TWI260868B - Method for describing a propagation behavior of a communication signal which is transmitted by a base station in a communication network and computer-readable data media including program code means for computer program - Google Patents

Abstract

In the method for describing the propagation behavior of the communication signal, calibration measurements are carried out at selected positions in the communication network. A model for the propagation behavior is determined using the calibration data. In this case, the modeling takes place using a Gaussian process which shows a measured physical property of the communication signal depending on location information or position information (""forward model"").

Description

1260868 IX. Description of the Invention: [Technical Field] The present invention relates to a model for generating a communication signal transmitted by a base station in a communication network. Radio communication systems, such as wireless local area networks, Bluetooth's Pan-European Digital Mobile Communications (GSM), Global System for Mobile Communications (UMTS) or Pan-European Digital Radiotelephone System (DECT) are the most used On the field. These are everywhere in industrial production facilities and office environments or in the health industry. The transmission characteristics of the electromagnetic field generated by the communication device radiating the communication signal substantially determine the performance characteristics of the communication device with respect to the coverage area, availability and transmission rate. First, the radio network operators are concerned with confirming the distribution of field transmission characteristics or signal characteristics, such as electromagnetic field strength, phase, transmission time, wave vector, bit error rate, signal noise ratio, etc. The network can confirm the system characteristics required for quality assurance after the network is installed, or diagnose the error status during the operation of the equipment. Furthermore, 'network service providers are concerned with how they can provide venue-related services to their customers. In this regard, the location of the receiving device must be known. Since only the data generated during normal normal network operation should be used for the estimated side of the location, it is also suitable here to consider the characteristics of the signal. [Prior Art] The methods and procedures disclosed in the prior art mainly relate to terminals -6 - 1260868, such as DECT mobile parts or PDAs, and notebook computers equipped with wireless area computer networks. In some methods, such as those shown in [1], regionalization is based only on the network topology. In this case, the location of the terminal is determined by reference to the base station to which it is connected and the connection record of the terminal. The accuracy of this method will be limited because the base station to which the environmental terminal is connected will have a very large area to be identified as a possible venue. Other known methods attempt to estimate the position based on the received field strength of all available transmitters. In this case, a detailed physical model is sometimes used for radio wave transmission. However, detailed information about the environment must be known for this purpose. [2] revealed the use of knowledge about the electromagnetic transmission characteristics of various walls within buildings. However, such knowledge is generally not readily available. Therefore, under normal circumstances, an intensity map will be made based on the transmission model of the wave transmission, and then used in the regional limitation. One point estimation of the receiver position is usually implemented according to the field intensity map [3], [4]. [5] describes a one-way, stochastic nonlinear filter method used to estimate the position of a DECT mobile phone. A transfer model, in this case a statistical nonlinear model, is also generated or used here as the basis for estimating the position. In most of the model-based procedures disclosed in the prior art, models for constructing transmission behavior occur with reference to calibration measurements, wherein the physical variables that characterize the transmission behavior, such as the aforementioned field strength, are at predetermined locations. 1260868 (corrected position) is measured. The model used to describe the transfer behavior is determined by using the corrected positions and the transmitted variables measured at these locations. One of the disadvantages of the above-described model-based procedure is that many correction positions must be measured to obtain a model that is sufficiently accurate about one of the transmission behaviors, and one of the positional specifications that is sufficiently accurate. SUMMARY OF THE INVENTION Accordingly, the present invention is directed to the problem of relating to a specification-related procedure when generating a model of transmission behavior, wherein the procedure requires less calibration of the corrected position. The method and computer program for achieving the purpose include a code device and a computer program product for describing a communication signal transmission behavior, the signal is transmitted by a base station in a communication network, and reference is made according to related independent items. Characteristics. In the method for describing the transmission behavior of a communication signal, a physical property of the communication signal is measured at a selected location in the communication network, the physical property being combined with the associated location in each case. In this way, the physical properties constitute a characteristic of the communication signal transmission behavior. One of the transmission behaviors can be determined using the selected location or using the corresponding location information or location information associated with the selected location and the associated physical properties of the communication signal, which describes the transmission behavior. In this case, the model is generated using a Gaussian program that displays the measured physical properties ("forward model") that are dependent on the location information or location information. In order to distinguish between various models, note that in the case of the so-called forward model 1260868, the physical properties of the communication signal are described in terms of a location or a distance in the communication network. A regression or inverse model describes the location in the communication network based on the physical nature of the communication signal. One of the main advantages of the present invention is that the use of a Gaussian program to model the transmission behavior can substantially reduce the correction measurement with only minor errors in accuracy. These can be specified using, for example, a "design method" [2]. The computer program contains a code device that is structured such that when the program is executed on a computer, it can describe the transfer behavior according to all the steps of the method shown in the request. The computer program product contains code means stored on a machine readable medium which is structured such that when executed on a computer, all of the steps shown in the method of the present invention can be used to describe the transfer behavior. The computer program includes a code device having a structure such that when the program is executed on a computer, the model can be constructed according to all the steps shown in the method of the present invention, and the computer program product includes the code device and can be stored in the machine. The read medium is structured such that when the program is executed on a computer, the model can be constructed in accordance with all of the steps shown in the method of the present invention, both of