TWI320854B - Magnetic data processing device, magnetic data processing method, magnetic data processing program product, and magnetic measurement apparatus - Google Patents

Abstract

In a magnetic data processing device, an input part sequentially inputs magnetic data outputted from a two-dimensional or three-dimensional magnetic sensor. The magnetic data is two-dimensional or three-dimensional vector data that is a linear combination of a set of fundamental vectors. The magnetic data processing device stores a plurality of the inputted magnetic data as a data set of statistical population in order to update an old offset of the magnetic data with a new offset. An offset derivation part derives the new offset based on the old offset and the data set of statistical population under a constraint condition that the new offset be obtained as the sum of the old offset and a correction vector.

Description

IX. Description of the Invention: [Technical Field] The present invention relates to a magnetic data processing device, a magnetic data processing method, and a magnetic data processing program, and more particularly to a method for correcting two-dimensional and three-dimensional magnetic sensing One of the techniques of offset. [Prior Art] A conventional magnetic sensor that is fixed on a moving body (e.g., a mobile phone or a vehicle) detects the direction of the earth's magnetic field or geomagnetism. The magnetic sensor includes a set of magnetic sensor modules for detecting scalar components of magnetic field vectors in mutually orthogonal directions. The magnetic data outputted from the magnetic sensor is composed of one output combination of the four magnetic sensor modules, and thus the magnetic data is used as a linear combination of mutually orthogonal unit vectors (basic vectors). The direction and number of the magnetic data correspond to the direction and number of a magnetic field detected by the magnetic sensor. When specifying the direction or number of the earth's magnetic field based on the output of the magnetic sensor, it is necessary to perform a procedure for correcting the output of the magnetic sensor in order to eliminate the inherent temperature characteristics of the magnetic body due to the moving body or the magnetic sensor. The measurement error caused. One of the calibration procedures is called an offset and the procedure used to derive the offset is called calibration (see, for example, International Patent Publication No. 2004_003476). When the external magnetic field strength is zero, the offset is also the vector data and is defined as the magnetic data output from the magnetic sensor. Such measurement errors are eliminated by subtracting the deviation from the magnetic data output from the magnetic sensor. In a two-dimensional (2D) magnetic sensor, the offset corresponds to a - position vector of the center of the circle on which a magnetic data set is distributed. However, in fact, the distribution of the magnetic bead set of the 114821.doc 1320854 output self-pinning magnetic sensor does not form a perfect circle. The reason is that the output of the magnetic 4 & sensing 11 module inherently has a measurement error following the Gaussian distribution, and a magnetic field measured by the 2D magnetic sensor changes during the period in which the statistical data set of the statistical data set is stored. Since there is virtually no completely uniform magnetic field, a calculation error occurs during AD conversion. A statistical matrix data set required to derive the offset of the magnetic I1/sensor output, • a moving body including the 2D magnetic sensor fixed thereon is rotated, and φ causes the 2D magnetic sensor to rotate around The axis rotates parallel to a direction perpendicular to the direction of orthogonal sensing of the magnetic sensor module. In order to move a moving subject (e.g., a mobile phone or a vehicle that can be moved three-dimensionally) in this manner, it is necessary for the user to intentionally operate the moving subject to move in this manner. Therefore, the offset-pushing algorithm for deriving a shift of a 2D magnetic sensor is designed based on the assumption that the user is notified of the start of calibration and the user appropriately operates the moving subject. However, it is cumbersome and complicated for the user to perform the calibration operation. In the standard calibration method, a binary statistical decision is made to determine whether a reliable statistical parent data set has been stored. Therefore, when the user has not performed the calibration operation correctly, the calibration fails without storing a reliable statistical parent data set. This requires the user to repeat the operation to store a reliable statistical parent asset. A conventional three-dimensional (3D) magnetic sensor fixed to a moving body (e.g., a mobile phone or a vehicle) detects the direction of the earth's magnetic field. The magnetic sensor includes three magnetic sensor modules for detecting scalar components of the magnetic field vector in three orthogonal directions. The magnetic material 114821.doc 1320854 outputted from the 3D magnetic sensor is composed of one output combination of the three magnetic sensor modules, and thus the magnetic data is taken as one of the mutually orthogonal unit vectors (basic vectors). The 3D vector data of the sexual combination means that the direction and the number of the magnetic data correspond to the direction and quantity of one of the magnetic fields detected by the 3D magnetic sensor. When specifying the direction or number of the earth's magnetic field based on the output of the 3D magnetic sensor, it is necessary to perform a procedure for correcting the output of the 3D magnetic sensor to eliminate the magnetization or the magnetic sensing due to the moving body. Inherent temperature characteristics

The measurement error caused. One of the control values of this calibration procedure is called an offset. The offset is indicative of vector data of one of the magnetic fields caused by the magnetization component of the moving body as a function of the 3D magnetic sensor. Such measurement errors are eliminated by subtracting the deviation from the magnetic material output from the 3D magnetic sensor. It is possible to calculate the offset by obtaining the center of a spherical surface on which a magnetic data set is distributed. However, the distribution of 'actual' magnetic data does not form a perfect sphere.

The reason is =, following the Gaussian distribution, the output of the 3D magnetic sensor is inherently measured! 6 difference, measured by the 3D magnetic sensor - the magnetic field is stored during the calculation of the bias, the required magnetic data - during the period, due to the fact that the magnetic field is not uniform - and the calculation error will occur until the 3D domain The output of the sexy sensor obtains a digital value. Sexual data and through the storage V 2 ... store a large number of 磁性 magnetic data statistics program to derive the posture of the name or 2 = method _ ' unless the user intentionally changes the mobile j material set, and: more: partial update bias After shifting a magnetic 1 shift* to appear, it takes a longer time to get a new offset of I14821.doc 1320854. Usually, the magnetic data is stored only in two-dimensional distribution, and the apparent three-dimensional posture change of the magnetic sensor fixed on the vehicle is relatively rare. Thus, it is not necessary to wait until a magnetic material uniformly distributed on a spherical surface is stored in order to accurately update the offset of the magnetic sensor fixed to the vehicle.

International Patent Publication 2005-061990 has disclosed an algorithm that corrects the offset even if the distribution of the magnetic > material set is one dimension. However, due to its complexity, it is difficult to implement a procedure in accordance with the algorithm described in International Patent Publication No. 2-05-061990. SUMMARY OF THE INVENTION One of the goals of the dimensional (2D) magnetic sensor and the three-dimensional winter invention is to improve the usability of the (3D) magnetic sensor. Another object of the present invention is to provide a magnetic data processing device, a magnetic data processing method, and a magnetic data processing program, and a device capable of correcting an offset using a stored magnetic data through a simple program, regardless of statistical parent data. What is the distribution of the collection.

In a first aspect of the invention, a magnetic data processing apparatus for achieving the above object comprises: an input member for continuously inputting a magnetic material from (a) two magnetic waves (m), the magnetic material A 2D vector linearly combined as one of the -group-basic vectors; a storage component for storing a plurality of input magnetic data as a statistic parent data set, so as to use the new (four) to request the system - the old bias And the offset = the derivative m is derived from the new offset system as the old offset and the correction to ^ and the constraint, the base (four) old offset and the statistical parent data set to derive the new Offset, the correction vector is in the statistician 114821.doc 1320854

One of the basic vectors is a linear combination. In a second aspect of the present invention, a hair data processing apparatus includes an input member, a magnetic member for achieving the above object, and is used for continuously inputting from a three-dimensional (3D) magnetic sensor. The magnetic data is a 3D vector data linearly combined as one of a set of first basic vectors; a storage component for storing a plurality of input magnetic data as a statistical parent data set to use a new offset Updating an old offset of the magnetic data; and an offset deriving component for using the old offset based on the constraint that the new offset is obtained as the sum of the old offset and a correction vector And the statistical matrix data set to derive the new offset, wherein the correction vector is linearly combined with one of a set of second basic vectors defined in a major axis direction of the distribution of the statistical parent data set, and represents the correction vector The individual coefficients of the linear combination of the second basic vectors are weighted by a primary value ratio according to a distribution of the statistical base data set relative to the old offset Obtained by an individual coefficient of a position vector derived from the statistical parent data set without using the old offset, the temporary offset position vector being linearly combined with one of the second basic vectors. [Embodiment] 114821.doc • 9- The principle and algorithm for the first aspect of the present invention will be described in detail below with reference to FIG. The key point of this algorithm is that a magnetic data set distributed in the direction of a major axis with a larger dispersion is estimated to be the more prominent precursor element for updating the offset, and in the direction of the major axis with a smaller dispersion. A magnetic data set on the upper 77 cloth is estimated to be a less obvious element for updating the offset. The offsets corresponding to the old offset CG, the new offset C, and the temporary offset of the position vector of the end point "g" relative to the origin "〇" are linearly combined as one of a set of magnetic data base vectors 20 location information. That is, the offsets of the offsets are vector data represented within the xy coordinate system. The new offset C is derived based on the old offset c〇 and a magnetic data set stored to update the old offset c〇β using the new offset c as a store to use the new offset c. The master data set for updating the old offset magnetic data set may include a magnetic data set stored for a predetermined period of time and may include a magnetic data set including a predetermined number of magnetic data and may further include a magnetic data set The magnetic data set old offset including any number of magnetic data that has been stored at a particular time (e.g., when an offset update request is made) can be derived and can also be predetermined using the same method as the new offset c. Although the temporary offset is defined to be derived from the statistical parent data set without using the old offset c, this definition is introduced to define the constraints used to derive the new offset e and the temporary offset is actually Data that does not have to be derived. If the temporary offset is actually derived from the statistical parent data set without using the old offset Cq ', then the temporary offset is distributed near the center of one of the circumferences of the statistical parent data set. - position vector. However, if the statistician 114821.doc 2 (4) is not distributed in the vicinity of the narrow central angle of the parent-part of the statistical data set, then it is included in each element of the statistical-beauty set. An error will affect this to a large extent: the derivation result may result in a deviation from the true offset of one. For example, the 'consideration-statistical maternal data set is unevenly distributed near the narrow central angle arc and the statistical maternal data set is mutually orthogonal to the S-system uAu2, as shown in Figure i. In this case, due to the statistic: the variable of the volume data set is smaller in the direction of the 对应2 corresponding to the feature of the distribution corresponding to the smaller main value, so there is a higher probability derived from the statistic parent data set. A temporary offset is away from the true offset in the direction of the feature vector. On the other hand, in this case, since the variable of the statistical matrix data set is larger in the direction of the feature vector U1 of the distribution, there is a higher probability that a temporary offset system derived from the statistical parent data set is derived. It is near the true offset in the direction of the feature vector U2. Since the variables in the direction of the major axis of the distribution can be expressed as indices of the distribution using the principal values λ & λ 2 of the distribution, the apparatus estimates based on the ratio of the principal values λ ′ and λ 2 Statistical parent elements distributed in directions corresponding to the primary values, respectively. Specifically, first, a correction vector f as a new offset C relative to a position vector of the old offset Co and a temporary position vector g as a temporary shift relative to the position vector of the old offset C〇 may be It is defined in a coordinate system having coordinate axes α and β that coincide with the direction of the major axis of the distribution. That is, each vector of the correction vector f and the temporary position vector g can be a linear combination of one of the basic vectors of the principal direction of the five mysterious cloth. This point corresponds to the transition to the spindle value. If the component & 114821.doc 丄(10) 854 and fp of the correction vector f are weighted by the measurement of the corresponding main values U 丨 and u 2 of the distribution, the temporary offset is relative to the position vector g of the old offset 4 By deriving the component "and", it is possible to derive the correction by increasing the reliability of the statistical parent element in a direction with a larger spread and reducing the reliability of the statistical parent element in a direction with a smaller spread. Vector, however, such definitions of the correction vector £ and the temporary position vector g are also introduced to define a new offset c*

