WO2018232557A1

WO2018232557A1 - Exercise movement recognition method and apparatus, and electronic device

Info

Publication number: WO2018232557A1
Application number: PCT/CN2017/088972
Authority: WO
Inventors: 李荣清; 宋志聪; 华桂才
Original assignee: 深圳市酷浪云计算有限公司
Priority date: 2017-06-19
Filing date: 2017-06-19
Publication date: 2018-12-27

Abstract

Provided are an exercise movement recognition method and apparatus, an electronic device, and a computer readable storage medium. The method comprises: tracking an executed movement process of a moving object, and obtaining an original movement signal (S310); performing conversion on the original movement signal by means of a preset conversion algorithm, and extracting a plurality of features of the original movement signal from a converted signal (S330); acquiring a probability of occurrence of each pre-stored feature in various movement types (S350); screening, from the various movement types, a movement type with a maximum probability of simultaneous occurrence of the plurality of features (S370); and obtaining an exercise movement recognition result according to the screened movement type (S390). The technical solution provided by the invention has high exercise movement recognition accuracy, fewer recognition calculations, and high recognition efficiency.

Description

Method and device for identifying motion action, electronic device

Technical field

The present disclosure relates to the field of data processing technologies, and in particular, to a method and device for identifying a motion action, an electronic device, and a computer readable storage medium.

Background technique

With the full development of national fitness, more and more people are involved in sports and fitness. Accompanied by, smart fitness equipment has become very common. Irregularities in exercise and fitness can cause a variety of adverse effects, such as increased risk of sports injuries, reduced muscle coordination, and reduced fitness efficiency. Therefore, in order to perform more efficient and standard exercise, it is necessary to recognize human body movements, for example, to recognize the action of waving a badminton racket, and to recognize the action of playing table tennis.

The identification methods of the existing sports equipment control actions include collecting motion trajectory data of various actions, establishing a sample database, and then collecting motion trajectory data of the motion to be detected to identify the motion of the human body to exercise the sports equipment through trajectory similarity matching. This method can only recognize whether the motion trajectories of the two types of motions are similar, but the motions with similar trajectories cannot be distinguished, and thus the accuracy of the motion recognition is not high.

Another way is to collect the motion parameters of the human motion through the sensor, and then determine the motion of the human body by extracting the feature values of the motion parameters and matching the feature values. This method can realize the recognition of the human body movement to a certain extent, but since the feature values of many actions are close, the accuracy is not high by the method of matching the feature values.

In summary, the prior art has low recognition accuracy for the manipulation of sports equipment.

Summary of the invention

In order to solve the problem that the recognition accuracy of the motion action existing in the related art is not high, the present disclosure provides another method for identifying the motion action to improve the accuracy of the recognition.

In one aspect, the present disclosure provides a method for identifying a motion action, the method comprising:

Tracking the motion process performed by the moving object to obtain the original motion signal;

Transforming the original motion signal by a preset transform algorithm, and extracting from the transformed signal Taking a number of features of the original motion signal;

Obtaining the probability of pre-storing each feature in various action types;

Screening from a variety of action types the types of actions in which the plurality of features are most likely to occur;

The recognition result of the motion action is obtained according to the selected action type.

In one embodiment, the screening of the types of actions that have the highest probability of occurrence of the plurality of features from among the various types of actions includes:

Calculating the probability that each of the action types has the plurality of features simultaneously according to the probability that each feature appears in various action types;

According to the probability that the plurality of features exist simultaneously for each action type, the action type with the highest probability is selected.

In an embodiment, the obtaining the recognition result of the motion action according to the selected action type includes:

Determining whether the selected action type is greater than one type, and if not, filtering the action type as a recognition result of the motion action;

If yes, the original motion signal is matched with the sample motion signal corresponding to each selected motion type to search for a sample motion signal that matches the original motion signal, and the action type corresponding to the matched sample motion signal is used as The recognition result of the motion action.

In an embodiment, the tracking of the motion process performed by the moving object obtains the original motion signal, including:

Collecting, by the sensor on the moving object, the triaxial acceleration component and the angular velocity of the moving object;

Obtaining a posture and a position of the moving object according to the collected three-axis acceleration component and an angular velocity;

An acceleration change waveform signal, an angular velocity change waveform signal, a posture change waveform signal, and a position change waveform signal are obtained according to changes in the triaxial acceleration component, the angular velocity, the posture, and the position of the moving object.

