KR20040057028A - A Hybrid Multiuser Interference Cancellation Scheme Using the Clustering Algorithm Based on the Dynamic Programming and its Apparatus - Google Patents

Abstract

PURPOSE: A hybrid multiuser interference cancellation method using a dynamic programming-based grouping technique and an apparatus therefor are provided to prevent the deterioration of detection performance due to a signal power difference by forming at least a group so that each of the received user signal power differences can be minimized. CONSTITUTION: A hybrid multiuser interference cancellation apparatus(300) comprises a channel estimator(301), a sorting part(302), a cluster calculator(303), and a hybrid interference canceller(304). The channel estimator(301) discriminates use signals from a plurality of signals received from the external and calculates the power of each of the discriminated user signals. The sorting part(302) arranges the calculated user signal power in the order of a descending series. The cluster calculator(303) forms more than a group through a dynamic programming-based grouping technique using the sequential numbers based on the arranged user signal power.

Description

{A Hybrid Multiuser Interference Cancellation Scheme Using the Clustering Algorithm Based on the Dynamic Programming and its Apparatus}

The present invention relates to a multi-user detection method for eliminating user-to-user interference caused by multiple users in a communication system using a code division multiple access (Code Division Multiple Access) method, in particular a multi-user detector with a hybrid interference cancellation technique The present invention relates to a method and an implementation structure of performing interference cancellation using a group configuration technique using an optimization method for effective interference cancellation.

Code division multiple access based on spread spectrum technology has been used as an important multiple access in second and third generation wireless mobile communication such as IS-95 and IMT-2000.

In the dual direct sequence code division multiple access (Direct Sequence Code Division Multiple Access) method, a plurality of users share time and frequency, thereby distinguishing users by a signature sequence.

However, full orthogonality between user signature sequences is not guaranteed, so there is a cross correlation between signature sequences, resulting in multiple access interference (MAI) that limits the transmission capacity and performance of a communication system. Will be added.

In addition, there is a near-far problem caused by differences in signal power of users. In order to solve this problem, a multiuser detection method has been proposed to eliminate interference between users.

The optimal detector for the multi-user detection method is theoretically found, but this method is also unsuitable for practical implementation because the computational complexity is given as an exponential function for the number of users. Thus, research has been conducted on quasi-optimal detectors with reduced computational complexity even in the presence of performance degradation.

Among the quasi-optimal detectors, the interference canceller is simpler than the linear detector because it does not require the inverse matrix calculation of the cross-correlation matrix, unlike the linear detector. According to the structure, it is classified into sequential interference cancellation (SIC) technique and parallel interference cancellation (PIC) technique.

A sequential interference cancellation technique that aligns the signal power of all users in order of magnitude, sequentially estimates and regenerates the signals, and then removes them from the received signal shows good bit error rate (BER) performance, but The processing delay time increases in proportion to the number.

On the other hand, the parallel interference cancellation technique detects the signals of all users at the same time and removes the regenerated signals of the remaining users except the users from the received signals. Since the multi-stage structure must be used, the computational complexity increases.

In addition, when the power control for the root problem is perfect, the parallel interference canceller performs better than the sequential interference canceller, and when the power control is not perfect, the sequential interference canceller performs better than the parallel interference canceller.

As a technology related to such interference cancellation, the right holder has a patent of LG Electronics Co., Ltd. [Patent Name: Multi-User Interference Cancellation Device, Application No. 10-2001-13752], which descends all the signal powers of the multi-users. By removing the interference after sorting by, the present invention provides a multi-user interference canceling apparatus and method having a large amount of computation but a good performance. However, such interference cancellation method also increases processing delay time in proportion to the number of users.

The technical problem to be achieved by the present invention is to solve this problem, by forming at least one or more groups so that the difference in the signal power of each received user, it is possible to prevent the overall detection performance degradation due to signal power difference It is to provide a hybrid multi-user interference cancellation method and apparatus using dynamic programming based grouping technique.

