Indonesian J our nal of Electrical Engineering and Computer Science V ol. 25, No. 3, March 2022, pp. 1297 1307 ISSN: 2502-4752, DOI: 10.11591/ijeecs.v25.i3.pp1297-1307 1297 The measur ement set r epr esentation of the body postur e based on gr oup theory Changjian Deng, Xiaona Xie School of Control Engineering, Chengdu Uni v ersity of Information T echnology , Chengdu, China Article Inf o Article history: Recei v ed Jun 22, 2021 Re vised Dec 20, 2021 Accepted Jan 11, 2022 K eyw ords: Attitude sensor Body posture Group theory Lie group Representation ABSTRA CT The foundation of measurement is the representation of measurement set. The paper proposes a body posture measurement set representation method based on concepts of group, elds, and ring. It attempts to e xplore the intrinsic relationship among numer - ous dif ferent measurement set. The attitude sensor is used to measure the attitude, and the measurement set representation and processing are analyzed based on the lie group theory in the paper . In the paper , the paper maps body posture data into image, and it is easier to identify the error of posture measurement data. Meanwhile, the simulation and test results sho w that the method can represent body posture data and detect it’ s defects easily . This is an open access article under the CC BY -SA license . Corresponding A uthor: Changjian Deng School of Control Engineering, Chengdu Uni v ersity of Information T echnology No. 24, Block 1, Xuefu Road, Chengdu City , China Email: chenglidcj@cuit.edu.cn 1. INTR ODUCTION Normally , measurement refers to a group of operations to determine the “quantity v alue”. The pur - pose of measurement is to mak e the quantity information contained in the object ob vious and comparable. In comple x and specic application, the measured quantity may no longer be a scattered string of irrele v ant data. These data and their changes may be manifolds or groups. Exploring these rela tionships is an important method to ensure the accurac y of measurement and control. This paper mainly studies this problem. The de- v elopment of group theory pro vides a basis for these studies. No w adays, group theory has man y applications [1]. F or e xample, it is also a po werful tool in signal processing [2], [3], signal representation [4], and me- chanics [5]. This paper is mainly concerned with the representation of group theory . The related research of representation includes tw o catalogues, one is signal representation and detection [6]-[9], and the other is deep learning and object tracking [10]-[12]. The paper focuses on measurement set representation based on group theory . The contrib utions of paper include: 1) it proposes image representation method; 2) it presents body posture representation and analysis method based on lie group. The paper is or g anized as follo ws: in section 2, we present the related w ork. In section 3, we present the measurement sets and group representation. Section 4 focusi ng on measurement set map. The e xperiments and simulation are presented in section 5. The section 6 is conclusion. J ournal homepage: http://ijeecs.iaescor e .com Evaluation Warning : The document was created with Spire.PDF for Python.