The constraints are used and the vectors of the correction vector f and the temporary position vector Lu actually do not need to be deduced. Deriving a new offset c by obtaining a new constraint c as the sum of the old offset Co and the correction vector factory as defined above, regardless of the distribution of the statistical parent data set, can be estimated in the distribution A magnetic data set in the direction of the main axis with larger dispersion is more statistically distributed to the parent element and distributed in the direction of the main axis with smaller dispersion - the magnetic retraction is less obvious when the parent element is not statistically significant. - (d) in this way

The exemplary technique of offset is to formulate the distribution as an "optimized question." Regardless of the calibration operation performed by the user, the device can derive the most likely new offset that can be derived from an actual execution operation and thus does not require the user to perform a predetermined operation. In the magnetic data processing apparatus of the invention, the constraint may be based on the fact that the ratio of the t-th main value to a larger main value is equal to or less than a predetermined threshold value, ff T m The weighting factor of the temporary position vector for the coefficient of the smaller main-sense plus m&-the basic vector in the direction of the main axis corresponding to the main values is zero. The apparatus discretely estimates the distribution I of the statistical dataset to discretely weight the temporary offset relative to the old offset based on the deviation. . Bits • These coefficients of the £ vector. Specifically, the weight of the coefficient of the temporary position vector § in the direction in which the value of the distribution is relatively small is. That is, there is no statistical maternal data set estimate in the direction that the value of the distribution is less than a critical value. In the magnetic data processing apparatus of the present invention, the (4) beam condition may be such that the coefficients of the correction vector are weighted by the coefficients of the temporary position vector, and the ratios of the principal values corresponding to the distribution of the statistical matrix data set are successively appealed. The value obtained by the weighting factor. * This device increases the substantial efficiency of the statistical maternal data set, since these weighting factors have a continuous correlation with the distribution. In addition, the apparatus simplifies the offset update procedure since the new offset can be derived from the distribution of the statistical parent data set without changing the program. In the magnetic data processing apparatus of the present invention, the individual weighting factors for the coefficients of the temporary position vector may be weighted by coefficients for the second basis vector in the direction of the major axis corresponding to the larger principal value. The factor is set to one to normalize. In the case where the estimation of the statistical matrix data set distribution other than the primary values of the distribution is introduced, it is not necessary to normalize the weighting factors by setting a larger weighting factor to one. For example, the maximum weighting factor may be based on the statistical matrix data set, based on the maximum distance between the magnetic data in the direction of the major axis corresponding to the larger principal value (ie, the main direction) and one has been derived as a circumference (having a vicinity) The ratio of the radius of the circle of the distribution of a part of the statistical maternal data set is set to be less than one. 114821.doc -13- In the magnetic data processing apparatus of the present invention, the magnetic parameter processing apparatus of the present invention minimizes the following 9 ^c) in order to realize the above-described objective offset derivation member derivable objective function f(c) -{Xc-j)T{xc .) is represented by qi=(qix,qiy)(i=〇, 1,2, where the magnetic data systems X" and "j" are as follows:

- R r qX 1 分丨7V丨- and W2~qf • · y =-2 Pregnancy 2 points 2 ~ and k.-g'fj » · · βΝ-\ 9n.\ -'R In this manual, all The vector is described as the transpose matrix of the row vector. The two = ^ quantity and the column vector system are in the respective transposition moments p car female side == to each transposed matrix [that is, using the form of () τ]. The new offset is derived by deriving the Γ beam condition as the distribution-optimization problem, as in the case of the 1 body = solution simple simultaneous linear square M m , which is described later in the body yoke example. That is, the new offset of the helmet can be derived from the statistical maternal data set. Included: The hardware resources of the components of the plurality of components in the device of the invention are functionally specified by their structure... The function is a combination of a hardware resource or a combination of these resources: 114821.doc 14 The function of each component is not necessarily achieved by a real I hip independent hardware resource. The present invention can be specified not only by a device but also by a program, a recording medium on which the program is recorded, and a method. In the case of technical deficiencies, the operation of the methods described in the patent application is not performed in the order stated in the scope of the patent application, but may be performed in any other order or simultaneously.

A plurality of specific embodiments of the first aspect of the present invention will be described in the following order. A. First embodiment [1. Overview] 1-1. Hardware structure 1- 2·Soft structure [2. Program] 2 -1. Overall process

2- 2. Buffer update 2 - 3 · Distribution estimation 2-4. Deriving a new offset through the optimization problem 2-5. Deriving a new offset when the distribution is two-dimensionally 2. 6 · When the distribution is substantially one Dimensional Derivation of New Offsets 2-7. Summary 第二 · Second Specific Embodiments • Overview • Distribution Estimation • Derivation of New Offset 114821.doc • 15-1320854 c. Other Specific Examples [i' Overview] 1-1. The hardware structure diagram 2 is a schematic view showing one of the appearances of a car 2 as an example of a mobile body to which the present invention is applied. The car 2 includes a 2-dimensional (2D) magnetic sensor 4 2 D magnetic sensor 4 detects the direction and intensity of a magnetic field by detecting the individual intensity of the magnetic field in two orthogonal directions (X, y) . A 2D magnetic sensor 4 constituting a part of the navigation system fixed to the car 2 is used to specify the traveling direction of the car 2. Figure 3 is a block diagram of a magnetic measuring device including a 2d magnetic sensor 4 and a magnetic data processing device i. The 2D magnetic sensor 4 includes X and y-axis sensors 30 and 3 2 ' which measure the X and y-direction components of one of the magnetic field vectors due to the terrestrial magnetic force. Each of the X and y-axis sensors 30 and 32 includes a magnetoresistive element, a Hall sensor, or the like, which can be any type of one-dimensional magnetic sensor (assuming it is directional) ). The X and y-axis sensors are fixed in the '3' and 32-series' so that the sensing directions are perpendicular to each other. The output of the beta X and y-axis sensors 3 0 and 3 2 is time-divided and input to a magnetic sensor interface (I /F) 22. After amplifying the inputs from the X and y-axis sensors 3 and 32, the magnetic sensor interface 2 2 analog to digitally converts the inputs. The digital magnetic data outputted from the magnetic sensor interface 22 is input to the magnetic data processing device through a bus bar. The magnetic data processing device 1 is a computer including a CPU 4A, a ROM 42 and a RAM. 44. The CPU 40 controls, for example, the overall operation of the navigation system. The ROM 42 is a non-volatile storage medium that stores a magnetic 114821.doc • 16·1320854 program for implementing the navigation system data processing program or various functions executed by the CPU 40. RAM 44 is a volatile storage medium that temporarily stores data for CPU management. Magnetic f-processing device The magnetic sensor 4 can be constructed as a single-wafer magnetic measuring device. 1-2. Software structure

Figure 4 is a block diagram of a magnetic data processing program. The magnetic data processing program 90 is stored in the ROM 42 to provide orientation information to the locator %. The position data is a 2D vector data 'which indicates the orientation of the earth's magnetic field. The magnetic data processing program 90 is constructed as a group of modules, such as a buffer management module 92, an offset derivation module 94, and an orientation derivation module %. The buffer management module 92 is a program portion that receives a plurality of magnetic data continuously outputted from the magnetic sensor 4 and stores the received data in a buffer for use in the offset update. The buffer management module 92 allows the CPU 40, the RAM 44, and the ROM 42 to function as input members and storage members. This buffer can be embodied not only by hardware, but also by software to embody a magnetic data set that is now stored in this buffer as a statistical parent data set. The offset derivation module 94 is a program portion that derives a new offset based on an old offset of one of the statistical parent data set and the offset derivation module 94 held by the buffer management module 92. The new offset to update the old offset ° offset derivation module 94 allows the CPU 40, RAM 44, and ROM 42 to function as an offset derivation. Since updating the old offset with a new offset causes the new offset to become an old offset, "old offset" is simply referred to as an "offset,• in the context of no misunderstanding. In fact, a shift for orientation data correction is set using a variable and the new offset is derived as a variable different from the variable. When the new offset is guided, it is set using a variable for orientation data correction. Therefore, the variable used for orientation data correction is the variable in which the old offset is stored. The azimuth derivation module 96 is a program portion that uses the offset maintained by the offset derivation module 94 to correct the magnetic material continuously output from the magnetic sensor to produce orientation data. Specifically, the azimuth derivation module % is used as the azimuth data to output the data '35 data including two components obtained by subtracting the components of the offset from the components of the magnetic data as the 2D vector data. The 疋 position 98 is a known program that specifies the position of the car 2 by automatic navigation. Specifically, the positioner 98 specifies the direction of travel of the vehicle 2 based on the orientation data and specifies the position of the "speed 2 relative to a base point based on both the direction of travel and the distance traveled. The screen displays the north and south maps on the screen by characters or arrows. The orientation data can be used not only for Firefly, East and West but also for display [2. Program] 2·ι·Overall flow Figure 5 is a flow chart showing a new offset derivation 窓..^ _ ., Yin Shouqian. When an offset update request has been made, the CPU & executes the buffer management module and the offset derivation module 94 to execute the sequence & ^ & The offset update request can be made at a predetermined time interval and can also be made by a clear indication from one of the drivers of the stalker. 2-2. Buffer update In the step of discussion, delete the magnetic data stored in the buffer (4), and store a magnetic data set (statistical parent data set) for deriving a new offset in the buffer 114842.doc. . Thus, in this procedure, a statistical parent data set for deriving the old offset is deleted. In step S102, a magnetic data input for deriving a new offset is stored in the buffer. When a plurality of magnetic data are continuously input from the magnetic sensor 4 without any change in the traveling direction of the automobile 2, the distance between the two consecutive input magnetic data (or values) is small. Storing a plurality of new magnetic data in a limited-capacity buffer wastes memory resources and causes unnecessary buffer update procedures. Furthermore, if a new offset is derived based on a new set of magnetic data sets, there is a possibility that an inaccurate new offset is derived based on an unevenly distributed statistical matrix data set. Whether it is necessary to update the buffer can be determined in the following manner. For example, if the distance between the last input magnetic data and the magnetic data stored in the buffer just before the last input magnetic data is less than a given threshold, then it is decided that the buffer does not have to be updated and the last input magnetic is discarded. The data is not stored in the buffer. In step S104, it is determined whether a specified number of magnetic data required to derive a precise new offset has been stored in the buffer. That is, the number of elements of the statistical parent data set is predetermined. Set a small amount of statistic dataset element to improve the response to the offset update request. The procedures of steps 81〇2 and sl〇4 are repeated until a certain number of magnetic data are stored in the buffer. 2 - 3. Distribution Estimation Once a certain number of magnetic data is stored in the buffer, the distribution of the statistical parent data set is estimated (S106). The distribution is estimated based on the main value of the distribution of 114821.doc 19 丄:>/)4. When the magnetic data set is expressed by the following equation (1), the main values of the distribution are eigenvectors of a symmetric matrix A using a center (average value) from one of the statistical matric data sets. The sum of the vectors ending with individual magnetic data is defined by equations (2), (3), and (4). Qi = Uix,qiy)(i=〇, 1,2,· . ·(1)

