DeepDyve requires Javascript to function. Please enable Javascript on your browser to continue.
A Russian Continuous Speech Recognition System Based on the DTW Algorithm under Artificial Intelligence A Russian Continuous Speech Recognition System Based on the DTW Algorithm under Artificial...
Yu, Chunping;Wang, Xin
2022-09-19 00:00:00
Hindawi Journal of Robotics Volume 2022, Article ID 5777472, 11 pages https://doi.org/10.1155/2022/5777472 Research Article A Russian Continuous Speech Recognition System Based on the DTW Algorithm under Artificial Intelligence 1 2 Chunping Yu and Xin Wang International Education School, Zhengzhou Railway Vocational and Technical College, Zhengzhou 451460, Henan, China International Exchange and Cooperation O ce, Zhengzhou Railway Vocational and Technical College, Zhengzhou 451460, Henan, China Correspondence should be addressed to Chunping Yu; yuchunping@zzrvtc.edu.cn Received 18 July 2022; Revised 23 August 2022; Accepted 29 August 2022; Published 19 September 2022 Academic Editor: Shahid Hussain Copyright © 2022 Chunping Yu and Xin Wang. �is is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited. In order to improve the eƒect of continuous speech recognition, this paper combines the DTW algorithm to construct a continuous Russian speech recognition system and proposes a continuous Russian speech detection method based on VGDTW- MPCA with an unequal interval process. Moreover, considering the in‘uence of the correlation between variables on the synchronization of the DTW algorithm, this paper constructs a DTW algorithm on a local data set to synchronize in diƒerent variable groups. �en, this paper integrates the obtained data into complete 3D data for modeling. It can be seen from the simulation research that the Russian continuous speech recognition system based on DTW proposed in this paper has a high continuous Russian speech recognition accuracy. technological exchanges and contribute to the development 1. Introduction of the country, which is precisely the original mission of �e training goal of foreign language subjects is to cultivate foreign language talents given to foreign language talents [3]. applied, compound, and innovative talents with international At present, there is still a big gap between Russian vision and cross-cultural communication skills, familiarity teaching and English. �e teaching methods are more tra- ditional and the subject research is relatively lagging behind. with international conventions, ability to participate in in- ternational competition and cooperation, with a sense of �erefore, while establishing a reasonable and complete social responsibility, and adaptability to the needs of eco- curriculum system and innovative Russian teaching nomic construction and social development. �is is also the methods, the teaching staƒ should be strengthened. Con- “authentic” that Russian subjects should abide by [1]. Under struction is particularly important. As a key factor the background of the new era and historical opportunities, throughout the entire teaching process, teachers’ professional one of the important missions of the Russian subject is to quality and ideological realm are important conditions for enhance the country’s soft power and shape the country’s the survival of Russian teaching [4]. Russian teachers also hard image in state exchanges. Moreover, the Russian- need to change their concepts, strive to improve their own speaking talents cultivated under this goal and mission knowledge, and strengthen the integration of theory and should form the consciousness of national culture [2]. In practice. Combining and handling the relationship between teaching and scientiœc research, it is still necessary to make today’s world, Russia is still a pivotal regional power with an important position in the world. At the same time, Russia is eƒorts to transform into a “dual-teacher type.” At the same still a world leader in basic research, military T, and aerospace time, Russian teachers should continue to enhance their own technology. �erefore, Russian-speaking talents still need to knowledge reserves, strengthen connections with other give full play to their due role in Sino-Russian scientiœc and disciplines, and prepare for the challenges of the new era [5]. 2 Journal of Robotics curriculum transformation. From a specific point of view, From the perspective of teaching methods, the current teaching activities of public Russian courses in colleges and colleges and universities can give full play to the advantages of the school-enterprise cooperative teaching mode, and hire universities have a single defect in teaching methods. First of all, in content design, more emphasis is placed on guiding Russian-speaking staff with rich business and trade work students to learn basic Russian vocabulary and grammar, experience from the cooperative enterprises along the “Belt and the application of Russian in actual communication and Road,” or professional and technical personnel from scenarios. Knowledge is not paid enough attention, which countries along the “Belt and Road.” As part-time teachers of leads to insufficient teaching quality of public Russian Russian courses for students, teachers can enrich teacher’s courses in colleges and universities and it is difficult for Russian professional knowledge, and improve students’ professional education ability in the process of continuous students’ practical application ability to effectively fit with the knowledge they have learned [6]. From the perspective of practice, so