1298 ISSN: 2502-4752 2. MEASUREMENT SETS AND GR OUP REPRESENT A TION 2.1. Fields of measur ement The measurement processing often is undertak en in dif ferent domains, for e xample: the time -frequenc y electronic magnetic domain; the tensor mechanical domain; the thermotical domain; ener gy and mass domain; time and space domain; and so on. In nature, most quantity of entity to be measured is represented by signal as signal processing is algebraic in nature. The measurement processing is algebraic in nature: The real v oltage signal set V in time domain is a group. The real v oltage signal set V in time domain is a ring. The real v oltage signal set V in time domain is a eld. Generally , measurable sets normally itself and its map are elds. Domain is structured sets. In group analysis, lagrange theorem and homomorphism principle in group theory are the basis of the analysis of measurement set. 2.2. Continue medium measur ement sets The measurement set may come from dif ferent time test, dif ferent space test, or dif ferent circ u m stance test. In a ri gid body , architecture, material, quantity form their dif ferent part and their moti v ation form a continue measurement sets. Here we discuss tw o kinds of problem. The y may be useful in sensor , moti v ation analysis, and so on. 2.2.1. The rst kind of pr oblem is perf ormance analysis and defect detection of material in sensor tech- nology F or e xample, the analysis the dielectric coef cient, piezoelectric coef cient, piezo-resistance coef - cient, and so on. There are some related softw are e xamples of group in engineering applicati on. Puschel and Moura [13] proposed the ISO TR OPY softw are tools, it collections softw are which applies group theoretical methods to the analysis of phase transitions in crystalline solids. Method of analysis the measurement related coef cient: as an introduction, here gi v e an e xample of dielectric coef cient of cubic dielectric capacitance sensor . And it can be pro v ed that this coef cient is scalar quantity . Pro v e [14]: assume the tensor of dielectric coef cient is ϵ αβ ( α , β = 1 , 2 , 3 ), then: D α = P 3 β =1 ϵ αβ E β select axis, let E = E j D α = P α,β ϵ αβ E δ β 2 = ϵ β 2 E = E αy E ( α = x, y , z ) let axis ‘y’ as basis axis of rotation, do C 1 4 operation: z z = x, x x = z , then, D x = D z , D z = D x . , with the in v arity property of C 4 , so, D x = D x , D z = D z . , then we get: ϵ xy = ϵ z y ϵ xy = ϵ z y so, ϵ xy = ϵ z y = 0 , and it is same as: ϵ xz = ϵ y z = ϵ y x = ϵ z x = 0 then let E along the diagonal of a cube ( C 3 axis), then: E x = E y = E z = 1 3 E , then: Indonesian J Elec Eng & Comp Sci, V ol. 25, No. 3, March 2022: 1297–1307 Evaluation Warning : The document was created with Spire.PDF for Python.
Indonesian J Elec Eng & Comp Sci ISSN: 2502-4752 1299 D x = 1 3 ϵ xx E D y = 1 3 ϵ y y E D z = 1 3 ϵ z z E do C 1 3 and C 2 3 operation, and consider the in v arity of C 3 . Then D x = D y = D z and ϵ xx = ϵ y y = ϵ z z = ϵ , and get the solution. 2.2.2. The another kind of pr oblem is faults or distortion caused by material defects, the important tool is homotr opy gr oup No w adays, point defect research in semic on duct ors has g ained remarkable ne w momentum due to the identication of special point defects that can implement qubits and single photon emitters with unique characteristics. The defect of continuum can be represented using homotrop y group. Homotrop y groups in order parameter space T = G/H are used to represent and classify defects. The union of defects is obtained by multiplication of homotrop y groups. In the plane ordered space eld T , the winding number n is an importa nt parameter . F or the images with the same winding number n, the y can be changed into the same features by local adjustment. And the kind of defects wi th same n can be transformed into each other , which is barrier free in topology . The y are same homotrop y group. Method of analysis the defect of planar spin system, planar spin system [14]: Its Order parameter: S = i cosθ + j sinθ T ransformation group: G=T(1); T ϕ ( θ ) θ ϕ Isotropic group: H= T ( n ) , n = 0 , ± 1 , ± 2 , ..., Discrete group: H 0 = e . So, Q 1 ( T ) = H = Z , this is additi v e group of inte gers. 2.3. Measur ement sets r epr esentation method As sho wn in T able 1, dif ferent measurement processing has dif ferent method to represent their p a ram- eters. Some rules are : a.) P arameters (measurement sets) can represent the main property of object visually , mapping them into graph- based data though positi v e and ne g ati v e direction calculation (mapping). T able 1 gi v e some e xist method and its e xtending. An e xample is sho wn in Figure 1. b .) This graph-based measurement set should be measurable and has some dened mathematical fe ature (or Algebraic). T able 2 is an e xample of body posture measurement set and its group representation. T able 1. P arameters representation Properties Graph method Infra thermo-image Pix el temperature map to image Nature sound T ime-frequenc y mapping Automoti v e radar Geometry parameters mapping Soil temperature of dif ferent place T emperature map to image Figure 1. The diagram of measurement set map The measur ement set r epr esentation of the body postur e based on gr oup theory ... (Changjian Deng) Evaluation Warning : The document was created with Spire.PDF for Python.