In an embodiment, the original motion signal is transformed by a preset transform algorithm, and extracting some features of the original motion signal from the transformed signal, including:

The acceleration change waveform signal, the angular velocity change waveform signal, the attitude change waveform signal, and the position change waveform signal are waveform-converted by various calculation rules to obtain a transformed waveform signal. Dry feature.

In another aspect, the present disclosure also provides an identification device for an athletic action, the device comprising:

a signal acquisition module, configured to track a motion process performed by the moving object to obtain an original motion signal;

a feature extraction module, configured to transform the original motion signal by using a preset transform algorithm, and extract, from the transformed signal, several features of the original motion signal;

a probability acquisition module, configured to acquire a pre-stored probability of occurrence of each feature in various action types;

An action screening module, configured to filter, from various action types, an action type in which the plurality of features have the highest probability;

The motion recognition module is configured to obtain a recognition result of the motion action according to the selected action type.

In an embodiment, the action screening module includes:

a probability calculation unit, configured to calculate a probability that each of the action types has the plurality of features simultaneously according to a probability that each feature appears in each action type;

The action selecting unit is configured to filter out the action type with the highest probability according to the probability that the plurality of features exist simultaneously for each action type.

In an embodiment, the action recognition module comprises:

a determining unit, configured to determine whether the selected action type is greater than one type, and if not, the filtered action type is used as a recognition result of the motion action;

a matching unit, configured to perform waveform matching on the sample motion signal corresponding to each selected motion type, and search for a sample motion signal that matches the original motion signal, and corresponding to the matched sample motion signal. The action type is used as the recognition result of the motion action.

In still another aspect, the present disclosure also provides an electronic device, the electronic device comprising:

processor;

a memory for storing processor executable instructions;

Wherein the processor is configured to perform the identification method of any one of the above motion actions.

Moreover, the present disclosure also provides a computer readable storage medium storing a computer program executable by a processor to perform an identification method of performing any of the above-described motion actions.

The technical solutions provided by the embodiments of the present disclosure may include the following beneficial effects:

The present disclosure extracts several features of the original motion signal by transforming the original motion signal of the moving object, and then simultaneously appears according to the probability of occurrence of each feature in each known action type. The action type with the highest dry feature probability is recognized as the result of the motion action. Since the scheme does not need to match the waveform signals of the action to be tested one by one with the waveform signals of all known actions, the motion recognition can be realized according to the probability that several features of the unknown motion appear in the known actions, and the calculation amount is compared. Small, the recognition efficiency is high, and further, since the recognition of the motion does not depend on the motion trajectory and the eigenvalue matching, for the action with similar trajectory or the action with the eigenvalue close, by calculating the probability that several features of the unknown action appear in the known action, The recognition of the action can be realized, and thus the recognition accuracy of the motion action is high.

The above general description and the following detailed description are merely exemplary and are not intended to limit the disclosure.

DRAWINGS

The accompanying drawings, which are incorporated in and constitute in FIG

1 is a schematic diagram of an implementation environment in accordance with the present disclosure;

2 is a block diagram of an apparatus, according to an exemplary embodiment;

FIG. 3 is a flowchart of a method for identifying an action according to an exemplary embodiment;

Figure 4 is a flow chart showing the details of step 310 in the corresponding embodiment of Figure 3;

FIG. 5 is a schematic diagram showing waveform waveform transformation of a complex waveform signal to form a waveform signal having significant features, according to an exemplary embodiment; FIG.

6 is a schematic diagram showing the principle of performing feature extraction on a waveform transformation of an original motion signal according to a specified rule, according to an exemplary embodiment;

Figure 7 is a flow chart showing the details of step 370 in the corresponding embodiment of Figure 3.

FIG. 8 is a block diagram of an apparatus for identifying an action according to an exemplary embodiment;

Figure 9 is a detailed block diagram of the action screening module of the corresponding embodiment of Figure 8;

Figure 10 is a detailed block diagram of the signal acquisition module of the corresponding embodiment of Figure 8.

Detailed ways

The description will be made in detail herein with respect to exemplary embodiments, examples of which are illustrated in the accompanying drawings. The following description refers to the same or similar elements in the different figures unless otherwise indicated. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the present invention. Rather, they are only one of the aspects of the invention as detailed in the appended claims. Examples of devices and methods.

1 is a schematic diagram of an implementation environment in accordance with the present disclosure. The implementation environment includes at least one mobile terminal 110 and smart device 120.

The manner of association between the mobile terminal 110 and the smart device 120 includes the network association manner and/or protocol of the hardware, and the data association manner between the two. The smart device 120 also has different locations for installation depending on the type of sports equipment to which it is applied.