...(4) Since matrix A can also rewrite equation (5), matrix A corresponds to a multiple of a variable covariate matrix. • · · (5) Let λ丨 and λ2 become the eigenvalues of matrix a in ascending order. Let (1) and become the mutually orthogonal feature vectors that correspond to the person and the person 2 and have been normalized to size 1. In this embodiment, it is assumed that the matrix A is non-singular and the range of λ1 & λ2 is 114821.doc -20· 1320854 is λ丨> and 2> When the moment is smaller, the smaller eigenvector of the single A enters the 2 system zero, that is, the rank of the matrix A is one or less, and the matrix A is not considered, because the number of elements of the statistic sports material set is - or The division and the 豕 knife cloth are a perfect straight line. The eigenvalues of the eigenvalues must be zero or - positive real numbers, since the matrix A is defined by the matrix A is half positive definite moment (j car. 5 syllabary s ten parent data set of the public $ knife cloth based on smaller features The ratio of the value to the larger eigenvalue is 1:!/person 1 to estimate.

In step SH) 6, it is determined whether the distribution of the statistical parent data set is sufficiently one-dimensional. Specifically, the decision is affirmative when the following condition (6) is satisfied and negative when it is not satisfied. Λ2/λι> ΐ (6) When the statistical matrix is collected in a large extent along a specific circumference, the condition (6) is satisfied. If the decision in step S106 is negative, the distribution of the statistical parent data set is f-dimensional. When the statistical matric data set is unevenly distributed on the narrow circumference and the angular arc of the specific circumference, the distribution of the statistical maternal data set is substantially one-dimensional. 2-4. Deriving a new offset through optimization issues • An optimization problem for deriving a new offset will now be described. When the statistic parent-beat set includes two magnetic materials that do not exist on the same line, the circumstance of the statistic parent data set is assigned to the local system without the use of statistical techniques. The position of one of the centers of this circle is ton, ~) is obtained by solving the simultaneous equation (7) to obtain three equality constraints exist for two variables 'but equation (7) must have - answer, due to three flat 114821.doc •21 · 1320854 and other constraints are equal constraint redundancy

C 2 9\~~R 殳2殳2 ~ and its middle (7) 1 Λ^-1 7V /=〇τ <h • (8) 2 9〇~-jR.々'-R points 2 points 2 And βN-' qΝ_'- r (9) solution == using r" the simultaneous linear equation 〇. ) with -answer, then the distribution of the statistic parent data set on the circumference.

Xc=j- (10) However, 4t bandits right loss 2D magnetic sexy test 4 _ Tian ' 上 盔 祜 祜 + a journal benefit 4 inherent measurement error, then ..., to make Fang Wei (10 ) has a solution. The solution defined by the following - vector "e··俜引彳 equation (11) is over-statistical technology to obtain - plausible ^^l.doc -22- 丄 854 e=Xc'j ... (11) 6 Minimize "e"22 (ie eTe)"c" can be plausibly regarded as the center of the circumference of the distribution of the ten parent body. The problem of minimizing the value of 丨丨e丨丨z2 "c" is a problem for minimizing the objective function of one of the following equations (12). f(c) = {Xc - j)T [χ〇 _ Λ J 1 ··. (12)

2 5· § When the distribution is two-dimensional, the new offset is derived as shown in Fig. 6. 'When the distribution of the statistical maternal data set is two-dimensional, the statistical matrix data set is fully reliable as a whole, so the new offset is obtained by Minimize the objective function f(c) of equation (1) without deriving (S 1 G8) with any constraint of "C". When the tether is assumed to be non-singular in this embodiment, there is no need for a constraint to minimize the value of the objective function f(4), c"^ is equation (13).

C = {xTx)~lXTj (13) When the V, which satisfies the equation (9), is derived, the new offset (4) obtained by correcting the old offset in both directions is based on the statistic (four) material set. Do not use old offsets. Wait for the W虮-ST π遁 tribute set to use the old offset. - Algorithms for deriving based on statistical matric data sets without using old offsets can be implemented (as embodied here - using the various statistical techniques already proposed) and can also be 114821.doc • 23· 1320854 An algorithm using any statistical technique. • 2_6. Deriving a new offset when the distribution is essentially one-dimensional as shown in Figure 7, when the statistical matric data set is distributed over a narrow central angular arc of a particular circumference and thus When the distribution of the statistical maternal data set is one-dimensional (ie, linear), a new offset is derived (SUG) by determining the direction p of the corrected old offset as the main direction of the distribution. When the system is distributed near the specific line, the 'statistical matrix in the direction of the line•the distribution of the f-set is sufficiently reliable' and the distribution of the statistical matrix dataset in other directions is unreliable. In this case, In the direction other than the direction of the line, the old offset is not corrected, thereby preventing the offset from being updated based on unreliable information. When the statistical matrix data set is distributed near a specific line, the direction of the line is one correspond The direction of the feature vector of the larger feature value A1 (ie, the main direction) coincides and the direction of the feature vector h corresponding to the smaller feature value is perpendicular to the line. Therefore, in order only in the direction of the line Deriving a new value, • Minimizing a new offset c of the objective function of equation (12) is obtained under the constraints expressed by the following equation (14). u2 (c – c0) = 0 ... ( 14} The equation for solving the optimization problem of equation (12) under the constraint of equation (14) can be modified into its equivalent simultaneous equation using the Lagrangian multiplier method. When the number p is introduced and 1" is defined by the following equation (Μ), the simultaneous linear equation (16) of "X" is the above simultaneous equation. 114821.doc -24· c ;c =P .. (15) B3x =, where... (16) B3

A iTjl"A0 (17) (18) It should be understood from the above description that if the distribution of the statistical parent data set is substantially dimensional, the program for deriving the new offset in step S110 is used to solve the simultaneous connection. Linear equation (16). The answer "χ" must be uniquely specified, since the rank of matrix B3 must be 3. 2·7·Overview Now, reference will be made to Figures 6 and 7 using the space concept to illustrate steps S108 and H. If the statistical parent data is fully available, then the new bias is the old offset a and relative to the old one. Offset c. Only the statistical maternal lean set derivation (ie, the temporary position vector) τ machi position to the sum of 1 — is defined by the following equation (19). C ~ c〇+ g ... (19) The new offset position vector" is not the old offset "c0" and a correction vector, f" I14821.doc •25- and the correction to $"1 The basic vector in the direction of money - ... is characterized by the same linearity as 兮6". Therefore, corresponding to the center of the maternal data set based on the opposite side = individual reliability according to the position vector, v can be used in the corresponding knife on the main axis: upper: a vector Correction vector "f" The coefficient of the individual sin-weighted position vector v of the upward mutated maternal data set ^ and the obtained coefficient: Figure 6 is not the steps performed when the distribution of the statistic parent data set is sufficiently two-dimensional S108's Cheng φ, vouchers, ""r is obtained from the objective function of the simplification equation (12) in the private sequence without any of the above constraints. ", this program can be considered as a new bias. Move "c" as the old offset. . Executed with the constraint vector "f" - the constraint is executed, and the correction vector U is weighted by weighting the temporary position vector "g" of the two main axes of the distribution Obtained by factor T. In Figure 6, the correction vector "f" is not shown, since it is consistent with the position vector, g". The distribution of the statistical matrix data set in Figure 7 is essentially - dimension-time (4) S11G (4) m彳 (4) conditions are based on the old offset c. And the statistical maternal data set to derive - the new offset is imposed. This constraint condition is the new offset e as the old offset c. Obtained from the sum of a correction vector "f" by which one of the principals of the distribution will correspond to a larger principal value corresponding to the distribution (i.e., corresponding to a larger eigenvalue λι) The temporary position vector in the direction (or a main direction) "g" one coefficient 匕 weighted temporary position vector "g,, the -weighted pixel"" and will correspond to the distribution < minimum primary value ( That is, the distribution of the __ coefficient corresponding to the smaller eigenvalue λ2) is not obtained by weighting the position vector "g " a weighting factor "〇". 114821.doc •26- The key to the algorithm in this particular example (12^H^^, attacking the equation 2) is “the number, v, is below—the constraint parent data is at eight ·,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,, Between the main values of the distribution of the statistical maternal data set: in the case where the ratio is less than or equal to the predetermined critical value "t", one of the components of the temporary position vector "g" in the main direction:: main direction The weighting factor is again set to 1 and is used for one of the main axes of the distribution with a smaller distribution level, and the weighting factor of the component of the upward temporary position vector "g" is set to 0 '' Second Specific Embodiment * Overview In the first embodiment, when the distribution of the statistical parent data set is sufficiently two-dimensional and when the distribution is substantially one-dimensional, the distribution is discretely estimated and new offset" c" uses different techniques to derive. In a second embodiment, an explanation will be provided of a simple, highly accurate algorithm that eliminates the need to perform different procedures based on the estimation of the distribution as in the first embodiment and can also be used efficiently The maternal data set is statistically derived to derive - a more likely new offset. Figure 8 is a flow chart showing one of the new offset derivation procedures. In the same manner as the first embodiment, when an offset update request has been made, the CPU 40 executes the program of Fig. 8 by executing the offset derivation module 94. The procedure of the step S200 is the same as the procedure of the above-described step sl1 in the first embodiment. The procedure of step S2〇2 is the same as the procedure of S1〇2 in the specific embodiment of the above-described step 114821.doc •27· 1320854 in the first embodiment. The procedure of step S204 is the same as the procedure of step S104 described above. • Distribution estimation in the step, the statistical matrix data set-distribution indicator is derived. In fact, the distribution of the parent matric data set is derived as a continuous indicator by the following equation (2〇). Value to estimate.

