2.1 Quantitative running training data
Mentioned that there are three elements of scientific training, namely: quantification, personalization and periodicity. Among them, quantification is the most important, because if you cannot quantify the results of your training and competition, the other two cannot be implemented in detail. Quantification is the original abstract training results and subjective feelings are digitized, as if we can only feel thin or fat when it visually observe a person’s posture, without telling a standard or actual value[5][6]. For example, the BMI value can be calculated based on the height and weight, or use body weight to measure standard data such as the correct weight in kilograms, body fat and muscle mass ratio. With the rapid advancement of science and technology, in addition to all kinds of wearable technology or smart phones on the market, you can collect and record your own running training data [11]. There is also a cloud platform for analysis and calculation that can be used as a reference basis after training. This research is used the training data of the running watch as reference data for forecast [7][8].
2.1.1 Heart rate
Heart rate refers to the frequency of heart contraction and beats and the number of beats per minute (bpm). The heart rate of a normal person is 60 to 100 beats per minute when it is at rest. The heartbeat will start to accelerate during exercise, but the heart and lungs athletes with better functions will have a slower heart rate than normal people. The heart rate is controlled by the autonomic nerve [3]. When the sympathetic nerve activity increases, the heart rate will increase; when the vagus nerve activity increases, the heart rate will also slow down. The main fluid factors that affect the heart rate are adrenaline and norepinephrine in the circulating blood, and thyroxine. In addition, the heart rate is also affected by body temperature. If the body temperature rises by 1℃, the heart rate will increase by 12 to 18 times. The heart rate provides a lot of information [3] [9][10][12]. With correct information, you can evaluate your body's responsiveness, adaptability, energy consumption, and the suitability of your training plan. Using the heart rate monitor, plus the correct record data, you may get the following information:
-
Exercise suitable intensity when establishing aerobic and anaerobic systems
-
The correct length of time to stay in the appropriate training interval
-
Interval training and appropriate recovery time between the two training events
-
Effective evaluation of training courses
-
Anticipate the occurrence of overtraining, heatstroke and physical exhaustion in advance
-
Long-distance race pace strategy setting
When talking about how to use the heart rate during running to determine the intensity of running training, it must be understood that the intensity of running does not always directly correspond to the pace. Factors such as weather and terrain will affect the heart rate. When using the heart rate to monitor the training intensity, the most It is important to know what your maximum heart rate is. The maximum heart rate can be calculated using formulas to calculate the predicted value, or you can wear a heartbeat watch to increase the number of runs and speed measurements on the uphill section and in the playground[13][14].
Some coaches and athletes discuss how to effectively use heart rate in training. Nowadays, heart rate-based training is so common that athletes are too confident that the rate is a decisive factor reflecting their training and competition conditions. From another perspective, good use of a heart rate monitor can improve physical fitness and competition performance. When morale is low, over-excited, competing, inattentive and improper judgments, it will hinder wise training. At this time, the heart rate monitor is like a coach will accompany you to practice together. Understand how the heart rate works and use it with other tools for assessing intensity, plus some common sense gained from experience, the heart rate monitor can help determine whether such training is too much or not enough, whether recovery is complete, and physical progress. Increasing age is beyond the physiological irreversible factors that make the maximum heart rate slowly decrease [15] [16]. There are also situations where the maximum heart rate is getting higher and higher. In addition to the influence of age, each person’s maximum heart rate may also change after training. It is not necessary to measure the maximum heart rate, but it can also be seen from the maximum heart rate data that your physical condition is a good or bad reference benchmark. After all, measuring the maximum heart rate can cause injuries in addition to the original exercise posture. For long-distance aerobic endurance sports events such as marathon, middle and long distance running, triathlon, swimming, cycling, heart rate has become an indispensable and important, and one of the basic monitoring data for these sports. In sports, it is most common to measure the resting heart rate and maximum heart rate, and use the method of monitoring heart rate to increase the intensity of training. It is very dangerous for athletes to break through the limits of the body during training or competitions, so using this data, the trainer or athlete can clearly understand the state of their body. However, heart rate monitoring has become one of the most important data indispensable in wearable technology [17] [18].