In addition, the present invention not only reduces the computational complexity of such group members and interference cancellers, but also uses a hybrid programming-based grouping technique that can prevent additional cost increases (hardware and software costs). To provide a multi-user interference cancellation method and apparatus therefor.

1 is a view showing the structure of a general multi-stage hybrid interference canceller.

2 is a diagram illustrating an interference cancellation unit constituting a general multi-stage hybrid interference canceller.

3 is a diagram illustrating a configuration of a hybrid multi-user interference cancellation apparatus according to an exemplary embodiment of the present invention.

4 is a diagram illustrating a dynamic programming based grouping technique according to an embodiment of the present invention.

FIG. 5 is a diagram illustrating a trellis configured to calculate a grouping technique when a number of groups is determined according to an embodiment of the present invention.

FIG. 6 is a diagram illustrating a trellis configured to calculate a grouping technique when the number of groups is not determined according to an embodiment of the present invention.

※ Explanation of code for main part of drawing ※

300: mixed multi-user interference cancellation device

301: communication path estimation unit 302: alignment unit

303: group calculation / former 304: hybrid interference canceller

In order to achieve the above object, the hybrid multi-user interference cancellation method using the dynamic programming based grouping method according to the present invention is a hybrid multi-user interference cancellation method for removing interference between a plurality of user signals. When a user signal is received, calculating power of the received user signal, and numbering the calculated signal power in order of magnitude; b) sorting each given user number in descending order; c) respectively forming at least one user group such that each signal power according to the sorted user numbers is less than a center value or an average value within the same group; And d) performing parallel interference cancellation on each user signal in the formed group, and performing sequential interference cancellation between the formed groups.

In this case, step c) may include: calculating an average signal power of each formed user group to define an objective function of each branch; And calculating, for each branch, the sum of all objective functions imposed on at least one path by which the branches gathered at a particular node; Selecting a branch having the smallest sum of the calculated objective functions as an optimal path; And collecting a number assigned to each node in the selected optimal path to form a user group.

In addition, a hybrid multi-user interference cancellation apparatus using a dynamic programming based grouping method according to another embodiment of the present invention is a hybrid multi-user interference cancellation apparatus for removing interference between a plurality of user signals. A communication path estimating unit for calculating power of the received user signal, and numbering the calculated signal power in order of magnitude when received; An alignment unit for sorting each of the given user numbers in descending order; A group calculator / former for respectively forming at least one user group such that each signal power according to the sorted user numbers is less than a center value or an average value within the same group; And a mixed interference cancellation unit for performing parallel interference cancellation for each user signal in the formed same group, and performing sequential interference cancellation between the formed groups.

In addition, in a recording medium including a hybrid multi-user interference cancellation method for removing interference between a plurality of user signals according to another embodiment of the present invention, a) when a plurality of user signals are received from the outside, the received user signal Calculating power of each and numbering the calculated signal power in order of magnitude; b) sorting each given user number in descending order; c) a function of respectively forming at least one user group such that each signal power according to the sorted user numbers is less than a center value or an average value within the same group; And d) each user signal in the formed group performs parallel interference cancellation, and a recording medium storing a program including a function of performing sequential interference cancellation between the formed groups.

Hereinafter, exemplary embodiments of the present invention will be described in detail with reference to the accompanying drawings.

First, consider the reverse link in the discrete signal synchronization direct sequence code division multiple access system. That is, in a synchronized channel, all data of all users are received at the same time, so signals of different symbol times are statistically independent of each other. Therefore, consider one symbol time.

The data modulation method uses binary data signaling (Binary Phase Shift Keying, BPSK). In addition, a situation in which the user's signal power is known by a separate channel estimation technique. When K user signals pass through an Additive White Gaussian Noise (AWGN) communication channel, a received signal sampled in units of chip symbols is represented by Equation 1 below.