1300 ISSN: 2502-4752 T able 2. P arameters measurable representation P arameters Algebraic method Acceleration and Gyroscope signal data set Perfect linear matrix group Eulerian angle measurement data Perfect linear matrix group Defect data Homotrop y group When we map these data into image, it is easy to see some defect of sensor or measurement error . The Figure 2 is acc data of three dif ferent sensor , the Figure 3 is their gyro data. The measurement en vironment is static. So, it is v ery clear that there is something wrong in sensor 2. And this is true case. W e use Kalman lter and other methods to deal with the acc data and gyro data, then we get the Eulerian angle data. It is sho wn in Figure 4. Figure 2. The acc measurement set mapping Figure 3. The h yro measurement set mapping Indonesian J Elec Eng & Comp Sci, V ol. 25, No. 3, March 2022: 1297–1307 Evaluation Warning : The document was created with Spire.PDF for Python.
Indonesian J Elec Eng & Comp Sci ISSN: 2502-4752 1301 Figure 4. The Eulerian angle data set mapping In planar operat ion, the rigid status is dened by the angle of the joint. In n-dimensional real space, i t is S O (2) × S O (2) ... × S O (2) . In this Algebraic structure, the group operators are dene d based on addition operation. Group addition operation: q 1 , q 2 n , q 1 q 2 = q 1 + q 2 zero element: E = 0 = [0 , 0 , ..., 0] T in v erse operation: q n , q = q scalar product: α , q n , α q = q q ... q = α · q space rigid body posture often is dened by 3 × 3 matrix, the group operator is dened by the matrix product, that is SO(3). Group addition operation: R 1 , R 2 S O (3) , R 1 R 2 = R 1 R 2 zero element: E = I 3 = diag ([1 , 1 , 1]) in v erse operation: R S O (3) , R = R 1 scalar product: α , R S O (3) , α q = R R ... R = R · R ... · R . The measur ement set r epr esentation of the body postur e based on gr oup theory ... (Changjian Deng) Evaluation Warning : The document was created with Spire.PDF for Python.
1302 ISSN: 2502-4752 3. MEASUREMENT SET MAP 3.1. Fr om ODE (ordinary differ ential equation) to gr oup In sensor technology , the polynomials is used to stand for the standalone static characteristics of the sensor system and ordi nary dif ferential equation is used to stand for the dynamic feature of t he linear time in- v ariant sensor system. More generally , the state space equation is used to stand for the system motion equation. Polynomials are created from names, inte gers, and other v alues using the arithmetic operators + , , , and ÷ . Ordinary dif ferential equation ha v e dif ferent types of ODE problems. Here, we consider it in the algebraic w ay . W e use mapping to represent the characteristic transformation processing (see on Figure 5). Figure 5. The diagram of measurement set map The authors denote Algebras by C the set of comple x numbers. A C -algebra A is a C-v ector space that is also a ring [11], [12]. The y also proposed that lter spaces is algebras, lter as metrics. F or e xample, if module M is of is of dimension n with basis b = ( b 0 , ..., b n 1 ) and s C n , then: ϕ ( s ) = s = n 1 X i =0 s i b i (1) and its z-transform, fourier transform. Some realization of the abstract space model is sho wn in T able 3. T able 3. Realization of abstract model Concept Abstract Realized Shift opetator q T 1 ( x ) = x Space mark t n C n K-fold shift operator q k = T k ( q ) T k ( x ) Signal P s n t n P s n C n ( x ) Filter P h k T k ( q ) P h k T k ( x ) Assumption: the static and dynamic feature of (sensor or control) system, are often also d e ned by error and other parameters partial dif ferential equation. And in ph ysics w orld, the enti ty state is reasonable to be represented. So, no r mally it is represented by eld, ring, group, or other kinds of math domain. It is reasonable and con v enience that som e measurement quantity or quality are represented by group. An y this may be their inner comparable properties of measurement. Lie group analysis (symmetry anal ysis) nds point transformations which map a gi v en dif ferential equation to itself [15]. Exce pt SO(2), the Af f(2) and SO(3) are also often used. SO(3) is the group of rotations in 3D space, represented by 3 orthogonal matrices with unit deter - minant. It has three de grees of freedom: one for each dif ferential rotation axis. The in v erse is gi v en by the transpose: R S O (3) R 3 × 3 Indonesian J Elec Eng & Comp Sci, V ol. 25, No. 3, March 2022: 1297–1307 Evaluation Warning : The document was created with Spire.PDF for Python.