2.1.2 Cadence and stride length
Cadence is the frequency of steps, the number of times the legs alternate per unit time during running, and the pace per minute (ppm) of the soles of the feet (step per minutes, spm). Simply put, the faster the cadence, the faster the movement. However, the higher the efficiency in mechanics, the more the load on physical fitness. Also, because everyone’s physical condition is different, each runner needs to find a pace that suits him at different distances, and basically wants to run. Accelerating speed, increasing stride frequency or increasing stride length (running speed = stride frequency × stride length) is one of the important factors that determine running speed. "Moving your feet quickly" sounds like a good way to increase speed, but it is actually a skill that must improve aerobic capacity while allowing neuromuscular training. This kind of double training, for more important than modifying the running posture [30] [32]. The average leisure runner's stride frequency is between 150–170 steps per minute, and the number evolution may depend on factors such as leisure running, racing or fitness. Many well-trained runners, regardless of the heel or forefoot landing, too large or moderate span, each runner may have a different running style, but most of the steps can be maintained at more than 180 steps. Stride length is the distance of one step, calculated at the center of the foot. After taking a step, the distance between the centers of the feet is the stride. However, the stride length in running will vary according to different running styles, which is generally related to height [19] [20]. If it is a short-medium run, the stride length is about 80% of your height. For example, if your height is 170, then the short and medium running width is about 136 cm (the short and medium running generally refers to 3000 meters or less). Because the marathon is a long-distance running (mostly 10KM/21KM/42KM), the stride is between 55% and 70% of the height. Speed = cadence × stride length. This formula is quite easy to understand. When you want to increase your running speed, you can increase the frequency that you can step out per minute. The so-called "step frequency 180" means that you can take 180 steps per minute. And stride length is the distance you can cross each step, the product of the two is equal to speed [6][7].
The relationship between stride frequency and stride length can be described as a complementary relationship. In a 100-meter sprint, stride length is more important than stride frequency, because short distances require more muscle power and do not need to adjust physical fitness [37–39]. On the contrary, marathon steps are used. The importance of frequency is much greater than stride length, and the adjustment of rhythm and physical fitness is very important. An increase in stride means a reduction in possible stride frequency, and an increase in stride frequency may also mean a shortened stride length [33] [35]. The point is not to increase the efficiency of the stride frequency or stride length, but the degree of attenuation of the other after the efficiency is improved, and the degree of durability. The reason why stride frequency is more efficient than stride length is that the increase in stride length makes it difficult to extend the joints, while stride frequency is the room for improvement through muscle and nerve training. Speaking of triathletes, try to increase the stride length and reduce the stride frequency to speed up the running speed. In order to increase the stride length, the center of gravity must be increased by tens of centimeters every time you step. Such a long-stride running method will have some undesirable effects. In addition to consuming extra unnecessary energy, the runner will also reduce the running speed due to the influence of gravity [34] [36]. The last effect is the long-stride running method. When the footsteps fall back to the ground, a great impact will be generated, and such impact will accumulate day after day and year after year of running, causing one of the chronic sports injury factors for runners. It is also mentioned that whether it is a general amateur runner or a long-distance runner, when participating in a long-distance running competition, the faster the running speed, the increase rate of running stride will be significantly higher than the increase rate of stride frequency. The rate is five times the rate of cadence increase. However, running stride length is proportional to long-distance running performance and optimal running speed. Increasing stride length to increase running speed in long-distance competitions seems to be more important than increasing stride frequency [8].
2.1.3 Pace
Pace is one of the key factors in the regulation of physical fitness and running technology in road running. It is used in long-term endurance sports such as 10 km, half marathon or full marathon, road racing, and even triathlon. In the event, the pace is appropriate. In addition to completing the event, you can even get better results after training. In the spring, the coach will adjust the training to a week on the track and a week to practice the fartlek run; in summer [29] [31]. All speed schedules are changed to train on the track. Such training can force the runners to focus on their actual race pace, speed up or decelerate according to the situation, and the turning points and lines on the track can help you increase the competition [5]. If the physical fitness is strong, even if the pace is wrong, you can withstand a certain degree of error performance in the game. If you add various factors such as fierce confrontation in the competition, extremely challenging track, and difficult external environment. It may be impossible to overcome the mistakes caused by the wrong pace. So pace is the key skill necessary to win, if and when you set high goals for yourself. The mentioned heart rate monitoring is actually a bit similar to monitoring pace. According to different running distances and training intensity, both are important data for observing physical fitness and adjusting. The slower Easy Pace (EP) and Threshold Pace (TP), Interval Pace (IP), Repeat Pace (RP), four marathon pace training [10].
2.1.4 Calories
For a person who is losing weight or losing weight, how many calories you burn and reduce your intake of calories. Calories is a data that is easy to understand. More and more wearable technology is based on the time distance of exercise or heart rate. Multi-advanced data to calculate how the calories burned after exercise trains the body's ability to burn fat is a very important thing for long-distance runners, and it is even more important for people who want to use running to lose weight [25][27]. The unit of calories is calories. Fat burn refers to the grams of fat consumed by runners. The meaning of fat burn rate is the percentage of energy derived from fat. The meanings of these three are very different. Although high-intensity exercise consumes more calories per unit time, the fat-burning rate is far less than that of long-distance jogging. In long-distance sports events such as running and cycling, you can often see the setting of food replenishment stations [26][28]. This kind of long-term aerobic endurance exercise means that a lot of energy will be consumed, and the water and fat in the body Carbohydrates (sugars) need to be supplemented in a timely manner to maintain the continuous energy supplement of the game [11].