A Λ \*2

J •_· (20) where ' "k is a predetermined normal number. The corresponding main "m of the relational and statistical parent data sets must satisfy the following conditions (2". 〇^m2<l...(21) The spatial concept of m2 will now be explained with reference to Figure i. the Lord

The value of 匕 determines the reliability of the phase direction between the main values. Here, the coefficients of the equal-capturing position in the axial direction to the components of 1 g are represented by g« and gp in descending order of the corresponding principal values, and the position vector f in the main axis direction of the distribution The coefficients of the equal components are decremented by the corresponding principal values. and. In the present embodiment, the relationship between the temporary position in the temporary position, the correction vector £ and the call in this embodiment is expressed by the following equations and (23). Λ

Sa (22) II482I.doc 2S- ... (23) 1320854 Λ t m2

Λ'LV m2 it m2

The weighting factor fa/ga associated with the component corresponding to the major axis direction of the maximum principal value may be set to be less than "1%. Further, if the following conditions are satisfied, any definition may be made for the determined "ΓΠ2" , such that the weighting factors continuously correspond to the ratio of the principal values and are not limited to the definition of equation (2〇). • The range of values of Π12 is [0, 1; | or a subset thereof. • When person 2/into 1 =1. • If λ2 = 0 is permitted, then =. • In general, m2 decreases monotonically with increasing. Specifically, m2 can be defined, for example, by the following equation (24).

1 1 mi ~~r--sgn- 2 5 1 1 2 s 2 2 , 7 1 (24) where "S" and "k" are non-negative real numbers and "Sgn" is expressed by the following equation One of the symbol functions. Sgn(;c) lfe〇) -〇(x = 〇) ··· (25) <χ<〇) Since the derivation defined by equation (24) requires a very large number of calculations, it has input 1 and The calculation result of equation 2 of human 2 can be stored in a 2d 杳 114821.doc -29· 1320854 lookup table and then an approximation of ΙΠ 2 can be derived by referring to the 2D lookup table. The indices "s" and "k" are set depending on the specific embodiment, since the weighting effect varies depending on its value. Figures 9 and 1 show that depending on the index %, and "k" The graph of the weighted effect of the change. Figure 9 shows that in the equation s is fixed to "1" and "k" is set to 1/4, 1 and 4 in the eigenvalue ratio (λ^λΟ and weighting factor called The relationship between Figure 1 is shown in the equation where "k" is fixed to "1/4" and "s" is set to 1/2, at the eigenvalue ratio Lu (, /λι) and weighting factor Relationship between m; 2. When a magnetic sensor is attached to a moving subject or object (such as a mobile phone or a PDA) that changes at a higher speed or rate in a posture or posture, It can be expected that the distribution of a magnetic data set stored over a certain period of time will be relatively wide on average. In the case where the magnetic data distribution is not too wide, if the offset correction is made by significantly estimating the principal axis direction having a smaller main value When the magnetic data group is implemented, the offset accuracy will be relative Therefore, since a magnetic data group has low reliability, in the case where the present invention is applied to a moving object, the distribution of the magnetic data group tends to become relatively wide, and the parameters s and The value of k will be set so that the magnetic data group with the smaller principal direction of the major axis is clearly evaluated only if the distribution of the magnetic data is relatively wide. Here, with the 方程 in equation (24) The smaller the value becomes, the greater the weight in the direction of the major axis with the distribution of the smaller principal values becomes. Therefore, the equation (24) is applied to the case where the moving object of the attitude is changed at a higher angular rate. Next, you need to use a relatively large value to set the parameter s and use a relatively small value to set the parameter k. 114821.doc 1320854 On the other hand 'when a magnetic sensor moves or moves at a lower angular velocity When the subject or object (such as a car) is on, it is assumed that the distribution of the magnetic data set stored in the period between the special materials will be narrower on average. In this case, when the distribution of the magnetic data is not too wide, use - In the case of weight correction to evaluate the magnetic data group with a smaller principal value in the direction of the main axis, the offset accuracy will be improved, although such magnetic data groups are less reliable. In the case where equation (24) is applied to a moving object whose attitude is changed at a slower angular rate, a relatively small value is required to set the parameter s and a relatively large value is used to set the parameter k. • Derivation of new Offset When it is difficult to derive one of the optimization problems under a particular constraint, a relaxation problem for solving the optimization problem by relaxation constraints can be introduced. By applying this relaxation problem, this The specific embodiment implements a program for deriving a new offset Ct as a sum of the old offset C() and a correction vector f, the correction vector f being obtained by the coefficient of the temporary address vector g See Figure 1) The weighting factors that are weighted consecutively corresponding to the ratio of the principal values of the distribution of the statistical maternal data sets are obtained. The following are the details of this procedure. An unknown constant multiplier P is defined as the calculation of one of the required variables during the program and the C and p systems are grouped together into a vector "χ, which is defined by the following equation (26).

Jc = C LPJ ... (10) 114821.doc 丄) 鄕 54 In addition, the 'one matrix' B' is defined by equation (27) and the vector "b" is defined by the square (28). ...(27)

XT] w2i//c〇 ...(28)

The program for deriving a new offset is used to find one of the following simultaneous equations (29) in step S208. The vector x is uniquely specified, since the matrix must be non-singular.

Bx=b ... (29)

A solution to find a simultaneous equation (29) is equivalent to solving a minimum offset equation (12) under a constraint obtained by summing a new offset system as the sum of the old offset Co and a correction vector f. An optimization problem of the objective function, the component of the correction vector f being weighted continuously by the coefficient of the position vector g in the direction of the major axis of the distribution corresponding to the principal values corresponding to the distribution of the statistical matrix data set The value obtained by the factor fa/ga & fp/gp. In the second embodiment, it is easier to develop or improve the offset derivation module 94 and also reduce the data size of the offset derivation module 94. Since the statistical matrix is distributed according to the above, the branch (four) is offset. Material program. In addition, the second embodiment increases the efficiency of the use of the statistical matrix data set by the offset derivation module % and also allows the azimuth derivation module to correct the magnetic data using the most probable offset, since the old offset can be in the distribution The main axis direction 114821.doc • 32- The upper correction corresponds to the statistical matrix that none of the main values is zero. C. Other Embodiments The distance of the main value ratio of the data set. The first aspect of the present invention is not limited to the above specific embodiments and various and practical embodiments can be made without departing from the spirit of the present invention. For example, the present invention: the first aspect can also be applied to a magnetic sensor on a _PDA, a mobile phone, a two-wheeled motor boat or the like.

Next, a principle and algorithm for the second aspect of the present invention will be described in detail below with reference to FIG. The key point of this algorithm is that a magnetic data set distributed in the direction of a major axis with a larger spread is estimated to be a more pronounced statistical parent element for updating the offset and distributed in the direction of the major axis with a smaller spread. A magnetic data set is estimated to be a less obvious statistical parent element for updating the offset. The details are as follows. Corresponding to the end point "g, the offsets of the old offset CQ, the new offset C, and the temporary offset relative to the position of the origin "〇" are linearly combined as one of the basic data sets of the magnetic data. 3D position

The quantities °, i.e., the offsets of the offsets are the vector data represented in the xyz coordinate system. The new offset c is derived based on the old offset CG and a magnetic data set. The magnetic data set is stored to update the old offset using the new offset C. The magnetic data set is stored for use with the new bias. Moving the C to update the old offset statistical data set may include a magnetic data set that has been stored for a predetermined period of time and may include a magnetic data set including a predetermined number of magnetic data and may also include one included at a time (eg, A magnetic data set of any number of magnetic data that has been stored when an offset update request is made. 114821.doc -33· 丄 Ring 54 The old offset C〇 can be derived using the same method as the new offset C and can also be added. Although the temporary offset is defined to be derived from the statistical parent data set without using the old offset C〇', this definition is introduced to define the constraints used to derive the new offset c and the temporary offset is actually It is not necessary to derive the information. If the temporary offset is actually derived from the statistical parent data set without using the old offset C, then the temporary offset is a vicinity of the position vector of the center of the spherical surface of the statistical parent data φ set. However, if the statistical maternal data set is unevenly distributed over a portion of the spherical surface derived from the statistical maternal data set, an error included in each element of the statistical maternal data set will greatly affect the The derivation of the spherical surface, so it is possible to derive a temporary offset away from the true offset. For example, consider that a statistic parent data set is distributed in a circular distribution with mutually orthogonal eigenvectors U1, "and... and the statistic parent data set is in the direction of the eigenvector U3 corresponding to the smallest principal value. Minimized, as shown in Figure i. In this case, since the variable of the statistical parent data set is smaller in the direction of the feature vector U3 of the distribution, there is a higher probability from the statistical data set. The position of one of the temporary offsets is derived in the direction of the feature vector 4, since the direction of the system W is farther away from the true offset. On the other hand, in this case, the variable of the parent data set is The eigenvectors in the distribution are large, so there is a higher probability of deriving one from the statistic parent data set.