as to further improve the teaching staff [11]. In addition, the introduction of postgraduates with overseas the course evaluation model, many courses and teachers are currently affected by the traditional test-oriented education experience should be increased, especially bilingual teachers with certain knowledge backgrounds, to meet the needs of concept and they do not pay enough attention to the de- velopment of students’ application skills in the course Russian teaching in colleges and universities. We improve teaching process. In the current process of Russian teaching the educational level of imported foreign teachers and en- in colleges and universities, many teachers have the phe- sure the employment of foreign teachers. Foreign teachers nomenon of “emphasizing theoretical results and ignoring play a crucial role in Russian teaching. *e imported foreign practical application” when setting up a course evaluation teachers should have a master’s degree or above and have system. Most Russian course exams are conducted through certain teaching experience. *e subject should be Russian as a foreign language, or bilingual teachers with railway pro- written tests, and the written test scores are used as the result of students’ course learning, which makes it difficult for the fessional background knowledge [12]. *e core courses in the talent training program for course evaluation model to accurately reflect students’ learning [7]. Russian professionals include basic Russian, advanced Russian, Russian grammar, Russian reading, Russian audio- In terms of education training and job training, colleges and universities should do a good job in education ability visual, and Russian writing. *e curriculum focuses on training and job training for Russian teachers and funda- helping students lay a solid foundation and develop language mentally improve teachers’ information-based teaching by skills. However, it should be noted that the needs of regional building a good practice platform and collective training for economic development, and the employment direction of teachers. In addition, it provides opportunities for teachers students should also be taken into consideration when to communicate with relevant professionals in the course, so designing talent training programs and writing syllabuses. In terms of teaching methods, traditional teaching methods are as to strengthen the training of teachers of public Russian courses and achieve the effective construction of Russian still used, which are mainly explained by teachers and passively accepted by students. *is single model is boring to teaching faculty [8]. Prejob training for young and middle- aged Russian teachers can be increased, as professional today’s students and it is difficult to obtain good results [13]. training for young teachers who are about to take the stage, We give full play to the advantages of comprehensive learning basic teaching norms, understanding teaching re- colleges and universities and actively carry out inter-pro- quirements, and inviting old teachers to impart teaching fessional exchanges within the school. Russian teachers and experience. We implement the “mentor-apprentice” system: teachers of international trade, international law, and other employ experienced Russian professors or excellent English majors regularly hold academic forums to gain an in-depth teachers in our school as teaching tutors for young Russian understanding of relevant knowledge and cutting-edge teachers to help young Russian teachers grow as soon as theories. Teachers of Russian majors are encouraged to apply for scientific research projects jointly with teachers of other possible [9]. *rough activities such as teacher skills com- petitions, teaching seminars, and observations, the training majors to break down disciplinary barriers and promote interdisciplinary research. However, we make full use of of basic teaching skills for young Russian teachers will be strengthened, teachers will be encouraged to carry out research results in teaching, constantly update teaching teaching research, professional curriculum reform, scientific content, improve teaching quality, encourage students to research, actively participate in social services, and improve think and explore, and cultivate the spirit of scientific re- teachers’ comprehensive ability. From time to time, expe- search and innovation [14]. Advocate exchange visits and rienced experts and professors from outside the school are enrich teaching experience. We support the exchange of invited to our school to give special reports and academic teachers between our school and foreign institutions for further study or visiting exchanges, encourage teachers to lectures to improve the professional qualifications of teachers. Finally, we must work hard on the structure of actively apply for the China Scholarship Council project, participate in international seminars, and introduce the teachers [10]. In the process of carrying out public Russian courses in colleges and universities, the improvement of foreign advanced school-running experience and high- quality educational resources. In terms of inter-school co- teachers not only requires the efforts of teachers themselves but also requires colleges and universities to introduce as operation, it attaches great importance to establishing many high-quality Russian teachers as possible and promote contacts with domestic colleges and universities, regularly the structure of the teaching staff from “single” to “Diversity” organizes teacher training exchanges, and invites Journal of Robotics 3 authoritative scholars and experts to give teaching lectures mxn grid searched on the basis of the shortest distance D (m and share valuable teaching experiences. We strengthen and n),asshowninFigure1.