demodulation performance.

In the US-specific J' M "Met hods, Receivers and Equa li zers Having Increased Computate ional Ef fic iency" with the special spear J number 6, 707, 849, the Joint Maximum Likelihood Sequence Estimation (MLSE) and The technique of decision feedback method (DFS E) proposes an effective method to reduce complexity. This patent divides the channel dispersion length into two parts, one for the MLSE algorithm and the other for the DFSE algorithm. The number of states in the grid map can be determined according to the dispersion length used for the MLSE algorithm, thus reducing the MLSE algorithm. The number of states. At the same time, when calculating the branch metric, the different parts moved into all the states are calculated separately from the same part, which can further reduce the computational complexity. But when used for the MLSE algorithm's dispersion length ': compare ^, the computational complexity of the entire system is still high. In addition, the use of DFSE technology will cause errors to accumulate when adjusting the balance.

In the US patent "Maximum Like l ihood Sequence Es t ima tor wi th Var iable Number of States", the training sequence is used to estimate channel parameters, and Vi terbi is determined based on the estimated channel parameters. The number of states of the algorithm is the result of the balanced adjustment. This patent requires different processing of different state numbers, so the corresponding program space needs to be increased, and the resource consumption is relatively large. In addition, when the modulation symbol of the signal is large, the complexity of the system increases exponentially and is not used in the EDGE system.

Chinese patent 01112664. 7 proposes an adaptive 8PSK modulation equalization demodulation implementation method for EDGE technology in the third generation mobile communication system. First, the inverted signal is E-filtered and the output will be output. As a result, coherent demodulation is performed to determine the demodulated symbols, and then the coherent demodulated data is used to perform N iterative path searches (N _rf is the number of iterations, generally For 1 or 2), the sequence with the largest likelihood function value is selected from all possible sequences as the output sequence. Using this method, the computational complexity is reduced to

. Although the invention has a small amount of calculation and a high system achievability, the performance of the algorithm for the decentralized demodulation is not fully stipulated in the EDGE protocol, and is not suitable for implementation in the actual EDGE system. Summary of the invention

The object of the present invention is to provide a method and apparatus for 8PSK equalization demodulation for an EDGE system by reducing the number of states and selecting a maximum likelihood sequence from all possible sequences. The output sequence can greatly reduce the computational complexity and ensure the performance of the baseband demodulation system while satisfying the EDGE† specification.

In order to achieve the above object, the present invention provides a method for 8PSK equalization demodulation for an f EDGE system, comprising the following steps:

Step 1, using the ^ ³ ' ⁸ for the received I and Q signals;

Step 2: Perform correlation on the inverted signal to obtain an estimated channel parameter;

Step 3. Determine the time advance amount and the energy and the maximum signal according to the estimated value of the channel parameter, and estimate the value of the channel;

Step 4. Match and filter the inverted signal according to the estimated value of the energy and the maximum channel parameter;

Step 5. Using the estimated energy and the maximum channel parameter to estimate the singularity and matching the filtered signal, find the maximum likelihood sequence, and output the maximum likelihood sequence;

Step 6. Convert the output maximum likelihood sequence symbol value into a bit value.

The algorithm for finding the maximum likelihood sequence in step 5 is an algorithm for finding a maximum likelihood sequence using suboptimal, and the suboptimal algorithm for finding the maximum likelihood sequence includes the following steps: Step 51: From time The unit k=L starts, and each state of the signal outputted by the matched filter is calculated corresponding to the metric of the sequence of the L segment branch, and the surviving sequence corresponding to each state and the surviving sequence metric are stored; Step 52, k=k+l, connect the branch of this node into each state and the previous node connected to these branches to obtain a path; in the state transition, calculate the secondary branch metric of each transfer path;

Step 53: Select a maximum sub-branch metric from the calculated sub-branch metrics, save the maximum sub-branch metric as a branch metric of the state transition, and save the symbol value of the branch metric; Step 54: Branches in each path The metric is added to the surviving metric of the previous path stored in the previous state of the pair to obtain the cumulative metric of the four paths;

Step storm 55. Select and store a path with the largest cumulative amount from the two paths entering the state as the new surviving path, and the maximum metric is used as the surviving metric of the determinate, and all other non-surviving paths are deleted at the same time;

Step 56: If k <the length of the sequence to be demodulated, go to step 52; otherwise, compare the surviving metrics of the respective states to obtain a maximum surviving metric, and the surviving sequence corresponding to the largest surviving metric is the maximum likelihood sequence.