The smart device 120 may be disposed in a racket handle of badminton, tennis, or table tennis, or may be set in a baseball bat, or may be set in a soccer ball, a basketball, or the like. Further, the smart device 120 can also be applied to sports equipment such as squash rackets, cricket bats, golf clubs, and the like. In other words, the smart device 120 can be disposed in the active sports equipment to recognize the motion of the human body to exercise the sports equipment according to the activity of the sports equipment. The smart device 120 can also be worn on the wrist or ankle of the human body as needed, and recognize the movement of the human body during sports according to the movement of the wrist or the ankle. For example, the action of waving a racket, the action of waving a baseball bat, the action of playing a football, the action of shooting, and the like.

After the smart device 120 establishes a connection with the mobile terminal 110, the recognition result of the motion action can be uploaded to the mobile terminal 110 for display and storage. A Bluetooth connection can be adopted between the smart device 120 and the mobile terminal 110.

FIG. 2 is a block diagram of an apparatus 200, according to an exemplary embodiment. For example, device 200 can be smart device 120 in the implementation environment shown in FIG.

Referring to FIG. 2, apparatus 200 can include one or more of the following components: processing component 202, memory 204, power component 206, multimedia component 208, audio component 210, sensor component 214, and communication component 216.

Processing component 202 typically controls the overall operation of device 200, such as operations associated with display, telephone calls, data communications, camera operations, and recording operations, and the like. Processing component 202 can include one or more processors 218 to execute instructions to perform all or part of the steps of the methods described below. Moreover, processing component 202 can include one or more modules to facilitate interaction between component 202 and other components. For example, processing component 202 can include a multimedia module to facilitate interaction between multimedia component 208 and processing component 202.

Memory 204 is configured to store various types of data to support operation at device 200. These ones Examples of data include instructions for any application or method operating on device 200. The memory 204 can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as Static Random Access Memory (SRAM), electrically erasable programmable read only memory (Electrically Erasable Programmable Read-Only Memory (EEPROM), Erasable Programmable Read Only Memory (EPROM), Programmable Red-Only Memory (PROM), Read Only Memory ( Read-Only Memory (ROM), magnetic memory, flash memory, disk or optical disk. Also stored in the memory 204 is one or more modules configured to be executed by the one or more processors 218 to perform the method of any of the following Figures 3, 4, and 7 All or part of the steps.

Power component 206 provides power to various components of device 200. Power component 206 can include a power management system, one or more power sources, and other components associated with generating, managing, and distributing power for device 200.

The multimedia component 208 includes a screen between the device 200 and the user that provides an output interface. In some embodiments, the screen may include a liquid crystal display (LCD) and a touch panel. If the screen includes a touch panel, the screen can be implemented as a touch screen to receive input signals from the user. The touch panel includes one or more touch sensors to sense touches, slides, and gestures on the touch panel. The touch sensor may sense not only the boundary of the touch or sliding action, but also the duration and pressure associated with the touch or slide operation. The screen may also include an Organic Light Emitting Display (OLED).

The audio component 210 is configured to output and/or input an audio signal. For example, the audio component 210 includes a microphone (Microphone, MIC for short) that is configured to receive an external audio signal when the device 200 is in an operational mode, such as a call mode, a recording mode, and a voice recognition mode. The received audio signal may be further stored in memory 204 or transmitted via communication component 216. In some embodiments, audio component 210 also includes a speaker for outputting an audio signal.

Sensor assembly 214 includes one or more sensors for providing status assessment of various aspects to device 200. For example, sensor assembly 214 can detect an open/closed state of device 200, relative positioning of components, and sensor assembly 214 can also detect changes in position of one component of device 200 or device 200 and temperature changes of device 200. In some embodiments, the sensor assembly 214 can also include a magnetic sensor, a pressure sensor, or a temperature sensor.

Communication component 216 is configured to facilitate wired or wireless communication between device 200 and other devices. The device 200 can access a wireless network based on a communication standard such as WiFi (WIreless-Fidelity). In an exemplary embodiment, communication component 216 receives broadcast signals or broadcast associated information from an external broadcast management system via a broadcast channel. In an exemplary embodiment, the communication component 216 also includes a Near Field Communication (NFC) module to facilitate short range communication. For example, the NFC module can be implemented based on Radio Frequency Identification (RFID) technology, Infrared Data Association (IrDA) technology, Ultra Wideband (UWB) technology, Bluetooth technology, and other technologies. .