Here, if the normalized signature sequence vector of the k-th user is s _k , the signal amplitude is w _k , and the data symbol is d _k , the received signal r is a matrix S of which each column is the signature sequence of the user and the signal amplitude of each user. The noise matrix is added to the product of the diagonal matrix W and d representing the data symbol vector. N and K are the processing gain and total number of users, respectively.

The multi-user detection method of estimating a signal of another user and removing a signal component generated again from a received signal and then determining a symbol of the corresponding user is called an interference cancellation technique. Such interference cancellation techniques are classified into sequential interference cancellation and parallel interference cancellation according to the method of interference cancellation, and a hybrid interference cancellation is a combination of the two techniques.

In an embodiment of the present invention, a plurality of user signals are divided into a predetermined number of groups, and user signals belonging to the same group are canceled using a parallel interference cancellation technique, and then interference cancellation is applied to each group using a sequential interference cancellation technique. Use the method.

This hybrid interference cancellation method is defined as an interference cancellation unit (ICU), which is a basic element of many interference cancellation techniques.

1 is a diagram illustrating an interference cancellation unit 100 for a g th group of an m th stage.

In detail, the subscript in the variable x _g represents the group argument, the first subscript in the variable x _{m, g} represents the step, and the second subscript represents the group. The input is the remaining signal component, e _{m, g} 101, from which the estimated signal of users belonging to the previous group has been removed from the received signal. Signal estimation of users belonging to the current group is made by matching filter A _g 102.

The transformation matrix F _g 104 denotes a parallel interference canceller of users belonging to the group. The signal of the users whose interference has been removed by the parallel interference canceller is given by y ' _{m, g} 105, and at each step the signal determination of the users in the group is determined by the result 104 of the previous step, y _{m-1, g} ( 106) and the result of adding y ' _{m, g} to the soft decision (108).

The input e _{m, g + 1} 110 delivered to the next group passes through a signal regenerator 103 which regenerates the estimated signal with the interference removed by the parallel interference canceller 104 and removes it from the input of the current group (109). ) Signal. The interference cancellation unit of each group performs the same calculation as above.

2 is a block diagram of the entire multi-stage hybrid interference canceller 200 using the interference cancellation unit 202 when the number of groups is G. FIG.

When the number of users belonging to each group is 1, the multi-level hybrid interference cancellation technique is the same as the general multi-stage sequential interference cancellation technique, and when the number of users belonging to each group is equal to the total number of users, it is the same as the general multi-level parallel interference cancellation technique. Do.

Assuming K users and the number of users in each group equal u = K / G, the input signal e _1,1 201 of the first interference cancellation unit is equal to the received signal r, and the initial value transmitted from the previous step is y _{0, g} = 0 (203). Descending matrix product If defined as, the signal determination value for the input signal of each group and the users belonging to the group is as shown in Equation 2 below.

Through the above equation, it can be seen that the hybrid interference cancellation technique of the first step is equivalent to multiplying a received signal by a matrix representing a linear filter. Determination values of all the groups using the above equation are shown in Equation 3 below.

The same process as the first step is repeated for the higher steps. In the upper stage, since the remaining signals passing through the first stage repeat the same process, the signal determination value 201 in each interference cancellation unit is an estimate of the remaining signal components of the users, and each interference cancellation unit removes each component. Therefore, the signal decision value of the upper step is given as the sum of the decision value of the previous step and the decision value calculated in the current step. The output of the m-th group and the g-th group is shown in Equation 4 below.

Equation 5 below expresses the judgment for all groups of the m-th step. From the result of Equation 5, the hybrid interference cancellation process forms a linear filter using the signature sequence information and the group information of the users, and determines the signal of each user through the linear filter on the received signal.

When there is a difference in signal power of users when the power control for inducing the interference cancellation method proposed by the present invention is not perfect, the influence of the difference is considered.

First, we consider the effect of signal power difference of users in case of poor power control in parallel interference canceller, which is an interference canceller inside each group of hybrid interference cancellation scheme. And through this, we consider the effect of user signal power difference belonging to the group on the hybrid interference cancellation.