The invention also provides an 8PSK equalization demodulation device suitable for an EDGE system, comprising: a signal inversion module, a channel estimation module, a delay synchronization module Φ matching filtering module, and a balanced demodulation module and a symbol conversion module, wherein The signal 3 conversion module has an input end connected to the signal sampling output end of the receiver end, configured to invert the sampled 8PSK signal, and simultaneously output the inverted signal to the channel estimation module and the matched filtering module;

The channel estimation module is configured to correlate the input training sequence with the inverted signal to obtain an estimated value of the channel parameter, and output the estimated value of the obtained channel parameter to the delay synchronization module;

The delay synchronization module is configured to obtain a timing advance according to an estimated value of the input channel parameter, determine an estimated channel parameter with the largest energy, and simultaneously input the channel parameter estimated value with the largest energy to the matched filtering module. And equalization demodulation module;

The matched filtering module is configured to match and filter the channel parameter estimation value with the largest energy output from the delay synchronization module and the inverted nickname output from the signal inversion module, and The filtered signal is output to the equalization demodulation module;

The equalization demodulation module is configured to search for a maximum likelihood sequence according to the input filtered signal and the channel parameter estimated value with the largest energy, and output the obtained fishing symbol value to the symbol conversion module;

The symbol conversion module is configured to convert the input symbol value into a bit value corresponding thereto. The invention adopts the reduced state Vi terb i algorithm to perform equalization demodulation on the input baseband digital I and Q signals, and selects the maximum likelihood sequence as the output sequence from all possible sequences, under the premise of meeting the EDGE protocol specification. It can greatly reduce the complexity of i-time calculation, and solve various distortions of signals in the wireless channel in the EDGE system, such as the occurrence of zeros in the channel special 'I', especially the inter-symbol interference caused by the multipath effect. Illustration

1 is a basic schematic diagram of a channel model of an existing mobile communication system;

Figure 2 is a constellation diagram of 8PSK in the existing EDGE protocol;

Figure 3 is a flow chart of the method of the present invention;

4 is a disk diagram of a mesh transfer in the present invention;

Figure 5 is a schematic diagram of 8PSK subset splitting in the present invention;

Figure 6 is a structural view of the apparatus of the present invention. detailed description

The method of the present invention is applicable to 8PSK equalization demodulation of an EDGE system. By reducing the number of states, the maximum likelihood sequence is selected from all possible sequences as an output sequence, which can be greatly reduced under the premise of satisfying the EDGE protocol specification. The computational complexity, specifically, the flow of the method is as shown in FIG. 3, and includes the following steps:

Step 1. Invert the received I and Q signals by using ^ ^{Λ / 8} ;

Step 2: Perform correlation on the inverted signal to obtain an estimated channel parameter; Step 3: Obtain time advance quantity and energy and maximum channel parameter estimation value according to the estimated value of the channel parameter, thereby performing delay synchronization; since the space signal transmission has a certain time delay, the transmission signal and the reception signal are required. Synchronizing in time, the timing advance is to achieve this synchronization, to ensure the correctness of the demodulation.

Step 4: matching and filtering the estimated value of the channel parameter and the inverse-synchronized signal; Step 5: using the estimated value of the channel parameter and the matched filtered signal, finding the maximum likelihood column, and making the maximum likelihood sequence Output as a demodulation result;

For each of the above steps, the details are as follows,

In step 1, because of the protocol according to EDGE, the baseband modulation of 8PSK requires symbol rotation, that is, where k represents the index of the modulation symbol, and the rotation angle of each modulation is ³ ^/ ⁸ .

When demodulating 8PSK, the phase anti-rotation operation is required, and the phase anti-rotation operation is performed according to the following formula:

Yj{k) = _yj {k) -

After the formula is expanded:

3k

Ij(k) = Jj(k) cos ■Q ,(k)sin

,Ding"

Wherein, the received signal is received at the position of the jth sample point on the kth symbol, and ² ^^ is the real part and the imaginary part of the received signal respectively; for the inverted signal, » and ² ) are respectively inverted signals ^y » The real and imaginary parts.