114821.doc uses the main value of the distribution. Therefore, the device estimates the statistical parent element that is distributed in the direction of the main value according to the ratio of the main values λι, λ2, and λ3 of the -34-1320854. . In the case of a clear product, first, as a new offset C, the position vector of one of the offsets C is corrected vector f and a position of the temporary offset relative to the old offset CG It is defined in a standard system having coordinate axes β and γ that coincide with the direction of the major axis of the distribution. That is, what is the correction vector f and the position vector? § Each vector can be defined as one of the basic vectors of the main axis direction of the distribution. This point corresponds to the conversion to the spindle value. If the correction vector f is in the range of f „ Φ, fa, 6 and 6 are based on the corresponding main values U, ..., and u of the distribution, and the target θ l is also measured by the summer weighting of the temporary The offset is derived relative to the position vector g of the old offset C〇, which may be derived by increasing the reliability of the statistical parent element in a direction with a large spread, ... The correction vector f is deduced by reducing the reliability of the statistical matrix element in a direction with a smaller dispersion of the scatterers. However, the eigenvalues of the instrument forward If and the position vector g are also introduced to define Deriving the new value 狡 雅 导 祈 偏移 偏移 偏移 偏移 偏移 偏移 偏移 偏移 偏移 偏移 偏移 偏移 偏移 偏移 偏移 偏移 偏移 偏移 偏移 偏移 偏移 偏移 偏移 偏移 偏移 偏移 偏移 偏移 偏移 偏移 偏移 偏移 偏移 偏移 偏移 偏移 偏移 偏移 偏移 偏移 偏移 偏移 偏移 偏移 偏移The old offset c, derived from the sum of the correction vectors f as defined above, derives the new offset c under the constraints, and can derive the new offset i while estimating the distribution of the magnetic shell in the direction of the major axis with a larger spread. The material set is used to update the more obvious elements of the offset statistical parent and is distributed in a smaller A magnetic data set in the direction of the main axis of the dispersion is used for the less obvious elements of the statistical matrix of the 2 offsets. An exemplary technique for deriving a new offset in this way is to formulate the distribution as one of the most If the new offset is used as the best one of the distribution, the new offset can be derived by solving the simple simultaneous linear equation. As described later in the specific embodiment, that is, regardless of the knives of the statistical matric data set, the apparatus can be used to derive the most likely new offset through a simple procedure, which can be derived from the statistical maternal data set. In the magnetic data processing apparatus of the present invention, the constraint may be such that a ratio of an intermediate value of one of the primary values to a maximum primary value of one of the primary values is higher than a first critical value and the minimum primary value is Where the ratio of one of the maximum principal values is equal to or less than a second threshold, the second basis vectors for one of the principal directions corresponding to the smallest principal value of one of the principal values Position vector of one coefficient One of the weighting factors is zero and the ratio of the intermediate primary value to the maximum primary value is equal to or less than the first critical value and the ratio of the minimum primary value to the maximum primary value is equal to or less than the second critical value. In the case, the coefficient of the second basis vector for the direction of the major axis corresponding to the minimum principal value and the coefficient of the second base vector for the other principal direction of the principal direction corresponding to the intermediate principal value The individual weighting factors of the position vector of the person are zero. The device discretely estimates the distribution of the statistical parent data set and discretely weights the position of the temporary offset relative to the old offset c relative to the old offset c according to the discrete estimation result. In the case where the value of the distribution is the weight of the coefficient of the position vector g in the direction of the minimum value, in the case where the value of the distribution is small in the direction, the direction of the value of the distribution is in the direction of the intermediate value. The weight of the coefficient of the position vector g is "〇". That is, there is no magnetic data set estimate in the direction where the value of the distribution is less than a critical value. In the magnetic data processing apparatus of the present invention, the constraint condition may be that the coefficient of the positive vector is 114821.doc -36 - 1^20854 by weighting the coefficients of the position vector continuously corresponding to the distribution of the wash-ten matrix data set. The value obtained by the weighting factor of the ratio of the principal values. This device can increase the substantial efficiency of the statistical maternal data set, since these weighting factors have a continuous correlation with the distribution. In addition, the device simplifies the offset update procedure by deriving a new offset based on the statistical parent dataset distribution without changing the program. φ In the magnetic sensor shift derivation device of the present invention, the individual weighting factors for the coefficients of the position vector may be used by coefficients for the second basis vector in the direction of the major axis corresponding to the maximum principal value. The weighting factor is set to one to normalize. In the case where the estimate of the distribution of the statistical matrix magnetic data set other than the primary values of the distribution is introduced, it is not necessary to normalize the weighting factors by setting the maximum weighting factor to be -. For example, the maximum weighting factor may be based on a statistical matrix magnetic data set, based on a maximum distance between magnetic data corresponding to a major axis direction (ie, the main direction) of one of the largest principal values, and a derived spherical surface (having a The ratio of the radius of the sphere of the portion of the statistical parent magnetic data set distributed nearby is set to be less than one. " In the magnetic data processing apparatus of the present invention, the offset deriving member can derive C, which minimizes the following objective function f(c) 9 / in the constraint condition, in magnetic: the shell system is qi = (qix , qiz) (i = 〇, L 2,) when expressed, "X" and "j" are as follows: No 114821.doc -37- U20854

Since the device uses a ° to derive the new enthalpy, the distribution-optimization problem is used to derive a new offset, as described in the Simple Simultaneous Linear Square Statistic Parent Data tot, as described later. That is, whether or not to derive the most likely new: shift, the device can be derived from the statistical parent data set by a simple program offset. A magnetic measuring 兮偌 & magnetic sensor of the invention. D ^ The distribution of the above magnetic data processing device and the data matrix of the deduction can make the device transparent. #引可能的可能偏'' from the statistical maternal data set to the function of each component of the plurality of components in the device of the present invention: the hardware resource from the function by its structure, its function is one Program resources or a combination of these resources is implemented. The function of the plurality of # is not determined by the entity-independent hardware resource. The invention can be specified not only by the device but also by the program, the recording medium on which the program is recorded, and a method. Assuming that there are no technical obstacles, the operations of the methods described in the patent application are not necessarily performed in the order described in the scope of the patent application, but may be performed in the order of or in addition to the order of 114821.doc - 38 - 1320854. A number of specific embodiments of the second aspect of the invention will be described in the following order. A. First embodiment [1. Overview] 1-1. Hardware structure 1- 2. Software structure [2. Program] 2- 1. Overall flow 2-2. Buffer update 2-3. Distribution estimation 2 -4. Deriving new offsets through optimization problems 2-5. Constraints when the distribution is two-dimensional 2-6. Constraints when the distribution is essentially one-dimensional 2-7. Derivation when the distribution is three-dimensional New offset 2-8. Summary B. Second embodiment • Overview • Distribution estimation • Derivation of new offset C. Other specific examples [1. Overview] 1-1. Hardware structure Figure 12 is used as an application An action of one of the moving subjects of the invention 114821.doc -39 - 1320854 The mobile phone 3 includes a field (four) magnetic sensor 4. The 3D magnetic sensor 4 detects the magnetic field in three orthogonal directions (X The individual strengths on , y, z) are used to detect the direction and intensity of a magnetic field. Mobile Phone 3 - Display 2 displays various character or video information. For example, the display 2 displays a map and an arrow or a character indicating an orientation (or azimuth).

Figure U is a block diagram of a magnetic measuring device comprising a 3 〇 magnetic sensor 4 and a magnetic data processing skirt 3D magnetic sensor 4 comprising X, 乂 and z-axis sensors 30, 32 and 34. It detects X, 丫 and 2-direction components of one magnetic % to 1 due to terrestrial magnetic force. Each of the χ, 丫 and 2-axis sensors, Μ and 34 sensors includes a magnetoresistive element, a Hall sensor or the like, which can be any type of one-dimensional sensor (assumed) It has directions x, y & Z-axis sensors 30, 32 and 34 are fixed such that their sensing directions are perpendicular to each other. The outputs of X, mandrel sensors 30, 32 and 34 are time-divided and input to - magnetic Sexy sensor interface (I/F) 22. After amplifying the input from the isometric sensors 3〇, 32, and 34, the magnetic sensor interface 22 converts the analog input to analog input. The digital magnetic data from the interface 22 is input to the magnetic data processing device 1 through a bus bar 5. The raw data processing device is a computer including a cpu, a RAM 34 CPU 40 control (for example The overall operation ROM 42 of the mobile phone 3 is a non-volatile storage medium, which stores __magnetic resources = processing program or various utility programs (for example, a navigation program) for implementing the mobile body executed by (4) 4(). 44 is a volatile storage medium that temporarily stores data for processing by the CPU 40. Processing device 114821.doc 1320854 The 1 and 3D magnetic sensor 4 can be constructed as a single-wafer magnetic measuring device. 1-2. Software structure Figure 14 is a block diagram of a magnetic data processing program 9 magnetic data processing program 90 System, store μ〇μ 42 to provide orientation data to the navigation program (10). 11 Hai position data is a 3D vector data indicating the orientation of the earth's magnetic field. As a 3D vector data for gesture detection of, for example, a moving subject. The azimuth-bean material can be provided in other modes. The magnetic data processing program 9 is constructed as a module group, for example, a buffer management module 92, an offset derivation module 94, and an orientation derivation module %. The buffer official module 92 is a program portion that receives a plurality of magnetic data continuously outputted from the magnetic sensor 4 and stores the received magnetic data in a buffer (4) for use in the offset update. The buffer g-module 92 allows the CPU 40, the RAM 44, and the ROM 42 to be used as an input member and the storage member 4 buffer can be embodied by only (4) hardware, and can be embodied by software. A magnetic data set in the buffer is referred to as a statistical parent data set. The offset derivation module 94 is a program portion based on a statistical parent data set and an offset derivation module 94 held by the buffer management module 92. The old offset is preserved to derive a new offset and the old offset is used to update the old offset. The shifting derivation module 94 allows the CPU 4, ram 44 & R 〇 M 42 to be used as an offset deriving component. A new offset to update the old offset causes the new offset to become the old offset, so the old offset " bow is offset again in the context of no misunderstanding. In fact, an offset system for azimuth data correction is determined by a variable and the new offset is derived as a variable different from the variable 114821.doc 1320854. When the new offset is derived, it is set using the variable used for orientation data correction. Therefore, the variable used for orientation data correction is a variable in which the old offset is stored. The azimuth derivation module 96 is a program portion that uses the offset maintained by the offset derivation module 94 to correct the magnetic output continuously output from the magnetic sensor.

