CN107506041A - A kind of wearable mouse control method based on motion sensor - Google Patents

Abstract

The invention discloses a wearable mouse control method based on a motion sensor, through eight steps of data collection, action definition, data preprocessing, feature extraction, action classification, action segmentation, data and training error correction, etc. The head motion motion data transmitted by the sensor in the wearable device, while ensuring that it does not affect normal vision and easy operation, realizes high-precision control of the mouse using head motion. Compared with the eye-controlled, foot-controlled and exhalation-controlled mice that have appeared on the market, the present invention is easier to operate and more accurate to recognize. In addition, the invention has wide application, not only can be used conveniently by armless people, but also can help some normal people to free their hands. The invention has universal applicability and can be used not only to control computers, but also to control any other electronic equipment.

Description

A wearable mouse control method based on motion sensor

technical field

The invention belongs to the field of mobile perception computing, and in particular relates to a wearable mouse control method based on a motion sensor. Compared with the eye-controlled, foot-controlled and exhalation-controlled mice that have appeared on the market, the present invention is easier to operate and more accurate to recognize. In addition, the invention has wide application, not only can be used conveniently by armless people, but also can help some normal people to free their hands. The invention has universal applicability and can be used not only to control computers, but also to control any other electronic equipment.

Background technique

With the development of Internet technology, computers, smart phones and other devices are becoming more and more popular in people's lives, and the mouse has become the most common human-computer interaction system. However, armless people and some arm disabled people cannot use a mouse or a touch screen to manipulate devices in their lives. At the same time, wearable devices, such as smart watches and smart bracelets, are becoming more and more popular today. Sensors are also increasingly diverse and highly available: devices such as accelerometers and gyroscopes can provide precise information about the movement of a device. Some wearable devices have appeared on the market, such as eye-controlled devices that control the mouse through the movement of the eyeballs, foot-controlled devices that use the toes instead of fingers to control the mouse, and the length and strength of inhalation and exhalation. different combinations to realize the exhalation control class of mouse functions such as up, down, left, and right, and clicking, but they all have the disadvantages of inconvenient operation and difficult control. The use of head movements for control can achieve easy operation and accurate recognition, which can solve the above problems.

The present invention is committed to researching and exploring a wearable mouse control method based on a motion sensor by using the head movement data transmitted from the sensor in the wearable device.

The accuracy of action recognition is related to the sampling rate of sensor data. If the sampling rate is too high, the transmission cost will be too high, and it will easily cause congestion, causing considerable delay in data transmission and easy packet loss. If the sampling rate is too low, it will result in indistinct motion features and imprecise recognition results. Therefore, in order to ensure normal data transmission and accurate enough action recognition, it is necessary to select its sampling rate.

The transmission of sensor data is carried out via bluetooth or wifi. During use, unnatural shaking of the head and unsatisfactory network conditions may occur. Therefore, the way of data processing and transmission needs to be carefully designed. In practical applications, factors such as noise wave interference, lack of data transmission, and the trade-off between the imperceptibility of additional signals and the robustness of information transmission need to be considered.

Contents of the invention

In order to solve the above technical problems, the present invention provides a wearable mouse control method based on a motion sensor.

The technical scheme adopted in the present invention is:

A kind of wearable mouse control method based on motion sensor, is characterized in that, comprises the following steps:

Step 1: Conduct data collection, including:

Step 1.1: Select a Bluetooth or Wifi hotspot to connect the embedded wearable device and the controlled computer, and establish a Socket connection to transfer data. The port is set to 8088.

Step 1.2: Store the motion data in three directions obtained from the accelerometer and gyroscope in sequence. When the number of received data is greater than the window length, the first data will be removed. The sampling rate is set to 40 times per second.

Step 2: Action definition, specifically including the following sub-steps:

Step 2.1: Define that the user is facing the screen at the beginning, and the initial head position becomes the middle area.

Step 2.2: Define the click action. The action of snapping the head down and then snapping back to the middle area is defined as a click operation.