3ht

The present invention uses a look-up table method to obtain sine and cosine values. Because it is ² weeks ( 2>k7v

Cos I , . ( >k \

Sin

Period, so, in a change cycle (k = 0, 1...15), the values of 8 and 8" can be divided into 16 pairs, calculate the 16 pairs of data, create two tables, complete the phase anti-rotation Transport

( 3Κπλ . ( ^πλ cos sin

Count. Among them, in the two tables, one table stores the value of ^ ⁸ ), and the other one stores the value of ⁸ . The real and imaginary parts in the equation (9) are obtained by querying the two, thus obtaining the respective inverted signals t).

In the step 2, an estimated value of the channel parameter is obtained by correlating the training sequence with the inverted signal. In the EDGE system, each burst has 156.2 symbols, which occupies one time slot, about 5 s s, which is small compared with the channel fading period. Therefore, it can be considered that the channel time variation in a time slot has little effect. Because in wireless communication, the signal propagates through the air, there is a channel to influence it. If the mobile station is moving, the channel will change, but since the time slot of a time slot is very short, then the change of this channel can be considered negligible. The training sequence ^ 26 symbols (about 96.2 μ _δ ) is placed in the middle of the burst so that the channel parameters are considered to be constant in one time slot. Here, the sliding correlation method is used to obtain the parameters of the signal by using the orthogonal characteristics of the training sequence.

The estimated value of the channel parameter associated with the slipping of the signal after the training sequence can be expressed as:

Where t = 0,l,...,N , depending on the scope of the search, α ) is the symbolic value of the training sequence, ) represents the estimated value of the signal at the position of the 'J' sample points on the first symbol, ) Expressed as a signal after flipping.

In step 3, the advance of the time is found based on the estimated value of the signal obtained in the previous step, and the energy and the maximum estimated channel parameter are determined as the coefficients of the equalized, demodulated basic channel and matched filter. The process of delay synchronization is the process of finding the maximum energy sum in the channel parameter estimates. The formula for calculating the timing advance is:

∑4 = argmax ^ n(k)\ >

δ 04 241 The final (ie energy and maximum) channel parameter estimates can be obtained by time advance:

Where is the final channel parameter estimate at the J'th sample point position on the first symbol.

Among them, the timing advance is obtained by finding the sum of the energy of the gas, for example, the parameter estimates are a, b, c, d, e, then the energy sum of the different positions is

(laT+lbl ² ), ( |b| ² +|c| ² ), ( |c| ² +|d| ( |d| ² +|e| ² ), find the largest value, such as the second The value is the largest. That is the maximum energy sum, so TA = 2. The obtained time advance, you can get the energy and maximum by the formula g ) = ^. t+73⁄4) = ο, ι,.··,ζ-1 Channel parameter estimate g).

In step 4, the estimated channel parameter output in step 3 is used as the input of the matched filter module, and matched and filtered with the inverted signal. What needs to be noted is the coefficient of the matched filter, that is, the coefficient (FIR filter) required to filter the signal, which is conjugated with the energy and the estimated value of the largest channel parameter.

In step 5, the maximum likelihood sequence is searched for based on the channel parameter estimate of the input in step 3 and the matched filtered output signal, and the demodulated result is found.

In this step, the algorithm for finding the maximum likelihood sequence is a suboptimal algorithm for finding the maximum likelihood sequence. The present invention mainly focuses on the complexity of the implementation of the salty demodulation equalization, and the basic principle is based on the Viterbi algorithm. Specifically, the algorithm for finding the maximum likelihood sequence of the sub-optimal includes the following steps:

Step 51: First, initialization is required. Starting from the daytime unit k=L, each state of the signal outputted by the matched filter is calculated as a metric corresponding to the sequence of the L segment branch, and the surviving sequence corresponding to each state and the surviving sequence metric are stored;

Step 52, k=k+l, connecting the branch of the node into each state and the previous node connected to the branches to obtain a path; in the state transition, calculating a secondary branch metric of each transfer path, The calculation formula for the sub-branch metric is: Sub _ Branch _ Metric = Re^ i\ Z _k - ^S^ > where ^ is the estimated value of the symbol in the surviving path;