*eelementintheoptimalpathis communication with enterprises and continuously update thequantizationvalueoftheanisotropyofthesequenceinthe the knowledge system. Russian teaching work must be matching process, and the curved path represents the optimal “grounded,” and teachers must first go to practice [15]. path for the matching of the two sequences. As an inseparable whole, language and culture play According to the feasibility of the actual situation, the complementary roles in the teaching of Russian. Students curved path P needs to meet the following two conditions: integrate their understanding of Russian culture into the (1) Endpointconstraint:thefirstandlastelements p and process of learning the Russian language, which can better p oftheoptimalpath Pcorrespondtotheelementsat promote their mastery of the language. In this regard, foreign both ends of the diagonal of the distance matrix D, so teachers have unique advantages. Teachers and students who that the stretched and compressed new data, and the haveneverlivedinaRussian-speakingcountry, theyhavenever original data keep the same starting point and ter- experienced the national conditions, culture, customs, etc., of a mination point, as shown in the following formula: Russian-speaking country but only learn some superficial knowledge from books [16]. In the process of teaching and p � D(1,1), (4) working with teachers, foreign teachers can use personal ex- amples to explain to students or set up some special teaching p � D(m, n). (5) links for students to specifically reflect, so that students can learn relevant language knowledge in these processes, and at (2) Global constraints: the selection of the optimal path the same time enable teachers too. In particular, young teachers P is generally limited to a certain range. It is ben- have learned knowledge that cannot be learned in books and eficial to improve the calculation efficiency of the has a better understanding of Russian culture [17]. cumulative distance and avoid the appearance of *is paper combines the DTW algorithm to construct a abnormal paths. As shown in Figure 1, the optimal Russian continuous speech recognition system and im- path is bounded within two dashed diagonals. proves the Russian teaching effect through artificial intel- Local constraints define an optional preceding point for ligence methods. each point and specify the continuity of the midpoints of the path. *e continuity of the midpoint of the path avoids 2. DTW Continuous Speech excessive distortion and jumping off the original data. Recognition Algorithm As shown in Figure 2, the local constraint (1) can only have the following three options before point (i and j) in the 2.1. (e Principle of the DTW-MPCA Algorithm. We assume grid: point (i−1 and j), point (i and j−1), Point (i−l and j−1). that there are two-time series R and C, the length of the series *erefore, formula (2) can be obtained by formula (6): R is m, and there is R � (r , r , L, r , L, r ). *e length of the 1 2 i m sequence C is n, there is C � (c , c , L, c , L, c ). First, in 1 2 j n D(i − 1, j) + d(i, j) ⎧ ⎪ ⎫ ⎪ ⎪ ⎪ order to use the DTW method to synchronize the two-time ⎨ ⎬ D(i, j) � D(i, j − 1) + d(i, j) . (6) series, a path matrix d of mxn needs to be defined in advance. ⎪ ⎪ ⎪ ⎪ ⎩ ⎭ Among them, the elements of the (i and j) position of the D(i − 1, j − 1) + d(i, j) matrix represent the distance between the two points r and As shown in Figure 2, the local constraint (2) can only c . *e formula for calculating the distance is as follows: have the following three options before point (i and j) in the � � � � � � � � grid: point (i−1 and j), point (i−1 and j−1), point (i−1and d r , c � r − c . (1) � � i j i j j−2). *erefore, formula (2) can be obtained by formula (7): Among them, ‖ · ‖ represents the p norm, and p usually D(i − 1, j) + d(i, j) ⎧ ⎪ ⎫ ⎪ ⎪ ⎪ takes 2. ⎨ ⎬ D(i, j) � D(i, j − 1) + d(i, j) . (7) *e goal of dynamic time warping is to find the shortest ⎪ ⎪ ⎪ ⎪ ⎩ ⎭ distance D (m, n) between two-time series and the point D(i − 1, j − 2) + d(i, j) sequence of the optimal path P. *e optimal performance indicators of the shortest distance D (m, n) and the optimal path P are as follows: 2.2. Principle of the MPCA Algorithm. *e three-dimensional K data set for the unequal-length batch process is ⎡ ⎣ ⎤ ⎦ D(m, n) � min d(i(k), j(k)) , (2) X(I × J × K )(i � 1,2, L, I). Among them, I represents the k�1 number of batches, J represents the number of variables used for offline modeling, and K represents the reaction time. In P � argmin[D(m, n)]. order to analyze the batch process using multivariate statis- (3) tical methods, it is necessary to perform synchronization Amongthem,thereismax(m, n)≤ K≤ m+ n.*eshortest preprocessing on the 3D data set X(I × J × K ) (i � 1,2, L, I). distance D (m and n) is the sum of the local shortest distances *en, the complete 3D dataset X(I × J × K) after synchro- ofthetwosequencesalongtheoptimalpath.*eoptimalpath nization is expanded into a 2D matrix X(I × JK) based on P � (p , p , L, p ),p � (i(k), j(k)) is a point sequence in an batches, and the expansion method is shown in Figure 3. 