Step 2.3: Define the direction action. Turning the head from the middle area to the left, keeping it for a while until the cursor reaches the desired position, and then turning back to the middle area is defined as a left movement. Turning the head from the middle area to the right, keeping it for a while until the cursor reaches the desired position, and then turning back to the middle area is defined as a right movement. Turning the head upward from the middle area, keeping it for a while until the cursor reaches the desired position, and then turning back to the middle area is defined as an up movement.

Turning the head down from the middle area, keeping it for a while until the cursor reaches the desired position, and then turning back to the middle area is defined as a down movement.

Step 3: data preprocessing for eliminating noise and removing tooth waves;

Step 4: feature extraction, specifically including the following sub-steps:

Step 4.1: Because in the x-axis direction of the accelerometer, the click action is more prominent than the other four directions, so the value of the x-axis direction of the accelerometer can be used to distinguish the click action from the direction action.

Step 4.2: Because in the x-axis direction of the gyroscope, the left-right movement is more characteristic than the up-down movement, so the value in the x-axis direction of the gyroscope can be used to distinguish the left-right movement from the up-down movement. And because the characteristics of the left and right movements are just opposite, it is also possible to distinguish the left and right movements.

Step 4.3: Because in the z-axis direction of the gyroscope, the up and down motion is more significant than the left and right motion, so the value in the z-axis direction of the gyroscope can be used to distinguish the up and down motion from the left and right motion. And because the characteristics of the left and right movements are just opposite, it is also possible to distinguish the up and down movements.

Step 5: action classification, the specific implementation includes the following sub-steps:

Step 5.1: define a _tn ^x , g _tn ^x , g _tx ^x , g _tn ^z , and g _tx ^z as the lower threshold of the accelerometer in the x-axis direction and the upper and lower thresholds of the gyroscope in the x-axis and z-axis directions, respectively. The value of the threshold is determined by the pre-training step as the decision data of the decision tree. Define a _i ^x , g _i ^x , and g _i ^z as the values of the accelerometer x-axis direction, gyroscope x-axis direction, and gyroscope z-axis direction at the current time i.

Check the value of the x-axis direction of the accelerometer at the current time, if a _i ^x <a _tn ^x , the head starts to click, otherwise, go to step 5.2.

Step 5.2: Check the value of the x-axis direction of the gyroscope at the current time. If g _i ^x <g _tn ^x or g _i ^x >g _tx ^x , the head will start to move left or right, and go to step 5.3. Otherwise, go to step 5.4.

Step 5.3: If g _i ^x <g _tn ^x , the head starts to move to the left and the cursor on the screen moves to the left. Otherwise, the head moves to the right, and the cursor on the screen moves to the right.

Step 5.4: Check the value of the z-axis direction of the gyroscope at the current time. If g _i ^z <g _tn ^z or g _i ^z >g _tx ^z , the head will start to move up or down, and go to step 5.5. Otherwise, no action occurs at the current time.

Step 5.5: If g _i ^z <g _tn ^z , the head starts to move up and the cursor on the screen moves up. Otherwise, the head starts moving down and the on-screen cursor moves to the right.

Step 5.6: Once the action is recognized, the cursor will execute the corresponding action on the screen until the end of the action,

Step 6: Action segmentation. The obtained original data is time-continuous data. Action segmentation divides these time-continuous data into blocks, and the data contained in each block represents an action. According to the definition of action, each action contains two opposite parts. For example, a leftward movement first turns the head from the middle area to the left, and then from the left back to the middle area. The result is that there are two opposite waveforms in the same direction. Define two thresholds T _n and T _p in one direction to determine the start and end of the action. Wherein, T _n is the lower threshold and T _p is the upper threshold.

Step 7: Data pre-training, the specific implementation includes the following sub-steps:

Step 7.1: The user does 5 defined actions sequentially 5 times before use.

Step 7.2: Extract 5 pairs of extreme values of the obtained data, and multiply the obtained lowest peak and highest trough by a certain coefficient k as the trained personalized threshold.

Step 8: Error correction, when the user uses it, the user can see the real-time movement of the cursor on the screen. When an error occurs, the user can know that an error has occurred and can immediately start correcting it. If the cursor moves in the wrong direction, the user can perform the same action and then make a contact. For example, when the user moves to the left and finds that the cursor moves to the right, the user can also perform a right action to correct this error. In this way, the previous action is finished, and the subsequent action will not be affected by the previous error.