In an exemplary embodiment, the apparatus 200 may be configured by one or more Application Specific Integrated Circuits (ASICs), digital signal processors, digital signal processing devices, programmable logic devices, field programmable gate arrays, Implemented by a controller, microcontroller, microprocessor or other electronic component for performing the methods described below.

FIG. 3 is a flowchart of a method for identifying a motion action, according to an exemplary embodiment. The scope of application and the execution subject of the recognition method of the motion action, for example, the method is applied to the smart device 120 of the implementation environment shown in FIG. As shown in FIG. 3, the identification method may be performed by the smart device 120, and may include the following steps.

Step 310: Track a motion process in which the moving object is performed to obtain an original motion signal.

The moving object may refer to a sports equipment that is controlled to perform a motion process, or a body part such as a wrist or an ankle of a human body that controls the sports equipment. When there is a sports equipment capable of moving, the smart device 120 may be disposed inside the sports equipment, and the original motion signal is obtained by tracking the motion process of the sports equipment. For a stationary sports equipment (such as a horizontal bar) or a situation where no sports equipment is used (such as running), the smart device 120 can be worn at the position of the wrist or ankle of the human body, and the original motion signal can be obtained by tracking the activity of the wrist or the ankle. . Specifically, the smart device 120 can set a sensor, and during the moving object activity, continuously collect motion parameters of the moving object to obtain an original motion signal.

Optionally, as shown in FIG. 4, step 310 specifically includes:

Step 311, continuously collecting the three-axis acceleration component and the angular velocity of the moving object by using the sensor on the moving object;

Wherein, the sensor is carried on the moving object, and the activity is followed by the moving object, and the timing is collected. The triaxial acceleration component of the moving object. The sensor can include a three-axis accelerometer and a gyroscope. The three-axis accelerometer continuously collects the three-axis acceleration component of the moving object, that is, the component of the acceleration on the X-axis, the Y-axis, and the Z-axis. The angular velocity of the moving object can be acquired by the gyroscope.

Step 312: Obtain a posture and a position of the moving object according to the collected three-axis acceleration component and an angular velocity;

The posture of the moving object refers to the inclination of the moving object in the three-dimensional space. By blending the triaxial acceleration component and the angular velocity measurement results, the pitch angle, the yaw angle, and the roll angle of the smart device 120 in the three-dimensional space can be obtained. The MCU (microprogram controller) in the smart device 120 can solve the pitch angle, yaw angle and roll angle of the moving object according to the measurement results of the three-axis accelerometer and the gyroscope by using the existing two-stage extended Kalman filter algorithm. Wherein, the moving object rotates around the x-axis in the three-dimensional space, and the angle θ is a pitch angle, and the angle of rotation around the y-axis

For the roll angle, the rotation angle γ around the z-axis is the yaw angle. The pitch angle, yaw angle, and roll angle of the moving object form the pose of the moving object at that moment.

In addition, the smart device 120 starts from power-on or a certain trigger point (such as changing from a stationary state to a motion state), assuming that the initial position is the coordinate origin, and the acceleration measured according to the three-axis accelerometer is in the X-axis, the Y-axis, and the Z-axis. Upper component, using formula

The coordinates of the smart device 120 on the X-axis, the Y-axis, and the Z-axis can be separately calculated, thereby obtaining the coordinates of the smart device 120 in the three-dimensional space at each moment, and the coordinates in the three-dimensional space form the position at that moment.

Step 313, obtaining an acceleration change waveform signal, an angular velocity change waveform signal, a posture change waveform signal, and a position change waveform signal according to changes in the three-axis acceleration component, the angular velocity, the posture, and the position of the moving object.

The sensor periodically collects the three-axis acceleration component and the angular velocity of the moving object, and according to the change of the three-axis acceleration component, the angular velocity, the attitude and the position of the moving object at each moment, an acceleration change waveform signal, an angular velocity change waveform signal, and a posture change waveform can be formed. Signal and position change waveform signals. It should be noted that the position of the moving object at different times constitutes the position change waveform signal through a connection. These waveform signals can be stored in the RAM (random access memory) of the smart device 120. It should be noted that the acceleration components on the X-axis, the Y-axis, and the Z-axis can be combined to obtain a linear acceleration variation waveform.