Parallel interference cancellation for soft decision can be expressed in the form of a linear filtering process on a received signal. The detector output of the parallel interference cancellation scheme is shown in Equation 6 below.

We can rewrite the output y _k of the k th user as shown in [Equation 7]. The first term represents the signal of the k-th user, the second term represents the interference components of other users remaining after the interference cancellation, and the remaining term represents the linearly filtered noise. The signal-to-interference and noise ratio of the k-th user is shown in [Equation 8].

Assuming BPSK modulation, the molecule of Equation 8 is calculated as shown in Equation 9.

The denominator representing the remaining interference and noise power of other users is given by Equation 10 because the signal and noise are independent of each other and the user signal is independent of each other.

Equation (9) and [Equation 10] to rewrite [Equation 8] is the same as [Equation 11].

Average signal amplitude to represent the difference between average signal power and users' signal power And the user signal amplitude w _k are defined as follows.

The cross correlation between signature sequences is assumed to be well chosen so as not to affect the signal-to-interference and noise ratio. The sum of the power ratios of the k-th user and the other users in the denominator of the signal-to-interference and noise ratios is expressed by Equation 13 below.

The signal-to-interference and noise ratio in Equation 11 is inversely proportional to the square of the difference between the average signals. Therefore, in the case of the parallel interference cancellation technique, the larger the difference between users in the signal amplitude and the average signal amplitude, the lower the signal-to-interference and noise ratio. . Therefore, parallel interference cancellation techniques perform well when power control is complete.

The difference in the user signal magnitude belonging to the g-th group of the hybrid interference cancellation scheme where u is the number of users belonging to each group considers the signal-to-interference and noise ratio. If the magnitude of the entire user signal is sorted in descending order and the user is divided into G groups, the g-th group user is expressed by Equation 14 below.

Since the interference cancellation unit of the g-th group is represented by the linear filter, the linear filter output value of the g-th group is similar to the parallel interference cancellation method, and is given by Equation 15 below.

The difference with the parallel interference canceller is that the remaining received signal e _{g, which} is left in place of the received signal r, is given as input and is given by the user's signature sequence matrix A _g , which belongs to the group instead of the entire signature sequence matrix S. Accordingly, the decision value for the k-th user belonging to the group g can be determined in the same way as the parallel interference canceller, and when the signal-to-interference ratio is calculated in the form of Equation 11, Equation 16 below is used.

Equation 16 shows the signal-to-interference and noise ratio of the parallel interference cancellation scheme from Equation 16, but the components of other users belonging to group g and users belonging to groups following group g exist in the interference signal power. The point is different.

The hybrid interference cancellation scheme is inversely proportional to the power ratio of users belonging to the same group as the parallel interference cancellation scheme since the influence of the difference of user signal power belonging to the g group is considered.

If the g group user signal magnitude is defined as a difference from the amplitude average, the power ratio of users belonging to the same group is expressed by Equation 17.

Equation 17 shows that the greater the signal power difference among users belonging to the same group, the smaller the signal-to-interference and noise ratio. In the same way, for other groups, the difference between the average power magnitude and the average power of each group can be expressed.

Equation 17 is expressed as the sum of the signals of each group, and since the signals of each group can be represented by the average and the difference of the signals of each group, a large difference between the average of the signals of each group reduces the signal to interference ratio. It can be seen that.

Therefore, in the case of the hybrid interference cancellation scheme, a group should be configured to reduce the difference in signal magnitudes of users belonging to each group.

In other words, the smaller the difference in the signal size of the users belonging to the same group, the better the performance of the hybrid interference cancellation scheme.

Hybrid interference cancellation technique removes interference by creating groups to have the same number of users in each group without considering the user's signal size. If power control is not perfect, the user's signal size is not the same. If the signal size is large, the performance degrades overall.

Therefore, if power control is not perfect, performance can be improved by forming a group in consideration of the signal size of users and then removing interference.