Indonesian J Elec Eng & Comp Sci ISSN: 2502-4752 1303 R 1 = R T det ( R ) = 1 Af f(2) is the group of af ne transformations on the 2D plane. It has si x de grees of freedom: tw o for translation, one for rotation, one for scale, one for stretch and one for shear . Subgroups include Sim(2). After the discussion abo v e, there are some methods to represent the measurement set mapping, the y are sho wn in Figure 5. The processing or analysis is al w ays undertaking de v eloping while a object is mo ving. The better method is the timely solution and wi thin the timeout by some distrib ution or human being action. And also, for the mo ving properties of objects, the best solution is v arying with time [16]. The representation of parameters measurable is sho wn in T able 4. T able 4. P arameters measurable representation General operation methods Lie Group the normalized measurement set Measurement set and its map nite elds Measurable mo v ement or error e xist rule Modern algebra the structure of measurement set 3.2. J oint space mapping Joint space is calculated by linear interpolation of each joint. F or attitude, it is possible to nd the rotation axis between the tw o attitudes, and then interpolate t he rotation angle. Linear interpolation of Lie group: A α = A 1 α (( A 1 ) A 2 ) , 0 α 1 . In Joint space: q α = q 1 α (( q 1 ) q 2 ) (2) q α = q 1 + α (( q 1 ) q 2 ) (3) q α = q 1 + α ( q 2 ) q 1 ) , 0 α 1 (4) for attitude: R α = R 1 α (( R 1 ) R 2 ) (5) R α = R 1 + α (( R 1 ) R 2 ) (6) R α = R 1 · ( R 1 ) 1 · R 2 ) α , 0 α 1 (7) 4. EXPERIMENTS AND SIMULA TION 4.1. Action r ecognition using lie netw ork based on sk eleton data set F or sk eleton data has follo wing feature [16]: each input is an element on the lie group. The sk eletal data can be e xhibitted as fully connected con v olution-lik e layers and pooling layers, it has a deep netw ork architecture of the Lie group. The train processing is sho wn in Figure 6. The measur ement set r epr esentation of the body postur e based on gr oup theory ... (Changjian Deng) Evaluation Warning : The document was created with Spire.PDF for Python.
1304 ISSN: 2502-4752 Figure 6. The training processing of lie netw ork 4.2. Body postur e r ecognition by MPU6050 It is the direct angle dat a obtained by the data fusion algorithm between gyroscope and accel eration sensor . The deplo ying place MPU6050 sensor is sho wn in Figure 7. Lik e sk eleton data and lie group feature, dif ferent body postures are sho wn in Figures 8-11. It is ob vious that the relationship of Euler angle fol lo ws the theory of rigid body mechanics. And it is foundation of recognition of body posture. Figure 7. The MPU6050 sensor deplo ying place on body Indonesian J Elec Eng & Comp Sci, V ol. 25, No. 3, March 2022: 1297–1307 Evaluation Warning : The document was created with Spire.PDF for Python.
Indonesian J Elec Eng & Comp Sci ISSN: 2502-4752 1305 Figure 8. The body posture no. 1 Figure 9. The body posture no. 2 Figure 10. The body posture no. 3 The measur ement set r epr esentation of the body postur e based on gr oup theory ... (Changjian Deng) Evaluation Warning : The document was created with Spire.PDF for Python.