S I is an autocorrelation function of the synthesized channel;

The conjugate value of the symbol of the transmitted signal;

Step 53: Select a maximum secondary branch metric from the calculated secondary branch metric, save the maximum secondary branch metric as a branch metric of the state transition, and save the symbol value of the maximum secondary branch metric;

Step 54: Add the branch metric in each path to the surviving metric value of the previous path stored in the corresponding previous level state to obtain the cumulative metric value of the four paths; before or after the step ⁵⁴ a state surviving symbol path history table, wherein the surviving symbol path history table is an array of length 5, and the symbol estimates stored on the surviving paths of the past 5 moments are saved ^

Step 55: Select and store a path with the largest cumulative amount from the two paths entering the state as a new surviving path, and the maximum metric is used as a surviving metric of the state, and all other non-surviving paths are deleted at the same time;

Step 56: If k<the long sequence of the sequence to be demodulated, go to step 52; otherwise, stop the iteration and compare the surviving metrics of each state, and the surviving sequence corresponding to the largest surviving metric is the maximum likelihood sequence.

The process of the above steps 52-55 can be represented by FIG. 4, which is a disk shape operation diagram of the mesh transfer according to the present invention. Among them, the arrow indicates the branch, the state 2 / and 2. +1 are the two states before the transition, and the state and /+16 are the two states after the transition. For the two adjacent states 2 / and 2 / +1 of the previous stage, there are four branches. In each of the roads, four sub-branch metrics need to be calculated, and the largest one is selected from the sub-branch metrics. As a branch metric, each branch metric is added to the cumulative metric of the previous path stored in state 2 i or 2i+l of the previous level to obtain the accumulation of four paths Branch-UpO, Branch-Upl, BranctuDownO, and Branch-Downl. metric. (It should be noted that the branch metrics in these four paths are in the 4 sub-branch metrics: ^ large value.) Then compare the two current states and +16 under the corresponding two current states, each current state is left A path with a larger cumulative metric is used as the surviving path. At the same time, the value t of the current state is used as the surviving metric and the input bit corresponding to the surviving path is stored in the corresponding buffer, and the next level of calculation is prepared.

As shown in the figure, for the state (/) at the next moment, the symbol value in the subset 0 shown in Fig. 5 is input. The value of the survivor metric is the largest value that compares the cumulative metrics in Branch-UpO and Branch_Upl. Similarly, for the state of the next moment (/+16), the input ^ is the symbol value in subset 1 shown in Figure 5. The value of the survivor metric is the largest value that compares the cumulative metrics in Branch-DownO and Branch-Down.

FIG. 5 is a schematic diagram of 8PSK subset splitting according to the present invention, dividing the 8PSK signal into a state with a small number, and according to this state, creating a disk map 0 of the mesh transfer as shown in FIG. 4. As shown, 8PSK is divided into two subsets, denoted as subset 0 and subset 1, respectively. Subset 0 contains the symbols 1, 3, 5, 7; and subset 1 encloses the symbols 0, 2, 4, 6. Thus the state transition in the grid diagram in the Vi teirbi: algorithm can be categorized as a transition of subset 0 and subset 1, rather than a shift of each symbol. As can be seen in Figure 4, each state - ""^"^"^"*- ₅ } represents each subset 0 or 1, rather than a value, but its corresponding symbol value needs to be stored in the survivor. The symbol path history table needs to be used in the calculation of subdivision metrics. Therefore, the number of whole states is 2 ⁵ = 32 instead of 8 ⁵ = 32768, so the computational complexity is greatly reduced.

The present invention also provides an apparatus for 8PSK equalization demodulation for an EDGE system. The structure of the apparatus is shown in FIG. 5. The apparatus includes: a picture number inversion module 301, a channel estimation module 302, and a delay synchronization module 303. The matched filtering module 304, the equalization demodulation module 305, and the symbol conversion module 306. The input end of the signal inversion module 301 is connected to the signal sampling output end of the receiver, and is used for inverting the sampled 8PSK signal, and simultaneously outputting the inverted signal to the channel estimation module 302 and the matched filtering mode: ^304;

The channel estimation module 302 is configured to correlate the input training sequence with the inverted signal. Obtaining an estimated value of the channel parameter, and outputting the estimated value of the obtained channel parameter to the delay synchronization module 303;