1 2 k k 4 Journal of Robotics j (i−1,j) (i,j) (i−1,j) (i,j) (m,n) (i−1,j−1) (i−1,j−1) (i,j−1) (i,j−2) Local constraints(1) Local constraints(1) Figure 2: Schematic diagram of local constraints. Amplitude Point number of sampling Figure 1: Schematic diagram of the principle of DTW algorithm. X(IxJxK) In order to obtain the average running trajectories of the measurement variables on multiple batches of data, this 12JK J J paper performs the normalization processing of formulas (1)–(4) on the two-dimensional matrix X(I × JK) after the X(IxJK) three-dimensional data set is expanded. *e standardized data set approximately obeys the multidimensional normal distribution and can show the fluctuation of process vari- Figure 3: Schematic diagram of a batch expansion of 3D data. ables in different batch intermittent operations, which has a certain degree of statistical significance. After the batch process, the 3D dataset is preprocessed, variables and time changes, and E represents the residual and the expanded 2D matrix X(I × JK) is analyzed using matrix containing secondary information and noise. traditional PCA. In fact, it extracts feature information based Statistics used to monitor multivariate batch processes 2 2 on the variance between different batches of data and the are the T statistic and the SPE statistic. Among them, the T covariance of different variables at different sampling times. statistic monitors the change information of the data in the *e variance matrix (X X)/(I − 1) of the two-dimensional pivot space, and the SPE statistic monitors the change in- matrix X(I × JK) is obtained by the singular value de- formation of the data in the residual space. During offline composition algorithm as follows: modeling, the T statistic and SPE statistic are computed for batch i as follows: T T T T X X X X X X L X X 1 1 1 1 1 1 1 K ⎥ 2 T −1 ⎡ ⎢ ⎤ ⎢ ⎥ ⎢ ⎥ ⎢ ⎥ (10) ⎢ ⎥ T � t S t , ⎢ ⎥ ⎢ ⎥ i i t i ⎢ ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ T T T T ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ X X X X X X L X X ⎥ ⎢ ⎥ ⎢ 1 2 3 K ⎥ ⎢ 2 2 2 1 ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ ⎥ T ⎢ ⎥ T ⎢ ⎥ ⎢ (11) ⎢ ⎥ SPE � e e , ⎢ ⎥ ⎢ ⎥ i i i X X 1 ⎢ ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ T T T T ⎥ ⎢ ⎥ � ⎢ ⎥. (8) ⎢ ⎥ ⎢ X X X X X X L X X ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ 3 1 3 2 3 3 1 K ⎥ ⎢ ⎥ I − 1 I − 1 ⎢ ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ ⎥ T ⎢ ⎥ ⎢ ⎥ ⎢ ⎥ e � X − PP X . (12) ⎢ ⎥ ⎢ ⎥ i i i ⎢ ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ M M M M ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ ⎥ ⎣ ⎦ Among them, e represents the i-th row of the residual T T T T X X X X X X L X X matrix, t is the score vector, and S is the covariance matrix K 1 K 2 K 3 K K i t estimated from the pivot score matrix obtained during Among them, X X represents the variance informa- modeling. tion between measurement data at the same time, and For the monitoring data of the MPCA online process, X X represents the covariance information between K2 K1 only the current time and previous values are known. *e different measurement data at different times. *e decom- sampling data in the actual monitoring process are imperfect, position process is as follows: resulting in insufficient known data to constitute a complete sampling of the intermittent process. *erefore, it is necessary to use data-filling technology when applying MPCA to online X � t p + E. (9) r r monitoring. *is paper adopts the method of filling the r�1 current sampling value, that is, it is estimated that the data Among them, R represents the number of retained after the current moment is the same as the current moment. principal components, t represents the relationship be- *e calculation of the T statistic for online monitoring tween batches, p represents the relationship between is performed at every sampling time k of batch i as follows: Amplitude Point number of sampling Journal of Robotics 5 2 T −1 H(X, Y) � H(X) − H(X|Y) T (k) � t (k)S t (k). (13) i i t i Moreover, the SPE statistic is the square of the error at � H(Y) − H(Y|X) the sampling instant k as follows: JK � H(X) + H(Y) − H(X, Y) SPE (k) � e e . (14) i c c�(k−1)j+1 1 1 � p(x)log + p(y)log (20) p(x) p(y) Among them, e is the residual matrix E � X − x y c new t P . new *e control limits for the T statistic and the SPE − p(x, y)log statistic are calculated under the assumption that the p(x, y) x,y modeled data are normally distributed. In this paper, kernel density estimation (KDE) is used to calculate the p(x, y) control limits. For any variable z, the probability density � p(x, y)log . p(x)p(y) function of its radial basis kernel function is expressed as x,y follows: As shown in Figure 4, the mutual information between 1 1 z − z i variable x and variable x in a batch of batch process data x i j f(z) � √�� � exp . (15) 2π (K × J) (J �10) is I(x , x )(i, j � 1,2, k, J). *e greater the 2c i j i�1 mutual information value between the two variables, the Among them, c is the nuclear parameter, which is de- closer the relationship between the two variables. *e termined by the cross-validation method in this paper. *en, mutual information value of the two variables x; and x; can the integral of formula (15) is calculated in the definition be calculated, and the specific calculation method is shown domain of the variable z. When the integral reaches the set in formula (21). confidence limit, the corresponding value is the control limit 1 1 obtained by the KDE method. Ix , x � p x log + px log i j i j p x p x i x x i j 2.3. Variable Grouping DTW-MPCA Method. Two random 1 − px , x log i j variables X and Y are given, p (x) and p (y) are the marginal (21) px , x x ,x i j i