The structure of the interference canceller proposed in the present invention has a structure including such group division, and the display example of such an interference canceller is attached to FIG. 3.

3 is a diagram illustrating a configuration of a hybrid multi-user interference cancellation apparatus according to an exemplary embodiment of the present invention, in which a grouping module calculates and transmits elements of each group to an interference canceller.

Referring to FIG. 3, the hybrid multi-user interference cancellation apparatus 300 according to an exemplary embodiment of the present invention includes a communication path estimator 301, an alignment unit 302, a group calculation / forming unit 303, and a hybrid interference cancellation unit. 304.

In detail, first, the communication path estimator 301 distinguishes each user signal from a plurality of received signals received from the outside, and then calculates the power of the distinguished user signal. At this time, the communication path estimator 301 assigns the user signals having a large power to 1, 2, 3,... N in order to display the distinguished user signals.

The sorting unit 302 sorts the calculated user signal power in descending order. In other words, the signals are arranged in order of 1, 2, 3 ... N in order of high power signal.

The group calculation / forming unit 303 forms at least one or more groups through dynamic programming based K-average grouping techniques using the numbers according to the user powers arranged in descending order. In forming, the power magnitude difference of each user signal belonging to the same group is minimized in order to prevent system performance degradation.

Dynamic programming is the solution to the least costly pathfinding problem, and is widely used for the limited resources and the distribution of these resources to a fixed number of people at minimal cost. The most representative application is the Viterbi decoding method of convolutional codes. Dynamic programming is the way to find the least expensive path through the path defined in the trellis and the nodes connecting the paths, where each node finds the least costly path among the branches encountered at that node. You'll make your choice and finally find the least costly route.

In other words, through dynamic programming, the optimal path is obtained and the group is composed of the numbers written in the states included in the optimal path. For example, the result of 1,4,7,10,12 If so, it indicates that five groups are configured, and the numbers in the result indicate the last user of that group.

Thus, one set is one user with the largest signal power, two sets are three users whose signal power is 2, 3, 4th, and 3 sets are 3 users with 5, 6, 7th, 4 sets are 8, 9, Three tenth users and five sets constitute a group of two 11th and 12th users, respectively.

The hybrid interference canceller 304 removes interference between the formed user signals using parallel interference, and removes interference between multiple users by using sequential interference cancellation between groups.

Then, the operation process of the group calculation / forming unit 303 of the main operation process of the hybrid multi-user interference cancellation device constituting such a structure will be described in detail with reference to the accompanying drawings.

First, the K-average grouping technique is an optimization technique for creating a group such that elements (signal power magnitudes) belonging to each group have a small difference from a center value or an average value, which is shown in FIG. 4.

4 is a diagram illustrating a dynamic programming based grouping technique according to an embodiment of the present invention.

First, the objective function is calculated as the sum of squared errors, and then n samples, x _1, x _2,... set S of x _n is K subsets s _1, s _{2,... When i} n that the subsets s _i of the sample number, m _i the average of the samples belonging to the s _i of when divided into s _K, which as shown in [Equation 18] below.

The sum of squared errors is given by (19).

For a given subset s _i , the mean value m _i can be said to be the representative or central value of s _i in the sense of minimizing [Equation 19].

Thus, the objective function calculates the n samples as the sum of Euclidean distances from the center value. The objective function depends on the division method of the group, and this value should be minimized to obtain the optimal division.

In addition, in order to obtain an optimal result in dividing the entire set having n samples into K subsets, all possible cases should be considered. The number of cases is K ^n, so the number of cases to be divided must be taken into account.

However, since the group division corresponds to the ordered group division, the optimal group division can be obtained using the dynamic programming without considering all cases. You must initially configure trellis to use dynamic programming. The method of configuring a trellis is classified into a case in which the number of groups is determined and a case in which the number of groups is not determined. 5 and 6 are attached to display examples of the trellis configuration method.

FIG. 5 is a diagram illustrating a trellis configuration when the number of groups is determined. As shown, the trellis includes nodes 501 and branches 502.