The delay synchronization module 303 is configured to obtain an advance amount of time according to the estimated value of the input channel parameter, determine the channel parameter estimation value with the largest energy, and simultaneously input the channel parameter estimation value with the largest energy to the matched filtering module 304 and the equalization. Demodulation module 305;

The matching filtering module 304 is configured to perform matched filtering on the estimated value of the energy output from the delay synchronization module 303 and the inverted signal output from the signal inversion module 301, and output the matched filtered signal to the signal. Equalization demodulation module 305;

The equalization demodulation module 305 is configured to find the maximum likelihood sequence according to the input matched filtered signal and the energy maximum channel parameter estimation value, and output the obtained symbol value to the symbol conversion module 306;

The symbol conversion module 306 is operative to convert the input symbol values into corresponding bit values. The equalization demodulation module 305 is the core of the apparatus for finding the maximum likelihood sequence, and the equalization demodulation module 305 uses the estimated channel parameter value and the output of the matched filter as the input of the equalization demodulation to find the maximum. The likelihood sequence, while outputting the result of the demodulation. In addition, it should be noted that when using fixed-point DSP to achieve equalization demodulation, the value range of the DSP has a value of P. When implementing metric updates in the Vi terbi algorithm, the saved metrics are most likely to overflow, beyond the representation range of the DSP. Therefore, in the implementation of the DSP, it is necessary to ensure that there is no overflow during the operation, and it can adapt to a large dynamic range.

The implementation of the equalization demodulation module 305 is further illustrated in Figures 4 and 5. Figure 5 is a schematic diagram of 8PSK subset splitting. The 8PSK signal is segmented into a smaller number of states, and a mesh map is created based on this state. As shown, 8PSK is divided into two subsets, denoted as subset 0 and subset 1, respectively. Subset 0 contains the symbols 1, 3, 5, 7; and subset 1 contains the symbols 0, 2, 4, 6. The state transition in the trellis diagram in the Vi terbi algorithm can then be seen as a transition of subset 0 and sub-1, rather than a shift of each symbol. In addition, its corresponding symbol value needs to be stored in the storage symbol path history table, and the symbol value needs to be utilized in calculating the secondary branch metric. FIG. 4 is a disk diagram of the mesh transfer according to the present invention. For the two adjacent states 2 i and 2 i+l of the previous stage, there are four branches. In the dish operation, the branch metrics of the four branches and the received signal need to be calculated (please specify that the secondary branch metric is How to calculate, select the largest one from the sub-branch metric as the branch metric), and each branch metric is added to the previous metric and the cumulative metric of the previous Luoyang stored in state 2 i or 2 i+l to obtain four Path Branch—UpO, Branch—Upl, Branch—Dou and Branch—Lengl cumulative metrics. (It should be noted that the branch metrics in the four paths are the maximum of the four sub-branch metrics.) Then, in the corresponding two current states i and i+16, the current state is left tired. Only one path with a larger metric value is used as the surviving path. At the same time, the metric of the current state is stored as a surviving metric and the input bit corresponding to the surviving path is stored in the corresponding buffer, and the next level of calculation is prepared.

In addition, in order to obtain a training sequence, the apparatus further includes a training sequence module, and the training sequence module is configured to output, to the channel estimation module, a training sequence for correlating with the inverted signal.

The method of the invention comprehensively considers the performance, complexity, stability and operation speed of the equalization demodulation method, and adopts the reduced state Vi terbi algorithm to perform equalization demodulation on the input baseband digital I and Q signals to solve the EDGE system. Various distortions of the medium signal in the wireless channel, such as zero occurrence of channel characteristics, especially inter-symbol interference due to multipath effects.

It should be noted that the above embodiments are only intended to illustrate the technical solutions of the present invention and are not intended to be limiting, and the present invention will be described in detail with reference to the preferred embodiments. The modifications and equivalents of the present invention are intended to be included within the scope of the appended claims.