j probability densities of the random variables X and Y, re- spectively. p (x and y) is the joint probability density of two px , x i j random variables X and Y, which can be obtained by kernel � px , x log . i j density estimation or histogram method. *e entropy of X is p x px x ,x i j i j defined as follows: Among them, (x , x ) represents the mutual informa- 1 2 H(X) � p(x)log . (16) tion value between the first variable and the second variable, p(x) and the first row of the matrix represents the mutual in- formation value between the first variable and all variables. *e joint entropy of random variables X and Y is defined In order to obtain the variable grouping with the greatest as follows: correlation between variables, the first row is used to search for the starting point. First, we compare pairs in the first row H(X, Y) � p(x, y)log . (17) p(x, y) of the matrix, and remove the smaller value until only one x y element remains in the first row as the basis for subsequent Conditional entropy is used to describe the uncer- variable selection. *en, the variable represented by the tainty of inferring two variables when a random variable remaining elements in the first row is judged, and the is known. *e definition of conditional entropy is as variable selection in the next row is performed with the new follows: search starting point indicated by the variable. It iterates in this way until the selected variable coincides with variable 1 H(X|Y) � p(x, y)log . (18) in the first row. Finally, a variable grouping that searches for p(y|x) x y the start and end points with the first row can be obtained. As shown in Figure 5, we search for the starting point in According to the property of entropy, we obtain the the first row to obtain the maximum value (x , x ) of the 1 3 following equation: first row and then search for the maximum value of the third row to obtain (x , x ). *en, the maximum value (x , x ) of H(X, Y) � H(X) + H(Y|X) � H(Y) + H(X|Y). (19) 3 4 4 9 the 4th row is searched, and the maximum value (x , x ) of 9 1 the 9th row is finally terminated. In this way, the mutual *erefore, the mutual information value is as follows: 6 Journal of Robotics (X1, X2) (X1, X3) (X1, X4) (X1, X5) (X1, X6) (X1, X7) (X1, X8) (X1, X9) (X1, X10) (X2, X1) (X2, X3) (X2, X4) (X2, X5) (X2, X6) (X2, X7) (X2, X8) (X2, X9) (X2, X10) (X3, X1) (X3, X2) (X3, X4) (X3, X5) (X3, X6) (X3, X7) (X3, X8) (X3, X9) (X3, X10) (X4, X1) (X4, X2) (X4, X3) (X4, X5) (X4, X6) (X4, X7) (X4, X8) (X4, X9) (X4, X10) (X5, X1) (X5, X2) (X5, X3) (X5, X5) (X5, X6) (X5, X7) (X5, X8) (X5, X9) (X5, X10) (X6, X1) (X6, X2) (X6, X3) (X6, X5) (X6, X6) (X6, X7) (X6, X8) (X6, X9) (X6, X10) (X7, X1) (X7, X2) (X7, X3) (X7, X5) (X7, X6) (X7, X7) (X7, X8) (X7, X9) (X7, X10) (X8, X1) (X8, X2) (X8, X3) (X8, X5) (X8, X6) (X8, X7) (X8, X8) (X8, X9) (X8, X10) (X9, X1) (X9, X2) (X9, X3) (X9, X5) (X9, X6) (X9, X7) (X9, X8) (X9, X9) (X9, X10) (X10, X1) (X10, X2) (X10, X3) (X10, X5) (X10, X6) (X10, X7) (X10, X8) (X10, X9) (X10, X10) Figure 4: Schematic diagram of the mutual information matrix. (X1, X2) (X1, X3) (X1, X4) (X1, X5) (X1, X6) (X1, X7) (X1, X8) (X1, X9) (X1, X10) (X2, X1) (X2, X3) (X2, X4) (X2, X5) (X2, X6) (X2, X7) (X2, X8) (X2, X9) (X2, X10) (X3, X1) (X3, X2) (X3, X4) (X3, X5) (X3, X6) (X3, X7) (X3, X8) (X3, X9) (X3, X10) (X4, X1) (X4, X2) (X4, X3) (X4, X5) (X4, X6) (X4, X7) (X4, X8) (X4, X9) (X4, X10) (X5, X1) (X5, X2) (X5, X3) (X5, X5) (X5, X6) (X5, X7) (X5, X8) (X5, X9) (X5, X10) (X6, X1) (X6, X2) (X6, X3) (X6, X5) (X6, X6) (X6, X7) (X6, X8) (X6, X9) (X6, X10) (X7, X1) (X7, X2) (X7, X3) (X7, X5) (X7, X6) (X7, X7) (X7, X8) (X7, X9) (X7, X10) (X8, X1) (X8, X2) (X8, X3) (X8, X5) (X8, X6) (X8, X7) (X8, X8) (X8, X9) (X8, X10) (X9, X1) (X9, X2) (X9, X3) (X9, X5) (X9, X6) (X9, X7) (X9, X8) (X9, X9) (X9, X10) (X10, X1) (X10, X2) (X10, X3) (X10, X5) (X10, X6) (X10, X7) (X10, X8) (X10, X9) (X10, X10) (X1, X3) (X3, X4) (X1, X3, X4, X9) (X4, X9) (X9, X1) Figure 5: Schematic diagram of mutual information matrix search. Journal of Robotics 7 between each measurement variable in the augmented information value between the variables obtained by searching for the starting point of the first row is the largest, matrix and all variables in other matrices is calculated, and the calculation method is shown in formula (26): and the sum of the mutual information values of this group of variables is the largest. Cov(m, n) ������ � ������ R(m, n) � , (26) Var(m) Var(n) 2.4. Mining and Analysis of Dynamic Characteristics of J×1 Variables. We assume that x ∈ R are data consisting of J Among them, m and n are any two variables in different process variables. *e data array consisting of sampled augmented matrices, Cov (m, n) is the covariance between samples of x is as follows: the two variables, Var (m) and Var (n) are the variances of variables m and n, respectively. *e larger the absolute value T N×1 (22) X � [x(1), x(2), x(3), L, x(N)] ∈ R . of R, the stronger the correlation between the two variables. *e closer it is to 0, the weaker the correlation. Among them, N is the number of samples, and each row Finally, the correlation between all measured variables of X represents a sample. In order to consider the dynamic in each augmented matrix is K-means clustered into two characteristics between different sampling times x (k), X is categories, and the category with the smaller correlation first extended to as follows: value in the two categories is used as the basis for ∗ ∗ ∗ ∗ ∗ X � x(1) , x(2) , x(3) , L, x(N − d) eliminating variables. *e idea of K-means clustering is to (23) divide the data set through continuous iteration, so that � X , X , X , L, X . 