In this case, the trellis configuration first sorts the user numbers according to the power level of the received signal in descending order through the communication path estimator 301 and the alignment unit 302.

Nodes consisting of sorted user numbers are given in two dimensions, with each column representing each group, and the row direction elements of each column representing the last user's number as a group of the column. The number of rows represents all the possible ways to have the last element of the group.

Therefore, limiting the number of rows to m allows you to specify the maximum number of users that can be assigned to a group. Therefore, the number assigned to each node equals the number of all users assigned to the current group.

For example, the first row of the first column indicates that there is one element in the first group. Every group must contain at least one element, so the number given to the node in the second column starts at 2. Also, users belonging to one group cannot belong to the next group, so branches always link next to or below when they are connected to the next column. After the last K-1st group decision, all branches should converge to one node.

That is, all remaining users are assigned to the last group so that there are no unassigned elements. Perform dynamic programming on a given trellis. As you move from left to right along each branch in the trellis, make up a group from the number assigned to the node that you started and the node that arrived. That is, the user with the number assigned to the arrived node is assigned as a group, and the average power of the group is calculated and defined as the objective function for the branch. do. At each node, the branches collected at the node are defined as the objective function for the path by calculating the sum of all objective functions imposed on the path along the trellis. When several branches gather at a particular node, choose the one with the least path objective function associated with each branch and discard the others. This process is repeated for all nodes in the same column. The nodes in each column after this process are associated with a group assignment method such that the sum of the calculation results of Equation 19 for each group is smallest when the nodes are reached. Repeating this process K times for each column determines how K groups are allocated to each node in the final Kth column. In the final K-th group 503, the smallest objective function is selected to find the optimal path, and the number given to the nodes passed by the optimal path is obtained. Since these numbers represent the number of the last element of each group, the elements with the number of the last element of the group are collected from the number immediately after the number of the last element of the previous group to be the elements of the group. Since the optimal path has K values, the method can be repeated to determine K groups, and the grouping is terminated by dividing each user into such groups.

Next, even when the number of groups is not determined, elements of the group can be assigned while determining the number of groups. In order to use this technique, a trellis as shown in FIG. 6 is defined.

Unlike in Fig. 5, each node 601 is given the same number in the row direction. This means that groups without elements are allowed in each group. When you have finished assigning a group, you can get the final group by excluding the group with zero elements.

Thereafter, the hybrid interference cancellation unit 304 performs hybrid interference cancellation on at least one or more groups formed through the above method, thereby preventing overall detection performance deterioration.

The drawings and detailed description of the invention are merely exemplary of the invention, which are used for the purpose of illustrating the invention only and are not intended to limit the scope of the invention as defined in the claims or in the claims. Therefore, those skilled in the art will understand that various modifications and equivalent other embodiments are possible from this. Therefore, the true technical protection scope of the present invention will be defined by the technical spirit of the appended claims.

As described above, the present invention can improve the processing performance of each basic interference cancellation unit by dividing a group of users to be processed in the basic interference cancellation unit of the hybrid interference canceller by using a grouping technique based on dynamic programming. have.

In addition, if power control is not perfect, performance may be degraded by large signal size differences among users belonging to the same group, so grouping processing is added before using the hybrid interference cancellation technique, but the objective function for dynamic programming The calculation is simple, so there is little additional hardware cost. Therefore, the grouping technique can be easily implemented in software.