In the above-mentioned wearable mouse control method based on the motion sensor, the specific realization of step 3 includes the following sub-steps:

Step 3.1: Use threshold denoising to remove noise. Define a noise threshold E _th , and only transmit data to the device when the received data is greater than the threshold, otherwise it is set to 0.

Step 3.2: Use the maximum value filter to remove the tooth wave. Maximum value filtering is to use the maximum value of all values in the window as the current filtered value. Define the filtering window as τ, then the filtered value r _t at the current time can be expressed as:

r _t =max{r _t-τ/2 ,...,r _t-1 ,r _t ,r _t+1 ,...,r _t+τ/2 }

In the above-mentioned wearable mouse control method based on the motion sensor, the termination of the action in step 5.6 is defined as follows:

For the left action, when the real-time data g _i ^x ≥ g _tx ^x , the cursor stops moving, and then the real-time data g _i ^x ≤ g _tn ^x is recognized, and the left action is terminated.

For the right action, when the real-time data a _i ^x ≤ a _tn ^x , the cursor stops moving, and then the real-time data a _i ^x ≥ a _tn ^x is recognized, and the right action is terminated.

For the up action, when the real-time data g _i ^z ≥ g _tx ^x , the cursor stops moving, and then the real-time data g _i ^z ≤ g _tx ^z is recognized, and the up action is terminated.

For the down action, when the real-time data g _i ^z ≤ g _tn ^x , the cursor stops moving, and then the real-time data g _i ^z ≥ g _tn ^z is recognized, and the down action is terminated.

For the click action, its action is in the same direction as the down action. When the real-time data g _i ^z ≤ g _tn ^x , the cursor stops moving, and then the real-time data g _i ^z ≥ g _tn ^z is recognized, and the click action is terminated.

In the aforementioned motion sensor-based wearable mouse control method, in step 6, the following rules are used to segment the data:

It is defined that the head is facing the screen in the middle area initially, when the gyroscope data is recognized and reaches a certain threshold for the first time, this point is marked as the point where the action starts, denoted as P _s . When the gyroscope data reaches a certain threshold for the second time in the opposite direction, this point is marked as the point where the action ends, which is denoted as P _e . Then the section between P _s and P _e is divided as an action.

In the above-mentioned wearable mouse control method based on the motion sensor, the specific realization of step 7 includes the following sub-steps:

Step 7.1: The user does 5 defined actions sequentially 5 times before use.

Step 7.2: Extract 5 pairs of extreme values of the obtained data, and multiply the obtained lowest peak and highest trough by a certain coefficient k as the trained personalized threshold.

Compared with the prior art, the invention has the advantages of more precise control, simple operation, no influence on normal things, scalability and universal applicability.

Description of drawings

Fig. 1 is a system frame diagram of the present invention.

Figure 2a is an example diagram of the extracted waveform signal.

Figure 2b is a graph of the extracted waveform signal after preprocessing.

Figure 3a shows the accuracy rate of user test results under the universal threshold.

Figure 3b shows the recall rate of user test results under the universal threshold.

Figure 4a is the variation curve of each action recognition accuracy rate with k value.

Figure 4b is the curve of the recall rate of each action recognition with the value of k.

Figure 5 shows the judgment accuracy of the personalized threshold.

detailed description

In order to facilitate those of ordinary skill in the art to understand and implement the present invention, the present invention will be described in further detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the implementation examples described here are only used to illustrate and explain the present invention, and are not intended to limit this invention.

The present invention is mainly based on wearable equipment and sensor technology, considering the motion characteristics of head movements, and proposes a wearable mouse control method based on motion sensors. The method makes full use of the movement data of the head movement, and realizes easy-to-operate and high-precision mouse control without affecting normal vision. The present invention can be used for armless persons, so that armless or arm disabled persons can also use the mouse, and at the same time, normal people can also use the present invention to free their hands.

The method provided by the invention can use computer software technology to realize the process. Referring to Fig. 1, a kind of wearable mouse control method based on motion sensor provided by the present invention comprises the following steps:

Step 1: Collect action data from the sensor. The specific implementation process is:

Step 1.1: Select Bluetooth or Wifi hotspot to connect the embedded wearable device and the controlled computer, and establish a Socket connection to transfer data. The port is set to 8088.