Step 330, transforming the original motion signal by using a preset transform algorithm, and extracting, from the transformed signal, several features of the original motion signal;

It should be explained that the original motion signal is subjected to various transformations by a preset transformation algorithm, and a waveform signal having obvious characteristics can be formed. The preset transform algorithm refers to being able to pass the original motion signal The transform forms a transform algorithm with a distinct signature waveform signal. For example, as shown in FIG. 5, the left figure is a waveform curve of the original motion signal. Since it is difficult to describe the characteristics of the waveform curve, the waveform curve can be transformed into the right waveform by frequency domain time domain transformation. Obviously, the waveform characteristics of the right figure are obvious, so that several features of the original motion signal can be obtained from the transformed waveform signal. For example, maximum value, minimum value, etc.

Optionally, the step 330 includes: converting the acceleration change waveform signal, the angular velocity change waveform signal, the attitude change waveform signal, and the position change waveform signal by using a plurality of operation rules to obtain a plurality of features of the transformed waveform signal.

Among them, various operation rules may include conventional addition, subtraction, multiplication, division operations, integration, differential operations, Taylor expansion, Fourier transform, time domain, frequency domain transform, and the like. As shown in FIG. 6, the acceleration change waveform signal, the angular velocity change waveform signal, the attitude change waveform signal, and the position change waveform signal can output a waveform signal having an obvious characteristic by specifying a calculation rule, and the waveform signal having an obvious characteristic can be output according to the output. Several features are obtained, which are several features of the original motion signal.

It should be noted that the specific operation rule is used to process the waveform signal without limitation, and only the characteristics of the transformed waveform signal need to be obvious. For example, if the acceleration has a significant maximum value through the derivative, the acceleration can be transformed, and the waveform feature obtained after the derivative transformation can be recorded as the feature x. By analogy, the acceleration change waveform signal, the angular velocity change waveform signal, the attitude change waveform signal, and the position change waveform signal are waveform-transformed by various calculation rules, and then feature a, feature b, feature c, feature d, ... can be obtained.

Step 350: Acquire pre-stored probability of occurrence of each feature in various action types,

It should be noted that the probability that each feature appears in various action types may be stored in advance in the storage unit of the smart device 120. The probability of occurrence of each feature under each action type may constitute a statistical matrix. The following is a form of statistical matrix:

Each column represents an action type, and each row represents a feature. For example, the probability k ₁₁ is the probability that the action type 1 exhibits the feature a, and k ₂₁ represents the probability that the action type 1 exhibits the feature b, k ₃₄ Represents the probability that feature type 4 appears feature c, and so on.

For different sports, the probability of occurrence of each feature in each action type can be separately calculated. Taking the badminton sport as an example, reference may be made to the above steps 310 and 330. First, by acquiring the motion signals of the samples under different actions during the badminton movement, the waveforms of the sample motion signals under different motions are transformed into waveforms with obvious features. The waveform signal, in turn, obtains several characteristics of the sample waveform signal under different motions. The sample motion signals under different actions can be collected multiple times and feature extraction can be performed. Finally, the probability of occurrence of each feature under each action is calculated, and the above statistical matrix is constructed. If you want to identify new actions or add new features, you only need to improve the statistical matrix, which is more scalable.

Step 370: Filter, from various action types, an action type in which the plurality of features have the highest probability;

Specifically, the n action 1 is performed, and then the probability that the feature a appears in the statistical action 1 is recorded as k ₁₁ . Similarly, the probability that each feature appears in each action is completed, and a statistical matrix is formed. For the newly generated action, if the new action has the features a, b, and c, and there is an action x in the statistical matrix, the probability that the features a, b, and c appear at the same time is the largest, then the action x can be considered as New action.

Optionally, as shown in FIG. 7, step 370 specifically includes:

Step 371: Calculate a probability that each of the action types has the plurality of features simultaneously according to a probability that each feature appears in each action type;

Specifically, it is assumed that the original motion signal of the newly generated action obtains the presence features a, b, and c through steps 310 and 330, and then each motion type is simultaneously generated according to the probability that each feature appears in each action type. The probability of features a, b, c. Assuming that features a, b, and c appear independent of each other, the probability that features a, b, and c appear simultaneously in action 1 is k ₁₁ ×k ₂₁ ×k ₃₁ . Similarly, the probability of simultaneous occurrence of features a, b, and c in each type of action can be calculated separately.

Step 372: Filter the probability of the action type according to the probability that the plurality of features exist simultaneously for each action type.

After calculating the probability that several features (such as features a, b, c) exist in each action type in the above step 371, the action type with the highest probability can be selected. This action type is taken as the recognition result of the newly generated motion action.

