Multidimensional Bernstein polynomials and Bezier curves: Analysis of machine learning algorithm for facial expression recognition based on curvature
Introduction
Bezier curves, which are constructed with the help of Bernstein basis functions, have a wide variety of applications in not only mathematics, engineering and medicine, but also almost all areas in recent years. Therefore, a variety of new applications and methods have been developed recently on these concepts (cf. [1], [2], [3], [4], [5], [6], [7], [8], [9], [10], [11], [12], [13], [14], [15], [16], [17], [18], [19], [20], [11], [12], [13], [14], [15], [16], [17], [18], [19], [20], [21], [22], [23], [24], [25]).
In this paper, in order to give a new dimension to these concepts, we investigate some theoretical properties of multidimensional unification of the Bernstein basis functions. We first give higher-order partial differential equations (PDEs) for these basis functions and their generating functions. Furthermore, by using these PDEs, we derive some recurrence relations for the multidimensional unification of the Bernstein basis functions. We also provide an image processing implementation of Bezier type curves for human facial expression recognition. By using C++ programming language with an computer vision library and machine learning algorithm, we give a facial expression classifier to show how Bezier type curves are used. While selecting the feature vectors to be used in the machine learning algorithm, we give a new method related to curvature of Bezier curves. Thanks to this method, a new dimension has been gained to the studies related to the facial recognition algorithms presented up to now.
Before giving implementation and computation, in order to convey the potential and motivation of the implementation in our work to the reader, we introduce not only our new results and identities related to multidimensional unification of the Bernstein basis functions and their generating functions, but also some open questions related to these results and Bezier type curves.
Here, we investigate some theoretical properties of multidimensional unification of the Bernstein basis functions.
In [25], Simsek defined the multidimensional(or m-dimensional) unification of the Bernstein basis functions as follows:where ; xj ∈ [0, 1], and with andand also
Generating functions for the multidimensional unification of the Bernstein basis functions are given by (cf. [25]):Note that there is one generating function for each value of and . Remark 1 If we set and in (1.1), we havewhere(cf. [17, p. 51, Eq. (13)], [4]). Remark 2 If we set and in (1.2), we havewhich denotes the unification of the classical Bernstein basis function given bywhere and with n ≥ b1s1 (cf. [31], [30, Theorem 1.1]). Hence, one can infer thatMoreover, if we set into (1.1), we getso that,which is given by Simsek in [30, Eq. (22)]. Remark 3 If we set and in (1.2), we havewhich denotes the classical Bernstein basis function given by(cf. [1], [21], [26], [27], [29], [30]), so that, when n < b1(cf. [17]; see also the references cited therein). Hence, one can infer that
In order to give an implementation of Bezier curves for human facial expression recognition in the next sections, we need to recall some properties of these curves which are related to Bernstein basis functions.
In [31], by using unification of Bernstein basis functions, Simsek unified nth degree of Bezier type curves with control points Pb, s as follows:
From the above curves, for we easily get nth degree of Bezier curves with control points by
Further properties of these curves are studied by many authors in many different areas (cf. [8], [9], [11], [14], [24], [27], [28], [29], [30], [31]).
Section snippets
PDEs for the multidimensional unification of the Bernstein basis functions and their generating functions
In this section, we give higher-order partial derivative formulas for the multidimensional unification of the Bernstein basis functions and their generating functions. By using these partial derivative equations, we derive recurrence formulas and identities for these basis functions.
Differentiating both side of (1.2), with respect to xr; 1 ≤ r ≤ m, we get the following theorem whose proof can be verified by direct derivative, with respect to xr; 1 ≤ r ≤ m, or induction on r. Theorem 1 Let 1 ≤ r ≤ m. We
Bezier curves with application of facial expression recognition based on curvature
It is well known that people are using facial expressions as complementary during communicating between each other in addition to verbal communication. Therefore, usage of facial expressions is now very common in the human and computer interaction applications. There exist many studies on facial expression recognition using Bezier curves with Hausdorff distance. On the other hand, in order to present a new dimension to the studies associated with the facial recognition algorithms given up to
Machine learning supported by statistical evaluations
In this section, by using statistical methods on feature vectors used during machine learning, it is shown how meaningful they are.
Conjecture and open questions
In this section, we state some open questions and conjecture related to subdivision property of the multidimensional unification of Bernstein (type) basis functions.
If y is a scaler, by using (2.8), in Section 2.2, we give some properties of subdivision property for the multidimensional unification of the Bernstein basis functions. However, if we replace y by the following conjecture and open questions appeared for subdivision property of the multidimensional unification of the Bernstein
Acknowledgments
The third author was supported by the Scientific Research Project Administration of Akdeniz University.
We thank the referees for their comments and valuable suggestions.
References (35)
The bernstein polynomials basis: a centennial retrospective
Comput. Aided Geom. Design
(2012)Pyramid algorithms, a dynamic programming approach to curves and surfaces for geometric modeling
The Morgan Kaufmann Series in Computer Graphics and Geometric Modeling
(2002)- et al.
Generating functions for the q-Bernstein bases
SIAM J. Discrete Math.
(2014) - et al.
A note on generating functions for the unification of the Bernstein type basis functions
Filomat
(2016) - User contract for Cohn-Kanade(CK+) database. Official web site:...
- Weka 3: Data Mining Software,...
- M. Acikgoz, S. Araci, On the generating function for Bernstein polynomials, in: American Institute of Physics...
Hausdorff distance between the offset curve of quadratic Bezier curve and its quadratic approximation
Commun. Korean Math. Soc.
(2007)- et al.
Destek vektör makineleriyle sınıflandırma problemlerinin çözümü için çekirdek fonksiyonu seçimi
Eskişehir Osmangazi Üniversitesi İBBF Dergisi
(2014) - et al.
A note on the generalized Bernstein polynomials
Honam Math. J.
(2011)
LIBSVM : a library for support vector machines
On the derivatives of Bernstein polynomials: an application for the solution of high even-order differential equations
Bound. Value Probl.
Analytic properties of plane offset curves
Comput. Aided Geom. Design
Formulas and algorithms for quantum differentiation of quantum Bernstein bases and quantum Bézier curves based on quantum blossoming
Graph. Models
Comprehensive database for facial expression analysis
Proceedings of the Fourth IEEE International Conference on Automatic Face and Gesture Recognition (FG’00), Grenoble, France
An approximation of circular arcs by quartic Bezier curves
Comput. Aided Des.
Detection and recognition of facial emotion using Bezier curves
IT Convergence Pract. INPRA
Cited by (17)
On the generating functions and special functions associated with superoscillations
2023, Discrete Applied MathematicsGenerating functions for polynomials derived from central moments involving bernstein basis functions and their applications
2024, Revista de la Real Academia de Ciencias Exactas, Fisicas y Naturales - Serie A: MatematicasCalculation of retention indices of essential oils with the aid of the Van den Dool and Kratz equation and Bézier curves
2024, Mathematical Methods in the Applied SciencesDetection of fatigue driving behavior based on facial expression
2024, Journal of Intelligent and Fuzzy Systems