Material to produce orientation data. The azimuth derivation module 96 allows the CPU 40, RAM 44, and ROM 42 to function as azimuth deriving members. Clearly speaking, the azimuth derivation mode

Group 96 uses as orientation data to output all or both of the three components obtained by subtracting the offset component from the component of the magnetic data as vector data. The navigation program 98 is a known route that searches for a route to a destination and displays the line on a map. Since it is easier to identify the map, the (4)® display shows 'the map orientation matches the real world orientation. Therefore, for example, when the field mobile phone 3 is rotated, the map displayed on the display 2 is rotated relative to the display 2 so that the map does not rotate with respect to the earth. The

The orientation data is used for this location. Of course, the orientation data can only be used to display north, east, and west by characters or arrows. [2. Program] 2-1. Overall Flow Figure 1 5 shows a flow chart of a new offset derivation program. When an offset update request has been made, the CPU 40 is executed by the execution of the offset derivation module 94 to execute the procedure of FIG. 2-2. Buffer Update In step S100, the magnetic data stored in the buffer is deleted, and the magnetic data set is stored in 114821.doc • 42· ==, 35 (4) - used to derive the new offset (statistics) Shell material set). From A, in this procedure, one is used to derive the old offset, and the parent data set is deleted. The memory S1G2' is used to derive the new offset magnetic data input and the buffer is within the buffer. When a plurality of magnetic data systems continuously change from the magnetic sensing to the attitude of the mobile phone 3, the distance between the two consecutive input magnetic data (or values) is less than the buffer of the limited capacity. Storing multiple new magnetic data in the device wastes memory resources and causes unwanted buffer updates. In addition, if a new offset is derived based on a set of new (four) sets, then there is a possibility that an inexact new offset is derived from an unevenly distributed statistical matrix data set. Whether it is necessary to be more = buffering benefits can be determined in the following ways. For example, if the distance between the magnetic ejection materials stored in the buffer before the last input magnetic material is less than a given threshold value, it is decided that it is not necessary to update the buffering benefit and discard the Finally, the magnetic data is input and not stored in the buffer. In step S104, it is determined whether a specified number of magnetic data required to derive a precise new offset has been stored in the buffer. That is, the number of elements of the statistical parent data set is predetermined. Set a smaller number of statistical maternal data sets. The vegetarian response improves the response to the offset update request. The procedures of steps 81〇2 and §1〇4 are until the characteristic number of magnetic data is stored in the buffer. 2-3. Distribution Estimation Once a specific number of magnetic data is stored in the buffer, the distribution of the statistical parent data set is estimated (indicated by S1Q8). This distribution is estimated based on the principal value of the H4821.doc -43 - 1320854 distribution. When the magnetic data set is expressed by the following equation (31), the main values of the distribution are the eigenvectors of a symmetric matrix A, and the symmetric matrix A is used from one of the statistical matrix data sets (average value). The sum of vectors starting and ending with individual magnetic data is defined by equations (32), (33), and (34). Qi一(qix,qiy,qiZ)(i = 0,1,2,.. .)...(31) A = XT X ...(32)

among them

... (33) ...(34)

Matrix A can also be rewritten as equation (3 5). /«〇 Let λι, λ2, and λ3 become the eigenvalues of the matrix eight in ascending order. Let h and A be mutually orthogonal feature vectors corresponding to λι, ^, and & and have been normalized to size 1. The range of λι, and person 3 handled by this specification is λι>〇, 2>0, and λ3 Μ. When two or more eigenvalues of the matrix 零 are zero, that is, when the rank of 114821.doc •44 · 丄JZUOJH·matrix A is one or less, ... the two are to be tested I matrix A, due to the statistical maternal data set The number of elements is one or two. The eigenvectors of these features must be zero bite - A ^ M and positive real numbers, since matrix A is a half positive definite matrix according to the definition of matrix A. The distribution of the statistical matrix is based on the ratio of the minimum (four) value to the maximum eigenvalue (10) and the ratio of the intermediate eigenvalue to the maximum eigenvalue. In step S1, it is determined whether the distribution of the statistical maternal body is sufficiently two-dimensional. Specifically, the decision line is affirmative when the following condition J (36) is satisfied and negative when it is not satisfied. Μ/λι>. And ... (36) Here, V, and, V are predetermined predetermined values. How to set the value of value tjt2 is a design option and it can also be set as needed/needed based on how the derivation feature of the offset is determined. When condition (36) is satisfied, the statistical matric data set is isotropically distributed from the center of the statistical parent data, Jiao, and #. The statistical maternal data set is distributed around the center of the isotropic _ ^ J "knife cloth buckle not statistical maternal data sets are evenly distributed on a specific spherical surface. In the next step, it is decided whether the distribution of the statistical maternal data set is sufficient. The decision is affirmative when the following condition (37) is satisfied and negative when it is not satisfied. Enter 3/in 1 Sti and person 2/person 1> (2 · (37) When condition (37) is satisfied, the statistical matrix data set is within the range defined by the - specific plane from the statistical parent data The center of the set is isotropically distributed. The statistical maternal data set is defined by a certain plane near 114821.doc •45-1320854. The knives indicate that the circumference of the circle of the statistical parent data circle surrounds the circle The isotropic set of the heart is unevenly distributed in the vicinity of a specific sphere. If the decision is H, then the statistical matric data set is substantially-dimensional (ie linear). : The distribution indicates that the statistic parent data set is on a shorter arc of one of the circles of a particular spherical surface or in the circle of the segment

All are distributed. _ The end of the illusion is not 2-4. Deriving the new offset through the optimization problem A optimization problem for deriving a new offset will now be explained. When the statistical maternal data set includes four magnetic data that do not exist on the same plane, the spherical surface of the parent data set is uniquely specified without use—statistical techniques. The center of the spherical surface is ζ 1 standing = (cx, cy, cz) is obtained by solving the simultaneous equation (38). Although four equality constraints exist for three variables, equation (38) must have a solution, since one of the four equality constraints is equal constraint redundancy. \q〇~q)T~ 9〇Γ^〇 ~ R C^i 9iTq, - R iq2-j)T 2 <h ~ R "9)T_ - R_ where

H4821.doc -46- ... (39) ··· (38) Tian. When the number of elements in the mother's data set is 5 or more, "j " is defined by the following equation (40).

^Tgl -R 92Tq2 -R

βΝ-\Τ(ΐΜ-\ -R ... (40) Here, if the simultaneous linear equation (41) for "c" has a solution, the solution distributes the statistical parent data The center of the spherical surface is set. Xc=j (41) However, if one considers the inherent measurement error of the 3D magnetic sensor 4, the equation (41) cannot actually have the answer. The following equation (42) Defining one of the vectors "e" is to introduce a plausible solution through _statistical techniques. e = X°'j ··. (42) Minimize Ne ll22 (ie eTe)"c" Similar to {not the center of the spherical surface closest to the distribution of the statistical maternal data set. When the matrix is singular, one is used to find the value of the minimized ||e丨丨22" 〇" The problem of optimization of the objective function of one of the following equations (43) is best solved. Objective function: (43) 2-5. Constraints when the distribution is two-dimensional as shown in Fig. 16, when the statistical matrix The distribution of the data set is two-sided) 'A new offset is obtained by correcting the old offset, '(ie the flat offset direction is limited to two ll482l.d〇c •47-) 4 The direction of incoming material (s丨i 2). The masterbatch is located near a particular plane and the distribution is perpendicular to the direction of the L from parallel to the plane: the statistic is sufficiently reliable The statistics in the direction perpendicular to the plane = the distribution of the material set is unreliable. In this case, perpendicular to the plane: the upward 'old offset is not corrected, thus preventing updates based on unreliable information Offset. When the statistical matrix data set is distributed near a specific plane and the distribution is discrete in a direction perpendicular to one of the planes, the direction perpendicular to the plane is corresponding to the minimum eigenvalue ～ eigenvector... The directions in which the directions are coincident and parallel to the plane are consistent with the directions of the feature vectors u丨 and U2 respectively corresponding to the intermediate feature values λ2 of the maximum eigenvalue range. Therefore, in order to derive a direction perpendicular to the plane The new offset does not correct the old offset, minimizing one of the objective functions of equation (43), the new offset (: is obtained under the constraints expressed by the following equation (44). c=c〇+plUl + (32U2 (Pi, |32: real number (44) Equation (44) is equivalent to the following equation (45). (45) U3T(C~Cq) « 〇 is used to solve the optimization problem of equation (43) under the constraint of equation (45) The equation can be modified to its equivalent valence equation using the Lagrangian multiplier method. When an unknown constant multiplier p is introduced and "χ" is depreciated by the following equation (46) ' &quot The simultaneous linear equation (47) of X" is the above simultaneous equation. H4821.doc -48· ...(46) X -1320854 ... (47) β4文=厶4

B 4 b4 2A «3 W3r Ο 2Χ τ τ _Μ3 & (48) (49) It should be understood from the above description that if the distribution of the parent data set is two-dimensional, it is used to derive new in step S112. The offset plume program is used to solve the simultaneous linear equation (47). The answer "X" must be only the ground... Anal knowledge, because the rank of the matrix Β4 must be 4. f π 2-6. When the distribution system is substantially one-dimensional, the constraint condition is shown in Fig. 17. When the statistical matrix data is only π* fruit, the distribution is essentially one-dimensional (ie linear), and the new offset system Deriving (SllG)e by correcting the direction of the old offset to the main direction of the distribution (SllG)e # The statistical matrix data set is distributed near the -spec line and the distribution is discrete in the direction of the line

The distribution of the statistical matric data set in the direction of the line is sufficiently reliable, and the distribution of the statistical maternal data set of X in other directions is not reliable. In the case of 'in the direction other than the direction of the line, the old offset is not correct, thereby preventing the offset from being updated based on unreliable information. , 114821.doc • 49- When the statistic parent data 八 is left in the direction of the line, ^ is near the specific line and the distribution coefficient value...the eigenvector direction corresponds to the largest feature corresponding to the intermediate eigenvalue = Therefore, the other directions are derived from the eigenvectors of the small-scale traits 3, which are represented by the equation (50), and the new offset is derived only in the direction of the line. Minimize the objective function of equation (4)) - the new offset e is in - by the following ί50) + from the soil a .. ... (50) c = c 〇 + /3 , u 方私 (50) The equation is equivalent to the following equation (51). The equation for solving the optimization problem of equation (43) under the constraint of equation (51) can be modified to its equivalent simultaneous equation using the Lagrangian multiplier method. . When an unknown constant multiplier P1 is introduced by the system and "X" is defined by the following equation (52), the simultaneous linear equation (53) of "X is the above simultaneous equation.

A ...(52) ...(53) P2_ where 114821.doc -50- 2^4 u2 Ο

U2T

U

T 0 M300 • .· (54) 2X~ U2TC〇M3rC〇...(55) It should be understood from the above description that if the distribution of the statistical maternal data set is substantially one-dimensional, then the step suo is used to derive the new offset. The program is used to solve the simultaneous linear equation (53). The answer, 'χ" must be _, because the rank of the matrix I must be 5. 2-7. Deriving a new offset when the distribution is three-dimensional When the distribution is three-dimensional, a new offset is derived without limiting the direction of correcting the old offset (S114). When the distribution is three-dimensional, that is, when the statistical parent data set is distributed in various directions from a center of the statistical matrix data set, the statistical parent data set is filled in all directions. Guilty. Therefore, in this case, in order to derive a new offset, it is not necessary to use the old offset, so a new offset can be derived based on the statistical parent data set without using the old offset. An algorithm for deriving a new offset based on a statistical matric data set without using the old offset can be an algorithm that uses one of the various proposed statistical techniques and can also be an algorithm that does not use statistical techniques ( Patent Application Nos. 2005-337412 and 2006-44289, which are hereby incorporated by reference. In this particular embodiment, a new offset uses a statistical technique to push 114821.doc -51 - 1320854. That is, the 'new offset at step s114', the c" is used as the answer to the finalization of the objective function for minimizing the equation (4) without any constraint conditions. 2-8·Overview