For a clear and concise description of the technical solution, it is assumed that only one action 1, 2, 3 exists in a certain motion, and for the newly generated motion action, the presence characteristics a, b, and c are obtained through steps 310 and 330, and the following can be adopted. The calculation process screens out the types of actions with the highest probability of occurrence of several features at the same time:

Q ₁ =k ₁₁ *k ₂₁ *k ₃₁

Q ₂ =k ₁₂ *k ₂₂ *k ₃₂

Q ₃ =k ₁₃ *k ₂₃ *k ₃₃

P ₁ =Q ₁ /(Q ₁ +Q ₂ +Q ₃ )

P ₂ =Q ₂ /(Q ₁ +Q ₂ +Q ₃ )

P ₃ =Q ₃ /(Q ₁ +Q ₂ +Q ₃ )

Max(P ₁ , P ₂ , P ₃ )

Where Q ₁ represents the product of the probability of occurrence of features a, b, and c in action 1, Q ₂ represents the product of the probability of occurrence of features a, b, and c in action 2, and Q ₃ represents the features a, b, and c in action 3. The product of the probability of occurrence, max(P ₁ , P ₂ , P ₃ ), represents the largest of P ₁ , P ₂ , and P ₃ . If P _{1 is the} largest, it means that the newly generated action is action 1, if P _{2 is the} largest, it means that the newly generated action is action 2, and if P _{3 is the} largest, it means that the newly generated action is action 3.

Assuming that the newly generated action has only one feature d, it can be seen from the statistical matrix that the probability of occurrence of the feature d for each action is k ₄₁ , k _42, k ₄₃ ..., wherein the action type with the highest probability of the feature d can be considered as Newly generated action.

Step 390: Obtain a recognition result of the motion action according to the selected action type.

Specifically, the newly generated motion action has several features. On the basis of the above steps, the action type with the greatest probability of having several features at the same time can be selected, and the selected action type can be used as the recognition result of the newly generated motion action. .

Optionally, the step 390 specifically includes: determining whether the selected action type is greater than one type, and if not, filtering the action type as the recognition result of the motion action;

It should be noted that if more than one type of action is selected (possibly small), that is, at least two action types are selected, and the action types have the same probability of having the same feature at the same time, then the filter may be filtered according to the pre-recorded Sample motion signals (such as sample waveform signals) of each action type, and these sample waveform signals are used as a template library to match the original motion signals of the new motion (such as the original waveform signals) with the sample waveform signals in the template library. The sample waveform signal with the highest matching degree with the original waveform signal is searched, and the action corresponding to the sample waveform signal is used as the recognition result of the new motion. Of course, if only one action type is selected, the selected action type is used as the newly generated motion action. Identify the results.

In other embodiments, when more than one type of action is selected, the selected action types may be sorted according to the priority of each action type in the current motion, and the action type with the higher priority is used as the motion action. Identification result. The priority of each action type may be determined according to the proportion of each action type occurring in the current motion.

For example, suppose that in a badminton sport, action 1 accounts for 39%, action 2 accounts for 35%, and action 3 accounts for 20%... (action 1, action 2, action 3... respectively represent different actions) Type), then action 1 can be considered to have a higher priority than action 2, and action 2 has a higher priority than action 3.

In addition, it should be noted that the present disclosure does not directly identify the motion action by means of waveform matching, but first selects the action type by means of probability statistics, and performs waveform matching when filtering out more than one type of action. Since there are many movements in a motion, there are more sample motion signals in the template library. For the original motion signals of the new motion, if the waveforms of all the sample motion signals in the template library are matched one by one, the data processing amount is compared. Large, motion recognition is inefficient. Based on the method of probability and statistics, the action type is filtered, the calculation amount is small, and the motion action recognition efficiency is high.

The present disclosure extracts several features of the original motion signal by transforming the original motion signal of the moving object, and then, according to the known probability of occurrence of each feature in each action type, the action type with the greatest probability of occurrence at the same time is taken as the motion action. Identification result. Since the scheme does not need to match the waveform signals of the action to be tested one by one with the waveform signals of all known actions, the motion recognition can be realized according to the probability that several features of the unknown motion appear in the known actions, and the calculation amount is compared. Small, the recognition efficiency is high, and further, since the recognition of the motion does not depend on the motion trajectory and the eigenvalue matching, for the action with similar trajectory or the action with the eigenvalue close, by calculating the probability that several features of the unknown action appear in the known action, The recognition of the action can be realized, and thus the recognition accuracy of the motion action is high.

In addition, other required parameters can be obtained based on the angular velocity, acceleration, position and attitude as needed. For example, you can also calculate the speed and intensity of the movement.