which are particularly suitable for performing the method of the present invention or One of the derivation measures, as explained below. The preferred derivation of the invention is derived from the accompanying claims. The derivation described below is related to the implementation of the method and software. The invention and the following derivation can be implemented in software and hardware, for example, a special circuit is used for the 1260868. The invention or its derivation described below can also be implemented by a storage medium having stored thereon a computer program, the invention or a coded device derived therefrom. The present invention or various derivations may also be implemented by an electric power, which includes a storage medium on which the computer program or the derivation code device is stored. In a communication network, for example, a communication signal of a radio network is transmitted by signal packet, or a signaling device (mobile station), such as a mobile phone, and an omnidirectional antenna for transmitting or receiving. Or a multi-sector antenna burst. Depending on the nature of the transmitted or radiated communication signal or signal packet, several distance-dependent parameters can be determined, which are one of the distance-dependent, or distance-dependent, parameters that determine the radiation or signal characteristics of the (signal) transmitter. A field strength of a communication signal or signal envelope 'one phase, one wave vector, one bit error rate or signal noise ratio. The field strength of a radiated communication signal has a natural dependence on the distance between a transmitter and a platform, thus providing information on the transmission behavior (transmission characteristics) and is particularly suitable for inventing and using a model of a Gaussian procedure. In a best derivation case, the calibration measurement is pre-applied, wherein the physical properties of the points are measured and the computer readable is included to execute the brain program product, including during execution, including The so-called mobile platform, for example, a [measured] between each other can be used as a transmission system, for example, a transmission time, that is, (pushing) based on the transmitter in accordance with the construction Set the correction point. The choice of calibration point -10- 1260868 can be used by an optimal "design" method or grid method, such as the method disclosed in [2] to specify a hexagonal grid. Larger communication networks have a variety of or composite base stations under normal conditions, each radiating a communication signal. In this context, it is quite suitable to generate a separate transmission model for the communication signals of each base station or each base station. In the case of more than one base station, the model is made in accordance with the procedures of the present invention and can serve as a basis for several applications in a communication network. These models may therefore be used for planning and/or device and/or operational and/or error state diagnostics and/or communication network quality assurance purposes. The transmission model made in accordance with the present invention may also be used in the localization/positioning of at least one mobile communication device in the communication network, the mobile communication device being programmed to receive communication signals and/or receive a plurality of communication signals. In a communication network containing several base stations, when a mobile communication device is regionalized or specified to be executed, the physical properties of the measurement must be specified, such as the field strength, which can be approximated at the relevant base station. It is specified in the Gaussian program model. The location that the mobile device must specify is generated by specifying the point/location that has the greatest likelihood. When specifying a location, the basic idea is to reverse the Gaussian program model in this case. When making a model of the transfer behavior, the forward model is generated as described above. In the process of position specification, the model will be reversed so that in the inverse model, the position can be displayed according to the possibility of physical properties. The invention, or a model made in accordance with the transmission behavior of one or more of the base stations of the present invention, is particularly suitable for use in a 1260868 environment of a digital cellular mobile radio system, such as a GSM/UMTS network, and for use in GSM/UMTS The telephone (mobile phone) is regionalized. When the present invention is applied, only the data available to the mobile phone is used in this case, thereby avoiding the need for high cost changes to the mobile station in the G S network or the G S network. The invention is also suitable for use in the context of other digital cellular radio systems, such as WLAN, Bluetooth based networks or a DECT network, and for regionalizing, for example, a DECT mobile phone. The invention is particularly well-suited for environments that are detrimental to signal transmission conditions, such as heavy noise or reflected signals, shielding or obstructing base stations, and internal room trends. Under such restrictive conditions, creating a physically accurate model would be impossible or extremely difficult. A typical implementation of the present invention will be described in detail below and illustrated by way of illustration, in which: [Embodiment] The description/process positioning system (GPPS) is based on the correction using one of the calibration measurements, and the following description relates to the communication network ( One of the mobile communication devices, such as the DECT network here, and is from the Gaussian Program Model (GPM), where the signal from the base station of the communication network and the field strength at a predetermined location, such as a corrected location, It is measured in the communication network (Fig. 1, 1 1 〇). The Gaussian Program Model (GPM) is adjusted for calibration measurements (Reference Point 1) (Figure 1, 000). In this regard, proper selection of the core function of the Gaussian program model is quite important. The core function of the Matern level is used here. 1260868 The adjusted Gaussian program model is then used for positioning (reference point 2). At this point, the approximation of the Gaussian program model (Fig. 1 '13〇) and its optimization (Fig. 1, 140) are determined. Also described is the procedure for the optimal selection or alignment of the corrected position (Reference Point 3) (Fig. 1, 100). 1. Determine the Gaussian program model based on the measured field strength. The GPP S included in the description is based on the transmission characteristics of the communication signals or the probability model or statistical model of the signal strength/field strength of the individual base stations. The models used here are based on Gaussian models or Gaussian program regression (GPR) and are often used to solve nonlinear regression problems in Bayesian systems [12, 9]. The procedure or model described below produces a signal or field strength for the (selected) base station. This program is analogous to all base stations. The consideration is a set of N calibration measurements, where the field strength yi of the selected base station (normally in dB) is at the known position in the communication network Xi, i = l, ..., N Measure. GPR generally requires an unknown function f: /1 is mapped by y^fUd + ei, with independent Gaussian noise ei, and the variation is a 2 . The basic model here assumes that f(Xi) is a Gaussian program, which means that the function f(xi) has a Gaussian distribution at point Xi under the condition that the mean is 0 and the covariation matrix is K. . K itself is derived from the core (covariation) function K(·, ·), where