Step 1.2: Store the motion data in three directions obtained from the accelerometer and gyroscope in sequence. When the number of received data is greater than the window length, the first data will be removed. The sampling rate is set to 40 times per second.

The specific implementation process of the embodiment is described as follows:

First, connect the wearable device and the control device (such as a computer) through Bluetooth or Wifi hotspot, establish a Socket connection and transmit sensor data, and set the port to 8088.

Then, the motion data in three directions obtained from the accelerometer and the gyroscope are respectively stored sequentially. When the number of received data is greater than the number of defined windows, the first data will be removed, and then the remaining data will be placed in the window and handed over to the computer program for processing. The sampling rate is set to 40 times per second. When the sampling rate is too large, the cost is too large and may cause congestion. When the sampling rate is too small, the motion features will not be obvious enough.

Step 2: Define the action. For each operation of the mouse, a corresponding head movement is defined.

Step 2.1: Assuming that the user is facing the screen at the beginning, the initial head position is defined as the middle area.

Step 2.2: Define the click action. The action of snapping the head down and then snapping back to the middle area is defined as a click operation.

Step 2.3: Define the direction action. Turning the head from the middle area to the left, keeping it for a while until the cursor reaches the desired position, and then turning back to the middle area is defined as a left movement. Turning the head from the middle area to the right, keeping it for a while until the cursor reaches the desired position, and then turning back to the middle area is defined as a right movement. Turning the head upward from the middle area, keeping it for a while until the cursor reaches the desired position, and then turning back to the middle area is defined as an up movement. Turning the head down from the middle area, keeping it for a while until the cursor reaches the desired position, and then turning back to the middle area is defined as a down movement.

The specific implementation process is:

Take left action as an example. A left action is defined as moving the cursor on the screen to the left. Once the head moves to the left from the middle area, the cursor starts moving to the left as well. The head will keep a certain distance in the left area, and the cursor will keep moving to the left. Once the head is swung back to the middle area, the cursor stops moving.

Step 3: Preprocess the data. The denoising operation is performed on the original data without affecting the extraction of motion features.

Step 3.1: Use threshold denoising to remove noise. Define a noise threshold E _th , and only transmit data to the device when the received data is greater than the threshold, otherwise it is set to 0.

Step 3.2: Use the maximum value filter to remove the tooth wave. Maximum value filtering is to use the maximum value of all values in the window as the current filtered value. Define the filtering window as τ, then the filtered value r _t at the current time can be expressed as:

r _t =max{r _t-τ/2 ,...,r _t-1 ,r _t ,r _t+1 ,...,r _t+τ/2 }

The specific implementation process is:

Define a noise threshold E _th , the value in the original data that is smaller than the threshold will be set to 0, and only when the received data is greater than the threshold will the data be transmitted to the device. The data after the threshold denoising is used to remove the tooth wave by the method of maximum value filtering. Maximum value filtering considers the left and right values of the current position together, and takes the maximum value as the filtered value of the position.

Step 4: Feature extraction.

The specific implementation process is:

Because in the x-axis direction of the accelerometer, the click operation is more significant than the other four directions, and there is no obvious difference in the four actions of the accelerometer, so you can use the value of the x-axis direction of the accelerometer to Distinguish between click actions and direction actions.

Because in the x-axis direction of the gyroscope, the left and right movements are more significant than the up and down movements, so the value in the x-axis direction of the gyroscope can be used to distinguish the left and right movements from the up and down movements. And because the characteristics of the left and right movements are just opposite, it is also possible to distinguish the left and right movements.

Because in the z-axis direction of the gyroscope, the up and down motion is more significant than the left and right motion, so the value in the z-axis direction of the gyroscope can be used to distinguish the up and down motion from the left and right motion. And because the characteristics of the left and right movements are just opposite, it is also possible to distinguish the up and down movements.

Step 5: Action Classification.