The procedures of steps S1, SU2, and S114 will now be described with reference to the concepts between Figs. 1, 6, and 7 m. If the statistical matric data set is assumed to be completely reliable, then the new offset e is considered to be the old offset 4 and the old offset c. The sum of the position-g" vector g of a spherical surface center derived from the statistical parent data is defined by the following equation (56). c=c〇+g (56) as the eigenvectors ui, U2 of the distribution, which are derived from the target of the minimum equation (43), without any constraints. Linear combination of one of the basic vectors in the same direction as U3. Therefore, the individual degree of the component according to the position vector V corresponds to the correction vector corrected from the position vector _'g" - a correction vector of the vector can be based on the statistical matrix data set in the corresponding main axis direction. The individual reliability is obtained by weighting the position vector ng" coefficients ga, §(} and since (see Figure η). As shown in Figure 16, step SU2 is performed when the distribution of the statistical matrix data set is two-dimensional. In the program, the column constraint is imposed when the new offset is derived based on the old offset center of the statistical matrix data set. The constraint is the new offset C system as the old offset C() and a correction vector. Obtained by the sum of "f", the correction vector "f" is by the position vector corresponding to the maximum principal value of the distribution (i.e., the principal direction corresponding to one of the distributions of the larger eigenvalues) ••g" one of the coefficients g« and the position in the direction of the main axis of the distribution corresponding to the intermediate value of the distribution (ie corresponding to the eigenvalue λ2 in the middle 114821.doc • 52·), Coefficient gp both weighted 罝Θ g g" And "Γ, and will be obtained by the weighting factor "G" of the distribution, the weighted position vector ν corresponding to the sub-master value (i.e., corresponding to the minimum eigenvalue λ3). The distribution of the statistical parent data set is essentially: in the procedure of the step SUO performed, the following constraints are imposed when the _new offset is derived based on the old offset 4 and the statistical parent data set. The new offset is obtained as the sum of the old offset and the -correction vector, which is to be at the maximum principal value corresponding to the distribution (ie, corresponding to the larger eigenvalue λ1) The position vector "g" one of the coefficients in the main axis direction (or a main direction) is a weighted position vector "g"-weighted/factor 1" and will be in the middle value corresponding to the distribution (ie a coefficient gp in the direction of the major axis of the distribution corresponding to the intermediate feature value λ2) and a coefficient gy in the direction of the main axis of the distribution corresponding to the smallest principal value of the distribution (ie corresponding to the minimum principal value λ3) One of the weighted position vectors "g" The weighting factor "0" is obtained. In the procedure of the step suo performed when the distribution of the parent data set is three-dimensional, no specific constraint is imposed, that is, in step S110, the new offset C is used as the old offset c. Obtained from the sum of the position vector "g", which is obtained as a solution to the optimization problem of the objective function for minimizing equation (43) without using any constraints B. Second Embodiment * Overview In the first embodiment, the distribution of the statistical parent data set is discrete 114821.doc -53 - 1320854 and when the distribution is two-dimensional, the new offset "C" is derived by setting the component of the correction vector "f" in the direction of the major axis of the principal value to zero, and when the distribution is one-dimensional, the new offset "c" The component of the correction vector "f" in the direction of the two main axes of the main value system and the minimum value is set to zero to derive. In a second embodiment, a description will be provided of a simple, highly accurate algorithm that eliminates the need to perform different procedures in accordance with the "Establishment of the Measure" in the first embodiment. Efficient use of statistical matric data sets to derive - more likely new offsets. Figure (10), Description - Flow chart of one of the new offset derivation procedures. In the same manner as the first embodiment, when an offset update request has been made, the CPU 4 executes the program of Fig. 18 by executing the offset derivation module 94. The procedure of step S200 is the same as the procedure of step S100 described above in the first embodiment. The procedure of step S202 is the same as the procedure of step S102 described above in the first embodiment. The steps §2〇4 and κ10 are the same as the procedure of the first step S1G4 described above. In the embodiment, the distribution estimation is performed in step S206, that is, the distribution of the parent data, that is, the distribution of the parent data set is derived. (5 8) Defined as the distribution derivation as the distribution derivation from the following equations (7)) and (λ γ, and as a continuous value to estimate m, =1 ' Ϋ 3 (57) 114821.doc -54- (58) % =1 1320854 where ' "k, and "k are predetermined normal numbers. The values of h and h determine the correlation between the principal values and the reliability of the corresponding principal direction of the statistical parent data set. Here, "m2&quot ; and "m3" must meet the following conditions (59): 0 sm2<l and 0 sm3 si ...(59)

The values of the coefficients h and h are adequately determined in accordance with a particular embodiment of the invention 'because the weighting effect depends on the values of the coefficients 1^ and h. In the case where the magnetic sensor is attached to a portable object (for example, a portable telephone and PDA that changes its attitude at a relatively fast angular rate), it is desirable to accumulate a magnetic data group at a predetermined time interval. One of the groups has a distribution that is relatively broad on average. In the case where the distribution of the magnetic data is not too wide, if the offset correction is performed by significantly weighting the magnetic data group having the major axis direction of the main value, the offset accuracy will be considerably degraded due to such Data groups have lower reliability. Therefore, in the case where the present invention is applied to a moving object, the distribution of the magnetic material groups tends to become wider, and the values of the coefficients h and k; will be set so that the distribution only in the magnetic data is equivalent. More

The magnetic data group with the smaller principal direction of the main axis in the month is evaluated using heavier weights. On the other hand, in the case where the magnetic sensor is fixed to a moving object (for example, a car whose attitude is changed at a relatively slow angular rate), it is desirable to accumulate a magnetic data group at a predetermined time interval. A distribution becomes relatively narrow on average. When the distribution of magnetic data is not too wide, magnetic data with a smaller main value is evaluated without using a heavier weight: Group: In the case of performing offset correction, the offset accuracy will not be changed. - It is not reliable to use such magnetic data groups. Therefore, in the case where the present invention 114821.doc - 55 · 1320854 is applied to a moving object, the distribution of the magnetic data groups tends to become narrower, and the values of the coefficients h and h should be set so that even When the distribution of magnetic data is relatively narrow, the magnetic data group with the main axis direction of the smaller main value is evaluated using heavier weights. The concept of in2 and called space will now be explained with reference to FIG. The coefficient of the equal component of the position vector g in the direction of the major axis of the distribution is represented by ga, § 卩 and § 7 in descending order of the corresponding principal values and the position vector f in the direction of the major axis of the distribution The coefficients of the components are represented by fa, fp, and 6 in descending order of the corresponding principal values. The relationship between the position vector g, the correction to If, m2, and the call is determined by the following equation (6〇). And (62) to express. £〇_ • (60) 8a [±_ S0 — ί,2] UJ 2 J m2 - Λ A Λ • (61) Λ 8y λ]AA . (62) Decide to make these weighting factors continuously correspond to the main value ratio Such pass-off equations are not limited to Fangmo), (4) and (62). In addition, the weighting factor fa/ga associated with the main axis of the spindle corresponding to the value of = 114821.doc -56 - 1320854 can be set lower than • Deriving a new offset when it is difficult to make a specific hit, & Under the condition that one of the optimization problems is deduced, the problem of relaxation of the optimization problem can be solved by loosening the constraint condition. By the m &~ relaxation problem, this embodiment is implemented = As the old offset C. and the sum of the correction vector f to derive a new offset C::, the correction vector m is successively corresponding to the statistical parent data by weighting the above-mentioned position vector g (see the coefficients ga, gPAgy of the graphs) The weighting factor of the ratio of the main values of the set is obtained. The following are the details of this procedure. Unknown blame ^multiplier Ρ, and Ρ2#, defined during the program period ^--aggregation is integrated by the following equations 63) 2Λ m2u2 1 W3"3 w2m2t 0 0 ~-(m% 2λχ 3 Β 114821.doc ··. (64) -57- ... (65) ... (65) b 2XTj m2U2 C〇琢The procedure for deriving a new low 狡 - _ ^ new offset in step S208 is used to find one of the following simultaneous equations (36) to answer. 〆 Π is the only designation, since matrix B must be non-singular.

Bx=b (66) Finding the simultaneous equation (66) is equivalent to answering a constraint obtained by the new offset system as the old offset c〇 and the correction vector. The optimization of the objective function of the poem minimizing the filament (4)), the correction to the component of if by weighting the position in the direction of the major axis of the distribution corresponding to the principal values to the coefficient of 4g continuously correspondingly distributed The factors of the main value ratios f/g, f/...: tribute: the set ^ Ia / ga fp / gp and the heart / g 丫, and the value obtained. • In the second embodiment, it is easier to develop or improve the offset derivation module 94 and also reduce the data size of the offset derivation module. Since the distribution of the statistical matrix data set is not required, the branching is not required. New offset derivation program. In addition, the second embodiment increases the efficiency of the use of the statistical matrix data set by the offset derivation module 94 and also allows the azimuth derivation module to correct the magnetic data using the most probable offset, since the old offset can be used in the The distance in the direction of the major axis of the distribution contiguously corresponds to the ratio of the principal values of the statistical matric data set, unless either of the principal values is zero. C. Other Embodiments 114821.doc • 58-34 34 The second aspect of the present invention is not limited to the specific embodiments described above, and various specific embodiments can be used without departing from the spirit of the invention. For example, the present invention is applied to a magnetic sensor in which the ID is set to PDA ± or a magnetic sensor that is fixed to the straight line. BRIEF DESCRIPTION OF THE DRAWINGS Fig. 1 is a schematic view showing a specific embodiment of the first aspect of the present invention. Figure 2 is a schematic illustration of one embodiment of the first aspect of the invention. Figure 3 is a block diagram of a particular embodiment of the first aspect of the invention. Figure 4 is a block diagram of a particular embodiment of the first aspect of the invention. Figure 5 is a first embodiment of the first aspect of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS FIG. 6 is a first embodiment of the first aspect of the present invention, and is not intended to be the first embodiment of the first aspect of the present invention. FIG. 9 is a graph showing a correlation between the first embodiment and the first embodiment of the present invention. Fig. 10 is a graph associated with the first party of the present invention, namely, the heart, and the embodiment. Figure 11 is a schematic view of a specific embodiment of the second aspect of the present invention. Figure 12 is a schematic view of a specific embodiment of the second aspect of the present invention. Figure 13 is a specific embodiment of the second aspect of the present invention. - Blocks ^ 114821.doc - 59 - 1320854 Figure 14 is a block diagram of a particular embodiment of the second aspect of the invention. Figure 15 is a flow chart showing a first embodiment of the second aspect of the present invention. Figure 16 is a schematic illustration of a first embodiment of a second aspect of the invention. Figure 17 is a schematic illustration of a first embodiment of a second aspect of the invention. Figure 18 is a flow chart showing a second embodiment of the second aspect of the present invention. [Main component symbol description] 1 Magnetic data processing device 2 Car/display 3 Mobile phone 4 2D (2D) magnetic sensor/3D (3D) magnetic sensor 5 Busbar 22 Magnetic sensor interface (I/ F) 30 X-axis sensor 32 y-axis sensor 34 z-axis sensor 40 CPU 42 ROM 44 RAM 90 magnetic data processing program 92 buffer management module 114821.doc -60- 1320854 94 offset derivation module 96 azimuth derivation module 98 locator / navigation program