Ki,j = k(Xi, Xj). Gaussian distribution of 1260868 based only on a number of corrected or corrected positions t Gaussian program (GP). Use the following relationship: v(t) = (k(t, χχ), . . . 7 k(fc/ χΝ))Τ (2.1) y ~ (yir · · · / υν)τ (2.2) q = k + a2i (2,3) The preset GCM can be generated for several correction positions t: E(f(t)|D) = V(t)TQ-ly (2.4) The variation is: var (f(t)[D) = k(t, t) - v(t)TQ-1v(t) · (2,5> These relations can be found in the basic introductory work on Gaussian procedures [1. 8,12,9]. To be fixed

The parameter 0 of the noise variation and the parameter 0 of the kernel function K are therefore quite heavy. These can be maximized by the logarithmic approximation of the training/correction data according to the model parameters: 62, Θ = arg ma:x(- logdet Q - yTQ - 1y) · (2.6) σ2, θ

Matern core functions When using GPM, the appropriate choice of core (common variation) function is quite heavy. The core function describes the correlation between the functions of two points. A common choice consists of a GP s with a squared kernel and a structure of: k(x, xT) = exp (-v|x - xl However, from [1 0, 6], the structure of such a core function is a random program. The environment is unnatural, and the result exemplifies the path as infinite smoothing, that is, if the covariation function initially has an infinite number of derivatives. -14- 1260868 In this case, therefore, the Matern-level covariation/core is utilized. The function [1 〇] allows parameterization of the continuity of the smoothness of the exemplified path by the parameter v. The experimental example shows that the GPM with the kernel function can make a hypothetical variation from the equation (2·5). Actual estimate.