0 −1 −2 −d different classes are independent and the same class is ∗ T T T compact. In this section, Euclidean distance is selected as Among them, there is X � [X , X , X , L, 0 −1 − 2 T the similarity measure between data samples, and the X ] . x corresponds to the variable obtained by shifting − d −i performance of the clustering algorithm is evaluated by the sampling of x at time i, which is as follows: using the squared error and objective function. *e brief X � x (1), x (2), x (3), L, x (N − d) steps of the K-means clustering method are as follows: (a) −i − i − i − i − i (24) it arbitrarily selects n samples from the input data samples � [x(i + 1), x(i + 2), x(i + 3), L, x(N − d + i)] . of the correlation between the measured variables as the initial clustering centers. (b) According to the mean of the *e matrix X is extended to obtain the augmented matrix ∗ (N− d)×J(d+1) clustered samples, the Euclidean distance between each X ∈ R as follows: sample, and the sample selected as the cluster center is T T T x (1) x (2) L x (d + 1) calculated, and the samples are divided according to the ⎡ ⎢ ⎤ ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ ⎥ minimum distance. (c) *e mean of the new cluster ⎢ T T T ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ x (2) x (3) L x (d + 2) ⎥ ⎢ ⎥ ∗ ⎢ ⎥ ⎢ ⎥ samples is recalculated. (d) b and c are iteratively looped ⎢ ⎥ X � ⎢ ⎥. (25) ⎢ ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ M M O M ⎥ ⎢ ⎥ until the cluster centers no longer change. From the re- ⎢ ⎥ ⎢ ⎥ ⎣ ⎦ T T T sults in Figure 7, it can be seen that the dimensions of the x (N − d) x (N − d + 1) L x (N) augmented matrix of different variables are different after screening, and the difference in the dynamics of different *e value of the lag parameter d in the general aug- mented matrix construction is 1 or 2, and for the system with variables leads to the difference in the actual lag order demand. strong dynamics, d can be selected by a specific iterative algorithm. If the value of the lag parameter d is small, the number of measurement values introduced into the data 2.5. Distributed Dynamic VGDTW-MPCA Model Modeling matrix may be insufficient, and the dynamic characteristics Strategy. After the corresponding augmented matrix is of the process data cannot be fully captured. If the value of constructedaccordingtothedynamicsofeachvariable,under the lag parameter d is large, it may lead to the introduction of the guidance of the block modeling idea, the augmented measurement value redundancy in the data matrix, which matrix of different variables can be divided into different may cause excessive interference to process modeling and blocks according to a certain relationship. *us, augmented monitoring. For this purpose, the lag parameter in the matrices of different chunks are different and augmented augmented matrix selected in this section is 4. In the data matrices of the same chunk are similar. *erefore, it is con- matrix, four measurements were introduced for every single sidered to use the mutual information between augmented variable to construct an augmented matrix for the single matrices of different variables to measure the intrinsic rela- variable. As shown in Figure 6, an augmented matrix is tionshipbetweensuchvariables.Amongthem,theprincipleof constructed by taking a batch of data as an example. Among mutual information has been introduced in the previous them, each column in the matrix represents the value of a chapter, and the schematic diagram of the grouping of dif- variable at all sampling times. ferent variable augmentation matrices is shown in Figure 8. *is approach fully captures the dynamic nature of all According to formula (21), the mutual information value variables. However, it inevitably creates redundancy in in- between the two variable augmented matrices is calculated dividual variables. Second, in order to accurately obtain the and the corresponding mutual information matrix is dynamic characteristics of different variables, it is necessary formed. *e formula for calculating the mutual information to filter variables for each augmented matrix. *e correlation 8 Journal of Robotics Variation X1 X2 X3 X4 X5 X6 X7 X8 X9 X10 Build an augmented matrix based on lagged order 4 ... Figure 6: Schematic diagram of constructing augmented matrix with different variables. ... ... Figure 7: Augmented matrix variable selection. value between a single measurement variable in the aug- mented matrix is as follows: a b p X , X m n a b a b IX , X � pX , X log . (27) ... m n m n a b p X pX a b X ,X m n m n Among them, X represents the a-th measurement variable of the augmented matrix m and X represents the b-th measurement variable of the augmented matrix n. *e formula for calculating mutual information between one augmented matrix and another augmented matrix is as follows: ... J J m n a b (28) I � IX , X . m m n a�1 b�1 Figure 8: Schematic diagram of augmented matrix grouping. Sampling time Journal of Robotics 9 Among them, J represents the number of all mea- k Time (k) surement variables in the augmented matrix m and J represents the number of all measurement variables in the augmented matrix n. 1 X We assume that the PCA model is obtained separately from the two grouped augmented matrix data as follows: X � T P + E . (29) i i i i I 1 Variation Among them, there is i �{1 and 2}. *e control limits T and Q are calculated at the same time. In the online i,lim