Claims (15)

In the mixed multi-user interference cancellation method for removing interference between a plurality of user signals, a) when a plurality of user signals are received from the outside, calculating powers of the received user signals, and numbering the calculated signal powers in order of magnitude; b) sorting each given user number in descending order; c) respectively forming at least one user group such that each signal power according to the sorted user numbers is less than a center value or an average value within the same group; And d) performing parallel interference cancellation on each user signal in the formed group, and performing sequential interference cancellation between the formed groups. Mixed multi-user interference cancellation method comprising a. According to claim 1, Step c) is Constructing trellis consisting of nodes and branches to which the aligned user number is assigned; And Forming at least one user group including at least one user number from the starting node to the arriving node along the branch of the configured trellis; Mixed multi-user interference cancellation method comprising a. The method of claim 2, The user group forming step, Forming a group from the user number added to the user number assigned to the starting node to the user number assigned to the arrived node; Mixed multi-user interference cancellation method comprising a. The method of claim 3, wherein Step c) is Calculating an average signal power of each of the formed user groups to define an objective function of each branch; And Calculating, for each branch, the sum of all objective functions imposed on at least one path by which the branches gathered at a particular node; Selecting a branch having the smallest sum of the calculated objective functions as an optimal path; And Collecting a number assigned to each node in the selected optimal path to form a user group Mixed multi-user interference cancellation method comprising a. The method of claim 4, wherein The objective function defined above is Wherein s _i is the number of user sets and m _i is the average value of signal power in the user set. According to claim 1, Step c) is Constructing trellis consisting of nodes and branches to which the aligned user number is assigned; And Forming at least one user group that does not include at least one user number from a node starting to a node arriving along the branch of the configured trellis; Mixed multi-user interference cancellation method comprising a. The method of claim 6, Step c) is When selecting an optimal path using the formed user group, the mixed multi-user interference cancellation method characterized in that the user group that does not contain any one user number is excluded. In the mixed multi-user interference cancellation apparatus for removing interference between a plurality of user signals, A communication path estimator configured to calculate power of each of the received user signals, and to number the calculated signal power in order of magnitude when a plurality of user signals are received from the outside; An alignment unit for sorting each of the given user numbers in descending order; A group calculator / former for respectively forming at least one user group such that each signal power according to the sorted user numbers is less than a center value or an average value within the same group; And Each user signal in the formed group performs parallel interference cancellation, and the hybrid interference canceller performs sequential interference cancellation between the formed groups. Mixed multi-user interference cancellation device comprising a. The method of claim 8, The group calculation / forming unit, After configuring the trellis consisting of the nodes and the branches to which the sorted user numbers are assigned, respectively, at least one user number from the node starting along the branch of the configured trellis to the arriving node is included. Mixed multi-user interference cancellation apparatus characterized in that it forms at least one user group. The method of claim 9, The group calculation / forming unit, Mixed user multi-user interference elimination device, characterized in that formed as a group from the user number added to the user number assigned to the starting node to the user number assigned to the arrived node. The method of claim 10, The group calculation / forming unit, The objective signal of each branch is defined by calculating the average signal power of each formed user group, and the sum of all the objective functions imposed on at least one path of the branches gathered at a specific node is calculated for each branch. Mixed multi-user interference cancellation device. The method of claim 11, wherein The group calculation / forming unit, And selecting a branch having the smallest sum of the calculated objective functions as an optimal path, and then collecting a number assigned to each node in the selected optimal path to form a user group. The method of claim 8, The group calculation / forming unit, After configuring the trellis consisting of the nodes and the branches to which the sorted user numbers are respectively assigned, at least one user number from the nodes starting along the branches of the configured trellis to the arriving nodes is included. Mixed multi-user interference cancellation device comprising forming at least one user group that does not exist. The method of claim 13, The group calculation / forming unit, When the optimal path is selected using the formed user group, the mixed multi-user interference elimination device, characterized in that excludes the user group that does not contain any one user number. In the recording medium comprising a hybrid multi-user interference cancellation method for removing interference between a plurality of user signals, a) when a plurality of user signals are received from the outside, calculating power of the received user signals, and numbering the calculated signal powers in order of magnitude; b) sorting each given user number in descending order; c) a function of respectively forming at least one user group such that each signal power according to the sorted user numbers is less than a center value or an average value within the same group; And d) a function of performing parallel interference cancellation on each user signal in the formed group, and performing sequential interference cancellation between the formed groups The recording medium in which the program is stored.