The specific implementation process is:

Step 5.1: define a _tn ^x , g _tn ^x , g _tx ^x , g _tn ^z , and g _tx ^z as the lower threshold of the accelerometer in the x-axis direction and the upper and lower thresholds of the gyroscope in the x-axis and z-axis directions, respectively. The value of the threshold is determined by the pre-training step as the decision data of the decision tree. Define a _i ^x , g _i ^x , and g _i ^z as the values of the accelerometer x-axis direction, gyroscope x-axis direction, and gyroscope z-axis direction at the current time i.

Check the value of the x-axis direction of the accelerometer at the current time, if a _i ^x <a _tn ^x , the head starts to click, otherwise, go to step 5.2.

Step 5.2: Check the value of the x-axis direction of the gyroscope at the current time. If g _i ^x <g _tn ^x or g _i ^x >g _tx ^x , the head will start to move left or right, and go to step 5.3. Otherwise, go to step 5.4.

Step 5.3: If g _i ^x <g _tn ^x , the head starts to move to the left and the cursor on the screen moves to the left. Otherwise, the head moves to the right, and the cursor on the screen moves to the right.

Step 5.4: Check the value of the z-axis direction of the gyroscope at the current time. If g _i ^z <g _tn ^z or g _i ^z >g _tx ^z , the head will start to move up or down, and go to step 5.5. Otherwise, no action occurs at the current time.

Step 5.5: If g _i ^z <g _tn ^z , the head starts to move up and the cursor on the screen moves up. Otherwise, the head starts moving down and the on-screen cursor moves to the right.

Step 5.6: Once the action is recognized, the cursor will execute the corresponding action on the screen until the end of the action. The end of the action is defined as follows:

For the left action, when the real-time data g _i ^x ≥ g _tx ^x , the cursor stops moving, and then the real-time data g _i ^x ≤ g _tn ^x is recognized, and the left action is terminated.

For the right action, when the real-time data a _i ^x ≤ a _tn ^x , the cursor stops moving, and then the real-time data a _i ^x ≥ a _tn ^x is recognized, and the right action is terminated.

For the up action, when the real-time data g _i ^z ≥ g _tx ^x , the cursor stops moving, and then the real-time data g _i ^z ≤ g _tx ^z is recognized, and the up action is terminated.

For the down action, when the real-time data g _i ^z ≤ g _tn ^x , the cursor stops moving, and then the real-time data g _i ^z ≥ g _tn ^z is recognized, and the down action is terminated.

For the click action, its action is in the same direction as the down action. When the real-time data g _i ^z ≤ g _tn ^x , the cursor stops moving, and then the real-time data g _i ^z ≥ g _tn ^z is recognized, and the click action is terminated.

Step 6: Action Segmentation

The specific implementation process is:

The raw data we get is time-continuous data, and action segmentation divides these time-continuous data into blocks, and each block contains data representing an action. According to the definition of action, each action contains two opposite parts. For example, a leftward movement first turns the head from the middle area to the left, and then from the left back to the middle area. The result is that there are two opposite waveforms in the same direction. Define two thresholds T _n and T _p in one direction to determine the start and end of the action. Wherein, T _n is the lower threshold and T _p is the upper threshold. Use the following rules to split the data:

Assuming that the head is facing the screen in the middle area initially, when the gyroscope data is recognized and reaches a certain threshold for the first time, this point is marked as the point at which the action starts, denoted as P _s . When the gyroscope data reaches a certain threshold for the second time in the opposite direction, this point is marked as the point where the action ends, which is denoted as P _e . Then the section between P _s and P _e is divided as an action.

Step 7: Data pre-training

The specific implementation process is:

Due to the different action habits of different people, the resulting action characteristic waveforms are also different. Therefore, the classifier defined in the above step is not universal, and a personalized classifier needs to be trained for different users.

Step 7.1: The user does 5 defined actions sequentially 5 times before use.

Step 7.2: Extract 5 pairs of extreme values of the obtained data, and multiply the obtained lowest peak and highest trough by a certain coefficient k as the trained personalized threshold. Personalize the classifier by using this personalized threshold as the threshold of the classifier.

Step 8: Error Correction

The specific implementation process is:

When the user uses it, the user can see the real-time movement of the cursor on the screen. When an error occurs, the user can know that an error has occurred and can immediately start correcting it. If the cursor moves in the wrong direction, the user can perform the same action and then make a contact. For example, when the user moves to the left and finds that the cursor moves to the right, the user can also perform a right action to correct this error. In this way, the previous action is finished, and the subsequent action will not be affected by the previous error.