i,lim k = K 2 1 monitoring phase, when new data are obtained, new T and k = m Q are calculated, respectively. In order to achieve a com- k = 2 k = 1 prehensive monitoring effect, this paper uses the Bayesian fusion technology to combine multiple statistics into a Figure 9: Schematic diagram of time slice cutting of the three- probability form. dimensional matrix. *e probability of misidentification of x monitored by the T statistic is defined as follows: 2 2 P (x|F)P (F) T T array used for cutting is in accordance with the pre- i i P 2 (F|x) � . (30) determined process and parameters under the normal batch P 2 (x) process operating conditions, and the process variables Moreover, the probability of P 2 (x) is defined as follows: follow the predetermined running track. However, an in- dustrial process cannot be completely repeated and the P 2 (x) � P 2 (x|N)P 2 (N) + P 2 (x|F)P 2 (F). (31) T T T T T i i i i i process variable trajectory must fluctuate under the influ- ence of random disturbances. *erefore, the J process Among them, x represents the data at a sampling time of variables of the two-dimensional time slice matrix X (I × J), the test data, N and F represent normal and abnormal after cutting the three-dimensional data under normal operating conditions, P 2 (N) and P 2 (F) can be simply T T i i operating conditions approximately obey the Gaussian designated as a and 1−a, and a is the confidence level for distribution. calculating the control limit. *e conditional probabilities Next, the PCA method is applied to the time slice matrix P 2 (x|N) and P 2 (x|F) are calculated by the following T T i i X (I × J) with a Gaussian distribution, which can extract formulas: the correlation information between the process variables at K sampling instants. Formula (35) is the PCA model of the ⎛ ⎝ ⎞ ⎠ P 2 (x|N) � exp − , (32) T time slice matrix: i,lim (35) X � T P + E , k k k i,lim (33) Among them, T represents the score matrix, P rep- P 2 (x|F) � exp− . T k k i resents the load matrix, and E represents the residual matrix. *e K load matrices represent the correlation in- *en, a weighted form is used to combine multiple formation between process variables and can reflect changes monitoring results into an overall probability indicator as in the internal operating mechanism of the process. follows: In order to capture the change of the load matrix, the principal component correlation degree is used as the sim- P 2 (x|F)P 2 (F|x) ⎧ ⎨ ⎫ ⎬ T T i i ilarity measure to measure the changing relationship between BIC � . 2 (34) T 2 ⎩ ⎭ P 2 (x|F) i�1 i�1 the two matrices and as the basis for the stage division. *e calculation method of the pivot degree is as follows: Similarly, the final probability indicator BIC of the Q T T T T statistic can be obtained. Usually, when the value of BIC or trace∧ U V∧ ∧ V ∧ v u Q u u (36) PCC(U, V) � . c u v BIC 2 exceeds the control limits a, an alarm is misidentified. λ λ i�1 i i Otherwise, the monitoring process is considered normal. Among them, U and V, respectively, represent the load matrix obtained by PCA in two different time slices, λ and 2.6. Description of Segmentation Strategy of the Ms-VGDTW- λ are their corresponding eigenvalues, c represents the u u u MPCA Algorithm. As shown in Figure 9, the datasheet number of load vectors, and there is ∧ � diag(λ , λ , L, λ ) 1 2 c v v v X(I × J × K) is obtained by vertically cutting the three-di- and ∧ � diag(λ , λ , L, λ ). v 1 2 c mensional matrix X (I × J) along the third dimension. *is A time series X � x(t ) is given. It is divided k i i�1 slice consists of the k-th sampling time of all batches of batch into z segments: X , j � 1,2, L, z, which satisfies X ≠∅, j j process data and is referred to by the researchers as the time X ∩ X � ∅,∀j ≠ j , j , j ∈ [1,2, L, z], ∩ X � X. *e j1 j2 1 2 1 2 j�1 j slice matrix in the batch process. *e three-dimensional data corresponding time interval is [t , t ], s , f ∈ {1,2, L, n}, s f j j j j 10 Journal of Robotics Preprocessing Assembly line data supply and control unit Endpoint detection OP-DTW Identification External Read in the Assembly line results display data and Feature extraction initialize the memory IP kernel OP-DTW Assembly line Eigenvectors of the reference mode and unknown modes Figure 10: System framework. continuous speech recognition system based on DTW proposed in this paper is shown in Figure 10. *e abovementioned framework constructs a Russian continuous speech recognition system based on DTW; then, the effect of the system is verified, the accuracy of speech recognition is counted, and the results shown in Figure 11 are obtained. It can be seen from the abovementioned research that the Russian continuous speech recognition system based on DTW proposed in this paper has a high continuous Russian speech recognition accuracy. Num Figure 11: *e accuracy of Russian recognition of the DTW-based 4. Conclusion Russian continuous speech recognition system. To improve the overall level of the teaching staff of public which represents the position of the start and end points of Russian courses in colleges and universities, it is necessary for the j-th line segment in the time series, and there is s � 1 the teachers of Russian courses in colleges and universities to and f � n. A linear function F (t ) � a t + b , ∀i ∈ [s , f ] z j i j i j j j carry out self-learning and self-improvement. *e school on the interval X is found to minimize the following ob- should help teachers of Russian courses to develop a sense of jective function: self-improvement, and at