The function form of the Matern core is: k(xfX〇= Mv(z) = Kv(2Vvz)· (2.7) • Γ(ν) where Γ (ν) is the gamma function, and Kv(r) is the correction of the quadratic ν The function, and z2 = Σ 1 wj(xj - x»j) 2 has an input scale length Wi. The parameter ν indicates the smoothness of the illustrated path ("fractal dimension") and can be estimated by equation (2.6). A Matern core, adjusting the GPs effective solution formula (2.6) requires the Matern core function formula (2.7) for all the parameters v, w. The correlation application of the numerical gradient is disclosed in, for example, Π 〇], which requires a Bayesian function. Multiple evaluations result in a large number of calculations. The analytical function of the derivative function used here is as follows: = Γ(ν)ψ(ν) (2.8) ~f~ =-all (κν一l(z) + Kv+ l(z))' (2.9) where Ψ (ν) is a multi-gamma gamma function (so-called P si function). Since there is no known closed form of the Bayesian function kv(z) with respect to the derivative function of the order ν, Here, in the approximate manner DKv(z) = ef1 (Kv+e(z) - Kv(4). 1260868 From this estimate, the gradient of equation (2.7) can be indicated as follows: M = ^ Mv(2) - (Kv_l(2) ^) + KV+ 1(2V^)) θΜ ^ = Μν(ζ)^ + log(Vvz) - ψ(ν) ( ζ Γ(νΓ Γ (κν_1(2Λ/νζ) + Kv+1(2Vvz)) + DKv(2> /vzjj . (2.10) According to the above equation, the derivative function of equation (2.6) can be calculated using the standard matrix geometry with respect to the model parameters σ2, ν, w. In the basic introductory work Gaussian program [11, 8, The required relationship is described in 12, 9]. Specifying the field strength model The following describes how the GPM for the signal transmission of the selected base station is generated. It is assumed that the signal transmitted by this base station is already at the correction point Xi, i = l,... N measurements are measured at the 。. In this case, the measurement points are considered in the second dimension, and it should be noted that the described procedure applies equally to the point of the third dimension.

Starting from the correction data D = {xi, yj N i = l of the selected base station, the following steps will be implemented: 1. The estimated position of one of the base stations is obtained by selecting 3 correction points Xi, It has the highest field strength number 値Yl and constitutes its focus. The location of the base station is usually not recorded. In rare cases, however, such location data exists and can be used in place of the above estimates. 2 · To obtain the GPM's uniform function, a linear model will be adjusted to accommodate the measurement, where the one-to-five level is chosen as a function of the Euclidean distance to the base -16-1260868. The field strength 値 if indicated in dB値 will be used directly as if it were already in logarithmic scale. Therefore, the following conforming transmission rules will be included in the model: When the signal with intensity 1 is received at the base station, the receiving strength at the base station-distance d is e X p (-d)' and the original measurement is The uniform function is decremented. 3. Equation (2.6) provides the best model parameters, such as the variation σ2, the Matern smoothing parameter v, and the input scale length Wi. Figures 2a and 2b show an example of a GPM with raw calibration data and smooth data obtained by possible GPM. The smoothed data presents a structure that is not visible from the original measurement data, such as 2 zones extending to the left and right sides of the base station. 2. Decide to use one of the GPM locations. Determining a location based on the aforementioned GPM will be explained below. The following assumptions are known by correction: - C correction measurement

- at the correction point Xi, i = l, ..., C - for the B base station - with the receiving field strength c i, j at the base station j, j = 1, ..., B where X j. - j = 〇 ‘If the signal of base station j cannot be received at location Xi.

Ci indicates the field strength vector of all signals that can be received at location Xi. During the test phase or application phase, when the mobile communication user is located at an unknown location that must be indicated, the field strength of the base station receivable at the site will be measured. -17- 1260868 S indicates the vector of the receivable field strength at the location that must be specified, and the s j component of the vector S is the field strength received by the base station h. Use "nearest neighbor" to determine a position (NNLoc) In the case of NNLoc, the vector s of the position that must be specified is compared with the corrected measurement c i,i = 1,...,i. Each "neighbor" or correction point for that location will be weighted based on the extent to which the measurement is consistent with the associated correction measurement c i . These rights will be considered for all receivable base stations and current field strengths. Based on the known and best match correction position and associated weights, the location t that must be specified for the mobile communication user can be estimated by interpolation. The POS program positioning system (G P P S ) is determined by the GPPS to determine a location based on the GPM described above. In this case, using the GPM of the individual base station, the proximity of the receiving field strength of the base station can be formed at the location t must be specified. Use the correction data to pull i' ci, ji is bit '· · · / C}, je {1, · · ♦ , B} The relevant GMP Mj is formed for individual base stations, and the model Mj is based on these correction points j. Data Dj, where the base station j can be received:

Dj= { (Xi,Cj,j) ·· Ci,j 关 〇 }. The model considered in the application phase is only a model of the base station that can be received at the location that must be specified. The approximate equation for the acceptable field strength at the location t that must be specified is derived as follows: (3.1) L(t) = Π p(sj|^jft). j:Sj off 0 1260868 at point t of the data Dj The assumed distribution of GPM is specified by ptj|Dj. This assumed distribution is a one-dimensional Gaussian distribution with uniformity and variation according to the equation (2.4). It is now illustrated to search for GPPS points where the common near-field strength of the received field strength is the maximum 値. The position t that must be specified is obtained by L (t) for the optimization or maximization of t. t = arg max L(t) = arg max Σ log p(sj|Djt). (3·2) t t j:Sj off o This optimized operating system L(t) is solved with respect to the gradient of t. The joint gradient information of equation (3.2) can be solved by a standard number optimization method, such as "scaling conjugate gradient", and the position t in the search can be estimated accordingly. Figure 3 illustrates the relevant formula. Another approach is to solve L(t) by the grid method, where the calculation is performed at grid point L(t) and the maximum 値 has been indicated. The GPPS can be clearly stated as follows: The high field strength of one of the signals received for a particular base station indicates that the user of the mobile communication is within a very small radius of one of the base stations. It is therefore known that a very low field strength indicates that the user is located within a very large radius or range. The superposition of these individual site estimates provides the final position t of the estimate. The best choice for calibration points (Fig. 1, 100) A suitable positioning system, such as GPPS, should be implemented with a minimum number of calibration points, so that the less work required for calibration is as good as possible. At the same time, the calibration point should optimally cover the relevant area of possible regionalization. -19- 1260868 Several methods can be known from prior art [7]. The method as described in [7] is selected here for the best choice of correction points, wherein the result of the method is a hexagonal grid of correction points. This manual refers to the following publications: [1] Peyrard, F, Soutou, C., Mercier, JJ: Regionalization of Mobile Stations in a WLAN, Proceedings of the 25th Annual IEEE Conference on Local Computer Networks (LCN'00), Tampa, Florida (2000) 1 36- 1 42 [2] Hassan-Ali, M., Pahlavan, K.: A new statistical model for site-specific indoor radio transmission predictions, based on geometric optics and geometric probabilities, IEEE Transactions on Wireless Communications 1 (2002) 112-124 [3] Howard, A., Siddigi, S., Sukhatme, GS: Regionalization experiments using wireless tera-networks, Proceedings of the 4th International Conference on Field and Service Robotics , Japan (2003) [4] Bahl, P·, Padmanabhan, VN: RADAR: — RF-based user location and tracking system in a building, Proceedings of IEEE INFOCOM 2000. Volume 2., Tel Aviv, Israel (2000) 775-784 [5] Rauh ? A., Briechle, K., Hanebeck, UD? Bamberger, J.5 Η of fm an n, C · : Regionalization of DECT mobile phones, based on a new nonlinear filter Technology , Proceedings of SPIE Bd. 5084, Aero Sense Symposium, Orlando, Florida (2003) -20- 1260868 [6] Gneiting, T. "Small basic correlation function", journal of Multivariate Analysis, 8 3 (2) : 493 - 5 08, 2002 [7] Hamprecht, FA· and Agrell, Ε· “Development of Materials: Spatial Sampling Design and Sub-Collection Selection”, “Τ·Ν·Cawse, ed·, Development of Materials for Combined and High-Production Experimental Design, John Wiley & Sons, 2002 [8] Mackay, DJ· "Introduction to the Bureau of Procedures" 'CM Bishop, ed·, Neural Network and Machine Learning, ν〇1· 168, NATO Asi Series. Series F , Computer and Systems Science, Springer Verlag, 1998 [9] Rasmussen, CE "Gaussian Program Evaluation and Other Methods of Nonlinear Recursion", P h. D. thesis, University of Toronto, 19 9 6 [10] Stein, M. “Interpolation of Spatial Data, Theory of Kriging Method”, Springer Verlag, 1999 [11] Williams, CK· “Gauss Program”, Μ·Arbib, ed·, Brain Theory and Neural Network Handbook, MIT Press, 2nd edn., 2002 [12] W Illiams, C.K. and Rasmussen, C.E. "Recursive Gaussian Procedures", D.S. Tour etzky, Μ. C. Mozer, and M.E. Hasselmo, eds., Development of Neural Information Processing Systems 8, MIT Press, 1996. BRIEF DESCRIPTION OF THE DRAWINGS FIG. 1 shows a process for determining the position of one of the GMs using a Gaussian Program Positioning System (GPPS) according to an exemplary embodiment; FIGS. 2a and 2b are diagrams showing a GP Μ (Fig. 2a), It has been adjusted using the original calibration data, as well as the GMP chart (Figure 2b), which has been smoothed using data from the GPM; 1260868 Figure 3 shows the equation used to specify the derivative function for the position t must be specified . Exemplary Embodiment: A Gaussian Program Positioning System (GPP s) among communication networks (D E C T networks) of a plurality of base stations. [Component Symbol Description] 100 Indicates the best correction position. 110 Performing the calibration measurement and collecting the calibration data. 120 Using the Gaussian program, the transmission behavior is made to the previous model.