It should be understood that the parts not described in detail in this specification belong to the prior art.

It should be understood that the above-mentioned descriptions for the preferred embodiments are relatively detailed, and should not therefore be considered as limiting the scope of the patent protection of the present invention. Within the scope of protection, replacements or modifications can also be made, all of which fall within the protection scope of the present invention, and the scope of protection of the present invention should be based on the appended claims.

Claims (5)

1. A wearable mouse control method based on motion sensor, is characterized in that, comprises the following steps: Step 1: Conduct data collection, including: Step 1.1: Select a Bluetooth or Wifi hotspot to connect the embedded wearable device and the controlled computer, and establish a Socket connection to transfer data; the port is set to 8088; Step 1.2: Store the motion data in three directions obtained from the accelerometer and gyroscope in sequence; when the number of received data is greater than the window length, remove the first data; set the sampling rate to 40 times per second; Step 2: Action definition, specifically including the following sub-steps: Step 2.1: Define that the user is facing the screen at the beginning, and the initial head position becomes the middle area; Step 2.2: Define the click operation; the action of moving the head down quickly and then quickly returning to the middle area is defined as the click operation; Step 2.3: Define the direction action; turn the head from the middle area to the left, and keep it for a while until the cursor reaches the desired position, then turn back to the middle area to define the left movement; turn the head from the middle area to the right, and keep it for a while until the cursor reaches the desired position Turning back to the middle area after the desired position is defined as moving right; turning the head from the middle area upwards, keeping it for a period of time until the cursor reaches the desired position and then turning back to the middle area is defined as moving up; Turn the head from the middle area down, keep it for a period of time until the cursor reaches the desired position, and then turn back to the middle area to define the downward movement; Step 3: data preprocessing for eliminating noise and removing tooth waves; Step 4: feature extraction, specifically including the following sub-steps: Step 4.1: Because in the x-axis direction of the accelerometer, the click operation is more significant than the other four directions, so the value of the x-axis direction of the accelerometer can be used to distinguish the click action from the direction action; Step 4.2: Because in the x-axis direction of the gyroscope, the left and right movement is more significant than the up and down movement, so the value in the x-axis direction of the gyroscope can be used to distinguish the left and right movements from the up and down movements; and because the left and right movement characteristics are exactly On the contrary, it is also possible to distinguish left and right actions; Step 4.3: Because in the z-axis direction of the gyroscope, the up-and-down movement is more significant than the left-right movement, so the value in the z-axis direction of the gyroscope can be used to distinguish the up-down movement from the left-right movement; and because the left-right movement characteristics are exactly On the contrary, it is also possible to distinguish the up and down movements; Step 5: Action classification, the specific implementation includes the following sub-steps: Step 5.1: define a _tn ^x , g _tn ^x , g _tx ^x , g _tn ^z , and g _tx ^z as the lower threshold of the accelerometer in the x-axis direction and the upper and lower thresholds of the gyroscope in the x-axis and z-axis directions; the threshold values are determined by the preset Determined by the training step, as the decision data of the decision tree; define a _i ^x , g _i ^x , and g _i ^z as the values of the accelerometer x-axis direction, gyroscope x-axis direction, and gyroscope z-axis direction at the current time i; Check the value of the x-axis direction of the accelerometer at the current time, if a _i ^x <a _tn ^x , the head starts to click, otherwise, go to step 5.2; Step 5.2: Check the value of the x-axis direction of the gyroscope at the current time. If g _i ^x <g _tn ^x or g _i ^x >g _tx ^x , the head will start to move left or right, and go to step 5.3; otherwise, go to step 5.4; Step 5.3: If g _i ^x <g _tn ^x , then the head starts to move to the left, and the cursor on the screen moves to the left; otherwise, the head moves to the right, and the cursor on the screen also moves to the right; Step 5.4: Check the value of the z-axis direction of the gyroscope at the current time. If g _i ^z <g _tn ^z or g _i ^z >g _tx ^z , the head will start to move up or