Claims

Claim

A method for 8PSK equalization demodulation for an EDGE system, comprising the steps of:

Step 1. Use the ^ to flip the received I and Q signals;

Step 3. Determine the time advance amount and the energy and the maximum channel parameter estimation value according to the estimated value of the channel parameter;

Step 5. Use the energy and the estimated value of the largest channel parameter and the matched filtered rhythm signal to find the maximum likelihood sequence, and ^! Taking the maximum likelihood sequence output;

2. The method for 8PSK equalization demodulation applied to an EDGE system according to claim 1, wherein in the step 2, the process of correlating the inverted signal is: using the training sequence to reverse the signal Make a sliding correlation.

3. The method for 8PSK equalization demodulation for an EDGE system according to claim 1, wherein said time advance of said step 3 is determined by the following formula

Wherein, the channel parameter estimation value at the J'th sampling point position on the first symbol is represented; the energy and the maximum channel parameter estimation value are obtained by the following formula:

Gj(k) = hj(k + TA) k = 0,l, - -,L - \

The method for 8PSK equalization demodulation applied to the EDGE system according to claim 1, wherein the algorithm for finding the maximum likelihood sequence in the step 5 is to find the maximum likelihood sequence by using the J3⁄4 suboptimal sequence. Algorithm.

5. The method for 8PSK residual demodulation for an EDGE system according to claim 4, wherein the suboptimal algorithm for finding a maximum likelihood sequence comprises the following steps:

Step 51: Starting from a time unit k=L, calculating, for each state of the matched filtered output signal, a metric corresponding to the sequence of the L segment branch, storing the surviving sequence corresponding to each state and the surviving sequence metric;

Step 52, k=k+l, connecting the branch of the node into each state and the previous node connected to the sub-masters to obtain ^two paths; calculating a sub-branch metric of each transfer path;

Step 53: Select a largest sub-branch metric from the calculated sub-branch metrics, save the maximum sub-branch metric as a branch metric of the state transition, and save the symbol value of the branch metric; Step 54: Branches in each path The metric is added to the surviving metric of the previous path stored in its corresponding previous state to get the cumulative metric of the four-way path.

Step 55: Select and store a path with the largest cumulative metric as a new surviving path from the path that enters the state to be entered, and the maximum metric is used as a fortunate measure of the state, and all other non-surviving paths are deleted.

6. The method for 8PSK equalization demodulation for an EDGE system according to claim 5, wherein the sub-branch metric in the escape step 52 is obtained by the following formula:

Where -' is the estimated value of the symbol in the surviving path;

S 1 is an autocorrelation function of the synthesized channel;

r: is the conjugate value of the symbol of the transmitted signal.

7. The method of 8PSK equalization demodulation for an EDGE system according to claim 5, wherein a surviving symbol path for each of said states is saved before or after said step 54 The path history table, the surviving symbol path history table is an array of length 5, and the array is used to store symbol estimation values on surviving paths of the past 5 moments.

8. An apparatus for 8PSK equalization demodulation for an EDGE system, comprising: a signal inversion module, a channel estimation module, a delay synchronization module, and a matched filtering module, and characterized in that: a balanced demodulation module, a symbol conversion module, The signal inversion module is connected to the signal sampling output end of the receiver end, and is used for inverting the sampled SPSK signal by ^3/8 , and simultaneously outputting the inverted signal to the channel estimation module and matching. a filtering module, configured to: correlate the input training sequence with the inverted signal to obtain an estimated value of the channel parameter, and output the estimated value of the obtained channel parameter to the delay synchronization module;

The delay synchronization module, according to advance estimates channel parameter inputs obtained time, while determining the maximum energy of channel parameter estimates, and ^ ¹ the highest energy channel parameter estimates simultaneously input to the matched filter Module and equalization demodulation module;

The matching filtering module is configured to perform matched filtering on the channel parameter estimation value that outputs the maximum energy output from the delay synchronization module and the inverted signal output from the signal inversion module, and output the matched filtered signal to the equalization solution. Tuning module

The equalization demodulation module is configured to search for a maximum likelihood sequence according to the input matched filtered signal and the channel parameter estimate with the largest energy, and output the obtained symbol value to the symbol conversion module;

The symbol conversion module is configured to convert the input symbol value into a bit value corresponding thereto.

9. The apparatus for 8PSK equalization demodulation of an EDGE system according to claim 8, further comprising a training sequence module, wherein said training sequence module is configured to output to said channel estimation module for use with The inverted signal is used to perform the relevant training sequence.