Assuming that the position coordinates of the smart device 120 at times t ₁ and t ₂ are (x ₁ , y ₁ , z ₁ ) and (x ₂ , y ₂ , z ₂ ), respectively, the displacement of the smart device at times t ₁ to t ₂ for

Derivation can get the speed of each moment.

Taking a swing or a swing motion as an example, the acceleration is known as a, and the mass of the racket or the bat is m. According to the formula F=m*a, the force on the racket or the bat at each moment can also be calculated.

Among them, the recognition result of the motion action and the speed and intensity of the motion action can be stored in the wisdom In the flash (storage unit) of the device 120, the mobile terminal 110 can read the data in the smart device 120 through the software APP, then display and store it, and synchronize to the cloud. When the software APP is reinstalled or replaced with the mobile terminal 110, the previously synchronized data can be acquired from the cloud through the account, and then stored and displayed at the mobile terminal 110.

The following is an embodiment of the apparatus of the present disclosure, which may be used to implement an embodiment of the method for identifying an action performed by the smart device 120 of the present disclosure. For details not disclosed in the embodiments of the present disclosure, please refer to the embodiment of the method for identifying the motion action of the present disclosure.

FIG. 8 is a block diagram of an apparatus for identifying an action, which may be used in the smart device 120 of the implementation environment shown in FIG. 1 to perform FIG. 3, FIG. 4, and FIG. 7 All or part of the steps of the identification method of the motion action shown in any one of the steps. As shown in FIG. 8 , the identification device includes, but is not limited to, a signal acquisition module 810 , a feature extraction module 830 , a probability acquisition module 850 , an action screening module 870 , and a motion recognition module 890 .

The signal acquisition module 810 is configured to track a motion process performed by the moving object to obtain an original motion signal.

The feature extraction module 830 is configured to transform the original motion signal by using a preset transform algorithm, and extract some features of the original motion signal from the transformed signal;

The probability acquisition module 850 is configured to acquire a pre-stored probability that each feature appears in various action types;

The action screening module 870 is configured to filter, from various action types, an action type in which the plurality of features have the highest probability;

The motion recognition module 890 is configured to obtain a recognition result of the motion action according to the selected action type.

The implementation process of the functions and functions of the modules in the above-mentioned devices is specifically described in the implementation process of the corresponding steps in the above-mentioned motion recognition method, and details are not described herein again.

The signal acquisition module 810 can be, for example, one of the physical structure sensor components 214 of FIG.

The feature extraction module 830, the probability acquisition module 850, the action screening module 870, and the motion recognition module 890 may also be function modules for performing corresponding steps in the above-described motion motion recognition method. It will be appreciated that these modules can be implemented in hardware, software, or a combination of both. When implemented in hardware, these modules can be implemented as one or more hardware modules, such as one or more dedicated integrations Circuit. When implemented in software, the modules may be implemented as one or more computer programs executed on one or more processors, such as the programs stored in memory 204 executed by processor 218 of FIG.

Optionally, as shown in FIG. 9, the action screening module 870 includes but is not limited to:

The probability calculation unit 871 is configured to calculate a probability that each of the action types has the plurality of features simultaneously according to a probability that each feature appears in various action types;

The action screening unit 872 is configured to filter out the action type with the highest probability according to the probability that the plurality of features exist simultaneously for each action type.

Optionally, the action recognition module 890 includes but is not limited to:

a matching unit, configured to: when the filtered action type is greater than one type, search for a sample motion signal matching the original motion signal by waveform matching according to a sample motion signal corresponding to each action type pre-stored and filtered, The action type corresponding to the matched sample motion signal is used as the recognition result of the motion action.

Optionally, as shown in FIG. 10, the signal obtaining module 810 includes:

a data acquisition unit 811, configured to continuously acquire a triaxial acceleration component and an angular velocity of the moving object by using a sensor on the moving object;

a posture position determining unit 812, configured to obtain a posture and a position of the moving object according to the collected three-axis acceleration component and an angular velocity;

The waveform determining unit 813 is configured to obtain an acceleration change waveform signal, an angular velocity change waveform signal, a posture change waveform signal, and a position change waveform signal according to changes in the triaxial acceleration component, the angular velocity, the posture, and the position of the moving object.

Further, the feature extraction module 830 includes:

The waveform transformation unit is configured to perform waveform transformation on the acceleration change waveform signal, the angular velocity change waveform signal, the attitude change waveform signal, and the position change waveform signal by using various calculation rules to obtain several characteristics of the transformed waveform signal.