130 indicates the approximation 140 solves the optimization problem and determines the maximum approximation as the position that must be specified

-twenty two-

Claims (1)

...... 1 · ' '*·* ···· · * .,. 1260868 ^ ii -T,., .·» «*<-·········· -»-.*♦ 1- · .-· ..··*,.,》.·,. ····. -ί-'t , rm i ^ X. Patent application scope: 9 3 1 3 7 6 1 9 "Methods for describing the transmission characteristics of a communication signal transmitted by a base station in a communication network and a computer-readable recording medium containing a code device for a computer program" Patent (200) Revised December 5, 1. A method for describing the transmission behavior of a communication signal transmitted by a base station in a communication network, wherein - the physical nature of the communication signal is measured at a location selected in the communication network. 'The physical property is associated with the relevant location in each case, wherein the physical property is characterized by a communication signal transmission behavior, - the model for the transmission behavior uses the selected location or the use of the selected location and the communication signal Determined by measuring the corresponding location information or location information related to the physical property, and the model describes the transmission behavior, The characteristics are as follows: - The model making process is generated by using a Gaussian process, and the physical 丨 贝 ' 显示 其 其 其 其 其 其 其 其 其 其 其 其 其 其 其 其 其 其 其 其 其 其 其 其 其 其 其 其 其 其 其 其 其 其 其 2 The method, wherein for a plurality of base stations having a communication signal in the fg network in each case, the model for the relevant communication signal transmission behavior is determined in each case. The method of item 1 or 2, wherein the location of the selection is specified using a "design method". 4 - The method of claim 2 or 2, wherein the communication network is a radio network, in particular According to the wireless local area network or Bluetooth or GSM or DECT or UMTS radio network, and / or communication signal measurement physical 1260868 X nature is an electromagnetic field transmission characteristics, especially a field strength, ~ phase, ~ transmission time, A wave vector, a one-bit error rate, or a signal-to-noise ratio. 5. A method of claim 2 or 2, wherein the model or models are used in the communication network And / or set and / _ operation (c 〇mmissi ο ning) and / or error status diagnosis and / or quality assurance purposes. 6 · If you apply for the method of claim 1 or 2, where the model or The plurality of models are used to locate a location in the communication network and/or to specify a location of the at least one mobile communication device, the mobile communication device or the plurality of mobile communication devices being configured to receive the communication signal and/or the number 7. The method of claim 6, wherein at least for the Gaussian process model, the communication signal can be received at a location that must be specified at one of the mobile communication devices, with respect to the location that must be specified The likelihood of measuring the physical properties of the 'correlated Gaussian process model' is specified in each case - where the location that must be specified is determined by optimization or maximization possibilities. 8 - A computer-readable recording medium containing a program code device for a computer program for performing all the steps of the method of transmitting the behavior as described in the first paragraph of the patent application when the program is executed on a computer. 9. A computer readable recording medium as claimed in claim 8 wherein the program is stored on a computer readable data medium. 1 0 - a computer-readable recording medium containing a program code device for a computer program, which is used to implement all the steps of the method for transmitting the behavior as described in claim 1 of the patent application when the program is executed on a computer And the code device is stored in a machine readable recording medium. -2-