114821.doc -61 -

Claims (1)

1320854 Patent Application No. 096,107,901 (5) Chinese Patent Application Substitute Replacement (98 丨〇月^10, Patent Application Range: i) A magnetic data processing apparatus comprising: an input member' for continuously inputting one or two The magnetic data output by the dimensional (2D) magnetic sensing thief, the magnetic data is a 2d vector data linearly combined as one of a set of first basic * vectors; a storage component for storing a plurality of input magnetic data as a statistical matric data set to update an old offset of the magnetic data using a new offset; and a lure deriving component for use as the sum of the (four) offset and the bifurcation positive vector in the new offset system Obtaining a new offset based on the old offset and the statistical parent data set, wherein the correction vector is a set of second basic vectors defined in a major axis direction of the distribution of the statistical parent data set One of the linear combinations, and the individual coefficients representing the linear combination of the second basic vectors of the correction vector are based on the distribution of the statistical parent data set The principal value ratio weighting a temporary offset is obtained relative to an individual coefficient of one of the old offset temporary position vectors derived from the statistical parent data set without using the old offset, the temporary position vector system One of the second basic vectors is linearly combined. 2. The magnetic data processing apparatus of claim 1, wherein the constraint is equal to a ratio of a smaller primary value to a larger primary value of the primary values Or less than a predetermined threshold value, one of the temporary position vectors for the coefficient of the second basic vector in the direction of the major axis corresponding to the smaller principal value of the primary values is zero. 114821 - 981002.doc 3. The magnetic data processing apparatus of claimant, wherein the constraint is that the coefficient of the correction vector is weighted by the coefficient of the temporary position vector continuously corresponding to the ratio of the main values of the statistical data set distribution The value obtained by the weighting factor. 4. The magnetic data processing apparatus according to any one of claims 1 to 3, wherein the individual weighting factors of the coefficients for the temporary position vector are The normalization is performed with reference to a weighting factor for the coefficient of the second basic vector in the direction of the major axis corresponding to the larger principal value and is set to one. 5. If any of the claims 丨 to 3 A magnetic data processing apparatus, wherein the offset deriving member derives a value "c", which minimizes the following objective function f(c) under the constraint: ί卜f Γ 9〇y〇-R 1 -q) 7 =2 q2 -R • · - qj • ·» .Uah ~ and among them A magnetic data processing apparatus according to claim 4, wherein the offset deriving means derives a value "c" which minimizes the following objective function f(c) under the constraint: Fip) = {Xc^j)T[xc^j) f where the magnetic data is represented by qi = (qix,qiy)(i = 〇,1,2,. · ·)' "X" and, , j" is as follows: / -yj, - Qo -R X.= Η .1 T ^ m-.v 1 2 - 2 ^9n-\ • · » βΝ-\ 9 ff-\ ~ where 1 N~\ ι·=0 〇7· A magnetic measuring device, comprising: the magnetic data processing device according to any one of claims 1 to 6; and the 2D magnetic sensor. 8. A magnetic data processing method, comprising: Continuously inputting magnetic data rotated from a two-dimensional (2D) magnetic sensor, the magnetic data being a linearly combined 2D vector data as one of a set of first basic vectors; storing a plurality of input magnetic data a statistical parent data set to update the magnetic data for one of the bird's offsets using a new offset; and 114821-981002.doc The new offset is used as the sum of the old offset and a correction vector, and under the force beam condition, the new offset is derived based on the old offset and the statistical matrix data set, where the correction vector is the statistical The linear coefficient of one of the second basic vectors defined in the major axis direction of the distribution of the maternal data set and the individual coefficients representing the linear combination of the first basic vectors of the positive sense are based on the statistical matrix One of the distributions of the data sets is weighted by a temporary offset relative to the individual coefficients of the old offset-temporary position vector to obtain 'this temporary offset is derived from the statistical parent data set without using the old bias Shifting the temporary position vector is a linear combination of one of the second basic vectors. A magnetic data processing program product that allows a computer to be used as: an input member for continuously inputting magnetic material output from a series, The magnetic data is used as a 2D vector data linearly combined with one of the first = inward; a storage member 'used to store a plurality of input magnetic data as one Calculating the parent data set to update the old offset of the magnetic data using a new offset; and an offset deriving component for using the new offset as the sum of the old offset and the forward direction 1 Obtaining a new offset based on the old offset and the statistical parent data set, wherein the sigma correction vector is the first statistical data set of the set of the correction vector defined upward One of the basic vectors of the principal axis of the distribution is linearly combined, and the individual coefficients H4821-981002.doc representing the linear combination of the two 'basic vectors are weighted by the ratio of the main values according to the distribution of the statistical data set. Transmitting individual coefficients of the temporary position vector relative to one of the old offsets to obtain 'the temporary offset is derived from the statistical parent data set without using the old offset'. The temporary position vector is the second basic vector One linear combination. 10. A magnetic data processing apparatus, comprising: an input member for continuously inputting magnetic material outputted from a three-dimensional (3D) magnetic sensor, the magnetic data being a group of first basic directions 2: a linearly combined 3 d vector data; a storage component for storing a plurality of input magnetic data as a statistical parent data set to update an old offset of the magnetic data using a new offset; and an offset derivation a means for deriving the new offset based on the old offset and the statistical parent data set under the constraint obtained by the new offset system as the sum of the old offset and a correction vector, wherein the new offset The correction vector is a linear combination of the main sense of the distribution in the main axis direction of the distribution of the statistical matrix data set, and the individual coefficients of the linear combination of the second basic vectors representing the positive vector of the weight are obtained by Obtaining according to the distribution of the statistical matrix data set - the main value ratio weighting - the temporary offset is relative to the individual coefficient of the old offset - the position vector, the temporary offset is from Statistics parent data set the offset is derived without using the old train position vector of the temporary offset to the linear combination of such second one of the basic amount. The constraint is in the first 11. The ratio of the intermediate value of one of the main values of the magnetic data processing device M4821-98I002.doc of claim 1 to one of the main values of one of the main values is one of the first a threshold value and the ratio of the minimum principal value to the one of the maximum principal values being equal to or less than a second threshold value is used for the direction of the principal axes corresponding to the smallest principal value of one of the principal values One of the position vectors of one of the second basic vectors is a weighting factor of zero and the ratio of the intermediate primary value to the maximum primary value is equal to or less than the first critical value and the minimum primary value is The coefficient of the second basic vector in the direction of the major axis corresponding to the minimum principal value and the corresponding to the intermediate principal value in the case where the ratio of the maximum principal value is equal to or smaller than the second critical value The individual weighting factor of the position vector of the coefficient of the other second basic vector on one of the principal axes is zero. 12. The magnetic data processing apparatus of claim 10, wherein the constraint is that the coefficients of the correction vector are weighted continuously by the coefficients of the position vector to correspond to the distribution of the statistical data set. The value obtained by the weighting factor of the ratio of the main values. The magnetic data processing apparatus of any one of clauses 10 to 12, wherein the individual weighting factors for the number of the position vectors are normalized with reference to the weighting factor 'the weighting factor is used for correspondence The coefficient of the second basic vector in the direction of the major axis of the larger principal value is set to one. The offset derivation member derives a value "c column objective function f(c) as claimed in any one of claims 10 to 12: a magnetic data processing apparatus, wherein the 1' is minimized under the constraint fic) = {Xc - j)T 114821-981002.doc -6 * 1320854 wherein the magnetic data is determined by qi = (qiX,qiy, When qiZ)(i = 〇,l,2,···) is expressed, "X" and "j" are as follows: "ΚΓ 9〇Τ9〇1 -R x = k-dosing / =— J 1 q ;qi-R U— βΝ-Χ^Ν-λ -R_ where R Τ NA 15. The magnetic data processing apparatus of claim 13, wherein the offset deriving means derives a value "c", which minimizes the following objective function f(c) under the constraint: f(c) = (Xc-j)T{Xc-j) f where the magnetic data is represented by qi = (qiX,qiy,qiZ)(i = 〇,l,2,...), "X" and "j" is as follows: "ΚΠ Vi 1 / = - 2 n βΝ-Ι^Ν-χ x= where 114821-981002.doc 16R N % ia〇• A magnetic measuring device comprising: according to request 1 to 15 The magnetic data processing of the middle-item is equipped with the 3D magnetic sensor. A magnetic data processing method includes: the magnetic material-linear combination is continuously input from a three-dimensional (3D) magnetic sensor, and the magnetic data is used as a set of first basic vectors of 3D vectors. Data; storing a plurality of input magnetic data as a statistical parent data set to update a new offset of the magnetic data using a new offset; and ^ as the old offset and a correction in the new offset system The new offset is derived based on the old offset and the statistical matrix data set obtained by the sum of the vectors, wherein the correction vector is one defined in the main axis direction of the distribution of the statistical parent data set One of the second basic vectors is linearly combined and the individual coefficients representing the linear combination of the second basic vectors of the correction vector are weighted by a primary value ratio according to a distribution of the s-statistical parent data set. Obtained from the individual coefficients of the position vector of the old offset, the temporary offset is derived from the statistical parent data set without using the old offset, the temporary offset The position vector is a linear combination of the second basic ones of 114821-98I002.doc -8 - 854 854. 18. A magnetic data processing program product, which allows a computer to be used as an input member for continuously inputting magnetic material outputted from a three-dimensional (3) magnetic sensor, the magnetic data being used as a a set of first basic vectors - a linear combination of 3D vector data; a storage component for storing a plurality of input magnetic data as - Count the parent data set 'to use - the new offset to update the old offset of one of the magnetic data; and the offset derivation component, which is attributed to (4) partial consumption (4) old offset and ΓΡ xin -3⁄4 -V -r- / 〇.. a constraint based on the sum of the old offset-correction vectors and the statistical matrix data set to derive the new offset, wherein the correction vector is defined by the principal direction of the distribution of the statistical parent data set a linear combination of the second basic vectors of the group, and representing the individual coefficients of the linear combination of the first and second θ^ basic orientations of the correction vector by one of the knives according to the statistical matrix data The main value ratio weighting a temporary offset is obtained from the individual #number clips of the old partial ^^ position inward. The (four) offset system (four) Η - 兽木攸 the statistic parent data set is derived without using the old offset, Temporary partial Chiang + / guide without linearly combining one of the quantities. Inch; - basic to 114821-981002.doc