1306 ISSN: 2502-4752 Figure 11. The body posture no. 4 5. FUTURE STUD Y There are some paper [17]-[21] that study the coding-encoding technology , in the future research of the manuscript we should consider the method based on concepts of group. In articles [22]-[25] research the mathematical e xpression of symmetry in ph ysics using group theory , in the future research of the manuscript we should studies symmetry in ph ysics more deeply . 6. CONCLUSION The paper proposes a body posture measurement set representation method based on concepts of group, elds, ring. It attempts to e xplore the intrinsic relationship among numerous dif ferent measurement set. Meanwhile, the paper maps body posture data into image. The contrib utions of paper are: 1) it proposes image representation method; 2) it presents body posture representation and analysis method based on lie group. A CKNO WLEDGMENT The paper is supported by 2019R YJ03 Open fund project of Sichuan K e y Laboratory of articial intelligence. REFERENCES [1] X. B. W ang, Q. L. Zhang, Y . Zhou, and H. X. Zhou, “Unication of signal transforms on groups: An introduction, SCIENTIA SINICA Informationis , v ol. 43, no. 12, pp. 1547-1562, 2013, doi: 10.1360/N112013-00153. [2] M. P ¨ uschel and J. Moura, Algebraic signal proce ssing theory: foundation and 1-D time, IEEE T r ansactions on Signal Pr ocessing , v ol. 56, no. 8, pp. 386-391, 2006, doi: 10.1109/TSP .2008.925261. [3] A. Sandryhaila, J. K o v ace vic, and M. Puschel, Algebraic signal processing theory: 1-D Nearest Neighbor models, IEEE T r ansac- tions on Signal Pr ocessing , v ol. 60, no. 5, pp. 2247-2259, 2012, doi: 10.1109/TSP .2012.2186133. [4] A. Goodall,“F ourier analysis on nite abelian groups: some graphical applications, Combinatorics, Comple xity , and Chance: A T rib ute to Dominic W elsh , 2007, doi: 10.1093/acprof:oso/9780198571278.003.0007. [5] Y . Gu, “Image singularities of green’ s functions for anisotropic elastic half-spaces and bimaterials, The Quarterly J ournal of Me- c hanics and Applied Mathematics , v ol. 45, no. 1, pp. 119–139, 1992, doi: 10.1093/qjmam/45.1.119. [6] L. Daniels and T . Bay , “Group theory and the rubik’ s cube’, Lak ehead Univer sity , 2014. [7] V . Dabbaghian-Abdoly , An algorithm for constructing representations of nite groups, J ournal of Symbolic Computation , v ol. 39, no. 6, pp. 671–688, 2005, doi: 10.1016/j.jsc.2005.01.002. [8] M. C. Colin, H. Geor ge, H. Ale xander , and F . R. Edmund, “Ef cient simple groups, Communications in Alg ebr a , v ol. 30, no. 10, pp. 4613–4619, 2002, doi: 10.1081/A GB-120023154. [9] B. Eick and C. R. B. Wright, “Computing subgroups by e xhibition in nite solv able groups, J ournal of Symbolic Computation , v ol. 33, no. 2, pp. 129-143, 2000, doi: 10.1006/jsco.2000.0503. [10] Z. Huang, C. W an, T . Probst, and L. V . Gool, “Deep learning on lie groups for sk eleton-based action recognition, In Pr oceedings of the IEEE confer ence on computer vision and pattern r eco gnition , 2017, doi: 10.1109/CVPR.2017.137. [11] N. Boumal, B. Mishra, P . A. Absil, and R. Sepulchre, “Manopt , a Matlab toolbox for optimization on manifol ds, J ournal of Mac hine Learning Resear c h , v ol. 15, no. 1, pp. 1455–1459, 2014. Indonesian J Elec Eng & Comp Sci, V ol. 25, No. 3, March 2022: 1297–1307 Evaluation Warning : The document was created with Spire.PDF for Python.