down, and go to step 5.5; otherwise, there is no action occurs; Step 5.5: If g _i ^z <g _tn ^z , then the head starts to move up, and the cursor on the screen moves up; otherwise, the head starts to move down, and the cursor on the screen also moves to the right; Step 5.6: Once the action is recognized, the cursor will execute the corresponding action on the screen until the end of the action, Step 6: Action segmentation. The obtained original data is time-continuous data. Action segmentation divides these time-continuous data into blocks, and the data contained in each block represents an action; according to the definition of action, each action contains two opposite Partial; for example, a leftward movement first turns the head from the middle area to the left, and then turns from the left back to the middle area; as a result, there are two opposite waveforms in the same direction; define two thresholds T in one direction _n , T _p to determine the start and end of the action; among them, T _n is the lower threshold, T _p is the upper threshold; Step 7: Data pre-training, the specific implementation includes the following sub-steps: Step 7.1: The user performs 5 defined actions sequentially 5 times before use; Step 7.2: Extract 5 pairs of extreme values of the obtained data, and multiply the obtained lowest peak and highest trough by a certain coefficient k as the trained personalized threshold; Step 8: Error correction, when the user is using it, the user can see the real-time movement of the cursor on the screen; when an error occurs, the user can know that an error has occurred, and can start to correct it immediately; if the cursor moves to the wrong direction, the user can Perform the same action, and then make a contact; for example, when the user moves to the left and finds that the cursor moves to the right, the user can also perform a right action to correct this error; in this way, the current action is over, and subsequent actions will not Affected by previous errors. 2. the wearable mouse control method based on motion sensor according to claim 1, is characterized in that, the concrete realization of step 3 comprises the following sub-steps: Step 3.1: Use threshold denoising to eliminate noise; define a noise threshold E _th , and only transmit data to the device when the received data is greater than the threshold, otherwise set it to 0; Step 3.2: Use the maximum value filter to remove the tooth wave; the maximum value filter takes the maximum value of all values in the window as the current filtered value; define the filter window as τ, then the current time filtered value r _t can be expressed as : r _t =max{r _t−τ/2 , . . . , r _t−1 , r _t , r _t+1 , . . . , r _t+τ/2 }. 3. the wearable mouse control method based on motion sensor according to claim 1, is characterized in that, the termination definition of action in step 5.6 is as follows: For the left action, when the real-time data g _i ^x ≥ g _tx ^x , the cursor stops moving, and then the real-time data g _i ^x ≤ g _tn ^x is recognized, and the left action is terminated; For the right action, when the real-time data a _i ^x ≤ a _tn ^x , the cursor stops moving, and then the real-time data a _i ^x ≥ a _tn ^x is recognized, and the right action is terminated; For the up action, when the real-time data g _i ^z ≥ g _tx ^x , the cursor stops moving, and then the real-time data g _i ^z ≤ g _tx ^z is recognized, and the up action is terminated; For the down action, when the real-time data g _i ^z ≤ g _tn ^x , the cursor stops moving, and then the real-time data g _i ^z ≥ g _tn ^z is recognized, and the down action is terminated; For the click action, its action is in the same direction as the down action. When the real-time data g _i ^z ≤ g _tn ^x , the cursor stops moving, and then the real-time data g _i ^z ≥ g _tn ^z is recognized, and the click action is terminated. 4. the wearable mouse control method based on motion sensor according to claim 1, is characterized in that, in step 6, uses following rule to divide data: Define that the head is initially facing the screen in the middle area. When the gyroscope data is recognized and reaches a certain threshold for the first time, this point is marked as the point where the action starts, denoted as P _s ; when the gyroscope data is in the opposite direction When a certain threshold is reached for the second time, this point is marked as the point where the action ends, which is recorded as P _e ; then the section between P _s and P _e is divided as an action. 5. the wearable mouse control method based on motion sensor according to claim 1, is characterized in that, the concrete realization of step 7 comprises the following sub-steps: Step 7.1: The user performs 5 defined actions sequentially 5 times before use; Step 7.2: Extract 5 pairs of extreme values of the obtained data, and multiply the obtained lowest peak and highest trough by a certain coefficient k as the trained personalized threshold.