2.2 Recurrent Neural Network
RNN also be called recursive neural network. It is the most commonly used neural network model in the field of natural language processing. It is a neural network that processes sequence data such as sound, language, and video. The network is a kind of neural network with short-term memory ability [15] [16]. Because the front input value and the back input value of RNN are related, it is most suitable for practical applications such as language translation, sentiment analysis, and weather forecasting. One of the characteristics of RNN is that the output of each layer in the multi-layer neural network is directly added to the layer itself. The input is derived from the loop (Self-Loop). With this network memory structure, it is possible to think about the previous input of the desired data [18] [19]. The meaning of the data, the following Fig. 1 is the basic structure of RNN. In Fig. 1, x is the input vector, y is the output vector, and h is the state vector that preserves the internal RNN [13–15]. Figure 2 is the expanded architecture flow of the RNN, t is the time step parameter, t-1 is the previous step, and t + 1 is the next step.
Figure 3 is a simple RNN calculation model method. The input vector 𝑥𝑡 at time point step t is connected with the internal state vector ℎ𝑡−1 before the update to become a longer vector in the form of [𝑥𝑡, ℎ𝑡−1], which is taken as the input excitation function 𝜎, its output is the internal state vector ℎ𝑡.
ℎ𝑡 = 𝜎(𝑊ℎ × [ℎ𝑡−1,𝑥𝑡]) (1)
The new internal vector ℎ𝑡, when passed through the internal loop in the next step, becomes the input vector of the fully connected layer of the second layer. In the fully connected layer of the second layer, use the excitation function 𝜎 to get the output vector 𝑦𝑡 at the time point of step t.
𝑦𝑡 = 𝜎(𝑊ℎ × ℎ𝑡) (2)
The values of ℎ𝑡 are arranged horizontally, indicating that the previous 𝑥𝑡 information has been consolidated. RNN uses this method to memorize past information.
Figure 3 RNN internal calculation structure
2.3 Long Short-Term Memory (LSTM)
LSTM is improved from RNN. This is a neural network with long-term memory ability. In order to solve the problem of disappearing gradient of recurrent neural network, it was developed a kind of cyclic neural network structure, because RNN will produce gradient disappearance in structure, causing long-term memory to be hidden by short-term memory. The most important feature of LSTM is that it can transfer the internal state vector stably for a long time. In order to solve the problem of RNN, LSTM adopts an improved memory management architecture [12] [13] [16][18]. The internal state vector ℎ uses a simpler calculation method to transfer. Even after a long time step, the transfer vector can be stabilized ℎ. But the LSTM operation situation mentioned above as shown in Fig. 4.
From Fig. 4, we can clearly see that the LSTM activation function has become more, and the element-wise multiplication and element-wise addition are also used. There are three more gates in it, which is LSTM, it is an important factor for better memory function, and these three gates are forget gate (𝑓𝑡) input gate (𝑖𝑡) output gate (𝑜𝑡), the following is the calculation formula of LSTM three gates.
LSTM is a technology to update the internal state vector ℎ𝑡 in a long time step. The choice of forgetting or additional information is determined by the forgetting gate 𝑓𝑡 and the input gate 𝑖𝑡. Forgot to read 𝑓𝑡 decides to memorize part of the message, and the input gate 𝑖𝑡 adds the repeated message here. Then the LSTM neural network can temporarily retain part of the information [17].
2.4 Gated Recurrent Unit (GRU)
LSTM also has a brother neural network called gate recurrent unit (GRU) neural network, GRU, which is an updated version of LSTM, a simpler version than LSTM, which can provide faster execution speed and reduce memory usage. In GRU, no explicit internal state transition is required. It is because the internal state is combined with the output vector 𝑦𝑡. In step 𝑡, the GRU neural network calculates the binary vector called Update Gate 𝑧𝑡, and the binary vector called Reset Gate 𝑟𝑡 These gates are in Update Gate Layer and Reset Gate Layer are calculated [12].
From Fig. 5, it use of element product, element sum, and activation function, except that the activation function 𝜎 and the internal state vector ℎ𝑡 are missing. The following are the calculation formulas for Update Gate, 𝑧𝑡 and Reset Gate 𝑟𝑡.
𝑧𝑡 = 𝜎(𝑊𝑧 × [𝑦𝑡−1,𝑥𝑡]) (6)
𝑟𝑡 = 𝜎(𝑊𝑟 × [𝑦𝑡−1,𝑥𝑡]) (7)
Compared with the forget gate and input gate of LSTM, the update gate action of GRU is simpler. Consider that for one value output by the GRU, remember the original value or replace it with a new value. Since the update gate 𝑧𝑡 does not update or touch a part of the memory, it can be memorized within a long step like LSTM, so the memory mechanism of GRU is more efficient than RNN.