the same time update the educational concept in the course teaching, constantly enrich the personal k j J � x t − F t . (37) knowledge structure, and improve their professional knowl- i j i j�1 i�s edge level and education level. At the same time, it is necessary to continue learning as an important task, make use of modern *e objective function J uses the residual sum of squares information-based teaching technology, constantly supplement between the original time series and its linear approximation new knowledge and new content, and learn the teaching and to measure the degree of fit between the two. *e smaller J is, research methods of Russian teachers in other schools. *is the better the linear approximation fits the original time paper combines the DTW algorithm to construct a Russian ∗ ∗ continuous speech recognition system and improves the series. *e optimal solution (F , X ) obtained by j j j�1 Russian teaching effect through artificial intelligence methods. solving the above optimization problem is called an optimal It can be seen from the simulation results that the Russian linear approximation of the time series X. continuous speech recognition system based on DTW pro- posed in this paper has high continuous Russian speech rec- 3. Russian Continuous Speech ognition accuracy. Recognition System Data Availability *e most flexible hardware device in the embedded system is the FPGA (Field Programmable Logic Gate Array), which is *e labeled dataset used to support the findings of this study very suitable for implementing this algorithm. *e Russian is available from the corresponding author upon request. Accuracy of continuous Russian speech recognition (%) 97 Journal of Robotics 11 [15] S. N. Nurullayevna, “*e key of effective communication is Conflicts of Interest pronunciation,” European Journal of Humanities and Edu- cational Advancements, vol. 1, no. 4, pp. 5–7, 2020. *e authors declare that they have no conflicts of interest. [16] A. A. Kazantsev, P. Rutland, S. M. Medvedeva, and I. A. Safranchuk, “Russia’s policy in the “frozen conflicts” of Acknowledgments the post-Soviet space: from ethno-politics to geopolitics,” Caucasus Survey, vol. 8, no. 2, pp. 142–162, 2020. *is work was supported by Zhengzhou Railway Vocational [17] C. Portela, P. Pospieszna, J. Skrzypczynska, ´ and D. Walentek, and Technical College. “Consensus against all odds: explaining the persistence of EU sanctions on Russia,” Journal of European Integration, vol. 43, no. 6, pp. 683–699, 2021. References [1] D. Ivanko, A. Karpov, D. Fedotov et al., “Multimodal speech recognition: increasing accuracy using high speed video data,” Journal on Multimodal User Interfaces, vol. 12, no. 4, pp. 319–328, 2018. [2] P. S. Praveen Kumar, G. *immaraja Yadava, and H. S. Jayanna, “Continuous Kannada speech recognition system under degraded condition,” Circuits, Systems, and Signal Processing, vol. 39, no. 1, pp. 391–419, 2020. [3] A. A. Karpov and R. M. Yusupov, “Multimodal interfaces of human–computer interaction,” Herald of the Russian Acad- emy of Sciences, vol. 88, no. 1, pp. 67–74, 2018. [4] O. V. Verkholyak, H. Kaya, and A. A. Karpov, “Modeling short-term and long-term dependencies of the speech signal for paralinguistic emotion classification,” SPIIRAS Proceed- ings, vol. 18, no. 1, pp. 30–56, 2019. [5] V. Arutiunian and A. Lopukhina, “*e effects of phonological neighborhood density in childhood word production and recognition in Russian are opposite to English,” Journal of Child Language, vol. 47, no. 6, pp. 1244–1262, 2020. [6] E. Alsharhan and A. Ramsay, “Improved Arabic speech recognition system through the automatic generation of fine- grained phonetic transcriptions,” Information Processing & Management, vol. 56, no. 2, pp. 343–353, 2019. [7] A. Krickovic and I. Pellicciari, “From “Greater Europe” to “Greater Eurasia”: status concerns and the evolution of Russia’s approach to alignment and regional integration,” Journal of Eurasian Studies, vol. 12, no. 1, pp. 86–99, 2021. [8] L. Henry and E. Plantan, “Activism in exile: how Russian environmentalists maintain voice after exit,” Post-Soviet Af- fairs, vol. 38, no. 4, pp. 274–292, 2022. [9] R. Allison, “Russian revisionism, legal discourse and the ‘rules-based’international order,” Europe-Asia Studies, vol. 72, no. 6, pp. 976–995, 2020. [10] A. A. Sushentsov and W. C. Wohlforth, “*e tragedy of US–Russian relations: NATO centrality and the revisionists’ spiral,” International Politics, vol. 57, no. 3, pp. 427–450, 2020. [11] J. Eriksson and R. Privalov, “Russian space policy and identity: visionary or reactionary?” Journal of International Relations and Development, vol. 24, no. 2, pp. 381–407, 2021. [12] L. Li, X. Fu, S. Chen et al., “Hydrophobic and stable MXe- ne–polymer pressure sensors for wearable electronics,” ACS Applied Materials & Interfaces, vol. 12, no. 13, pp. 15362– 15369, 2020. [13] A. Kashevnik, I. Lashkov, and A. Gurtov, “Methodology and mobile application for driver behavior analysis and accident prevention,” IEEE Transactions on Intelligent Transportation Systems, vol. 21, no. 6, pp. 2427–2436, 2020. [14] A. Moe, “A new Russian policy for the Northern sea route? State interests, key stakeholders and economic opportunities in changing times,” (e Polar Journal, vol. 10, no. 2, pp. 209–227, 2020.
http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png
Journal of Robotics
Hindawi Publishing Corporation
http://www.deepdyve.com/lp/hindawi-publishing-corporation/a-russian-continuous-speech-recognition-system-based-on-the-dtw-8jq0PRcxJo