Optionally, the disclosure further provides an electronic device, which can be implemented as shown in FIG. The intelligent device of the environment performs all or part of the steps of the identification method of the motion action shown in any of FIGS. 3, 4, and 7. The electronic device includes:

processor;

a memory for storing processor executable instructions;

The processor is configured to perform the recognition method of the motion action described in the above embodiments.

The specific manner in which the processor of the apparatus in this embodiment performs the operation has been described in detail in the embodiment of the method of identifying the motion action, and will not be explained in detail herein.

In an exemplary embodiment, a storage medium is also provided, which is a computer readable storage medium, such as a temporary and non-transitory computer readable storage medium including instructions. The storage medium stores a computer program that can be executed by the processor to perform the recognition method of the motion action in the above embodiment.

It is to be understood that the invention is not limited to the details of the details and The scope of the invention is limited only by the appended claims.

Claims

A method for identifying a motion action, comprising:

Tracking the motion process performed by the moving object to obtain the original motion signal;

Converting the original motion signal by a preset transform algorithm, and extracting, from the transformed signal, several features of the original motion signal;

Obtaining the probability of pre-storing each feature in various action types;

Screening from a variety of action types the types of actions in which the plurality of features are most likely to occur;

The recognition result of the motion action is obtained according to the selected action type.
The method according to claim 1, wherein the screening of the types of actions in which the plurality of features are most likely to occur simultaneously from the various types of actions comprises:

Calculating the probability that each of the action types has the plurality of features simultaneously according to the probability that each feature appears in various action types;

According to the probability that the plurality of features exist simultaneously for each action type, the action type with the highest probability is selected.
The method according to claim 1, wherein the obtaining the recognition result of the motion action according to the selected action type comprises:

Determining whether the selected action type is greater than one type, and if not, filtering the action type as a recognition result of the motion action;

If yes, the original motion signal is matched with the sample motion signal corresponding to each selected motion type to search for a sample motion signal that matches the original motion signal, and the action type corresponding to the matched sample motion signal is used as The recognition result of the motion action.
The method according to claim 1, wherein the tracking the motion process performed by the moving object to obtain the original motion signal comprises:

Collecting, by the sensor on the moving object, the triaxial acceleration component and the angular velocity of the moving object;

Obtaining a posture and a position of the moving object according to the collected three-axis acceleration component and an angular velocity;

An acceleration change waveform signal, an angular velocity change waveform signal, a posture change waveform signal, and a position change waveform are obtained according to changes in the three-axis acceleration component, angular velocity, attitude, and position of the moving object. signal.
The method according to claim 4, wherein the original motion signal is transformed by a preset transform algorithm, and the plurality of features of the original motion signal are extracted from the transformed signal, including:

The acceleration change waveform signal, the angular velocity change waveform signal, the attitude change waveform signal, and the position change waveform signal are waveform-converted by various calculation rules to obtain several characteristics of the transformed waveform signal.
An identification device for a motion action, comprising:

a signal acquisition module, configured to track a motion process performed by the moving object to obtain an original motion signal;

a feature extraction module, configured to transform the original motion signal by using a preset transform algorithm, and extract, from the transformed signal, several features of the original motion signal;

a probability acquisition module, configured to acquire a pre-stored probability of occurrence of each feature in various action types;

An action screening module, configured to filter, from various action types, an action type in which the plurality of features have the highest probability;

The motion recognition module is configured to obtain a recognition result of the motion action according to the selected action type.
The device according to claim 6, wherein the action screening module comprises:

a probability calculation unit, configured to calculate a probability that each of the action types has the plurality of features simultaneously according to a probability that each feature appears in each action type;

The action selecting unit is configured to filter out the action type with the highest probability according to the probability that the plurality of features exist simultaneously for each action type.
The device according to claim 6, wherein the action recognition module comprises:

a determining unit, configured to determine whether the selected action type is greater than one type, and if not, the filtered action type is used as a recognition result of the motion action;

a matching unit, configured to perform waveform matching on the sample motion signal corresponding to each selected motion type, and search for a sample motion signal that matches the original motion signal, and corresponding to the matched sample motion signal. The action type is used as the recognition result of the motion action.
An electronic device, comprising:

processor;

a memory for storing processor executable instructions;

Wherein the processor is configured to perform the knowledge of the motion action of any of claims 1-5 No way.
A computer readable storage medium, characterized in that the computer readable storage medium stores a computer program executable by a processor to perform the method of identifying the motion action of any of claims 1-5.