TWI497424B - Real time human body poses identification system - Google Patents

Description

Human posture motion recognition system

The invention is a motion recognition system, in particular to a human body posture motion recognition system.

Human body feature recognition technology has become popular in recent years, such as facial feature recognition, fingerprint recognition, and the like. However, there is no relevant research or significant progress in the judgment and identification techniques of limb movements. However, through the judgment of the human body movements, not only can the user's identity be judged, so that the user's identity can be better and more accurate, and the user's identity recognition effect can be accelerated; in addition to identifying the identity, good physical movements Judgment and analysis techniques can also be used as an important aid to train athletes. Take the big league's pitchers as an example. Many pitchers may not pitch the ball after a period of time. They may forget the coordination of the body and can no longer return to the peak state. If you can record the technique of analyzing the physical activity of the player, you will be able to let the player in the ball. The start of the season or the rehabilitation process quickly retrieves the previous pitching position and restores his skill. Therefore, the development of techniques for identifying physical activity characteristics is the primary purpose of the present invention.

In order to solve the problem that the current technology does not have the ability to judge the identity of the user according to the physical movement of the user, and there is no accurate record and analysis of the athlete's physical movement state, resulting in relatively inaccurate identification or the athlete can not restore the past, the problem Human body posture motion recognition system The system performs the technical effect of judging the possible identity of the user by analyzing the user motions taken, analyzing the gait of the user, and the resulting sound.

The present invention provides a human body posture motion recognition system, comprising an attitude motion determining device and a sound receiving module, the sound receiving module comprising a bottom plate and a plurality of sound receivers mounted under the bottom plate, each of the sound receiving Receiving a sound spectrum generated by a user walking on the bottom plate, and outputting each sound spectrum to the posture motion determining device, the posture motion determining device analyzing the frequency and volume of the sound spectrum to determine that the user is walking on the bottom plate The weight and the toe of the heel landing sequence and posture, the posture motion judging device is compared with a database to select a possible or the same user's possible corresponding identity, the database includes a plurality of user identity data and the user's corresponding weight The toe and heel landing sequence and posture.

The human gesture recognition system includes an integral camera positioned adjacent to the sound receiving module, capturing a corresponding direction of the bottom panel and capturing an action posture image of the user walking on the bottom panel, the motion posture image Including an external size, a step size, a height change rate, and a height change path of the user, the posture motion determination device analyzes the local image features and specific features of the motion posture image, and then searches for and outputs the data in the database. Identity information of users with approximate or identical external dimensions, stride size, altitude change rate, and altitude change path data.

The gesture action judging device includes a host and a display unit electrically connected to the host, an input unit, an image processor and an analog digital signal converter, and the image processor and the somatosensory electromechanical Connecting, receiving the captured image result of the somatosensory camera and outputting to the host; the analog digital converter and each sound receiving module The sound receiver is electrically connected, and receives the received result of each of the sound receivers, and outputs the result to the host after filtering, amplifying, etc.; the display unit and the input unit respectively serve as user outputs of the attitude action determining device Enter the interface.

Further, the human gesture recognition system includes a plurality of infrared cameras respectively electrically connected to the host, thereby assisting the host to identify the position of the user and the appearance characteristics thereof.

Wherein, the host determines the order and posture of the user's walking on the bottom plate, the toe heel landing position, and the approximate user identity: receiving the sound spectrum of the user walking through the bottom plate; filtering the sound spectrum After the noise is amplified, the sound characteristic signal of the sound spectrum is taken out, and the sound characteristic signal includes a length, a frequency, and an amplitude; the sound characteristic signal is compared with the database to find the same or similar sound characteristic signal. The corresponding user displays the identity information of the identified user.

The host performs a dynamic gesture image identity determination program to determine the user's possible identity with the captured dynamic gesture image, and the step includes: the somatosensory camera taking an image of the user through the bottom panel; the host will Image binarization and dicing processing to select images of local or specific features of a particular part of the user.

The host takes out the feature vector of the local or specific feature after the image analysis, and determines the head circumference, aspect ratio, height, and specific part of the user. The shape of the body width and the step size parameter; comparing the head circumference size, the aspect ratio, the height, the specific body shape width, the stride size and the like of the user, the parameters are selected by the database or Approximate user's possible or approximate identity data and the display generates user identity information.

Thereby, the present invention can determine the movements of the human body, the sounds that may be generated by walking, and store the movements of the limbs, the generated sounds, the walking postures and the personal user data in a database; when the user passes again, Determine the identity and achieve the technical effect of identification. The identification of the present invention is suitable for applications such as airports that need to quickly identify an identity. The present invention can be used with devices such as facial feature recognition, and can greatly reduce the risk of identity forgery and solve many technical problems of the prior art.

Please refer to the first to fourth figures, which are the human body posture motion recognition system of the present invention, which includes an attitude motion determining device 10, a sound receiving module 20 and one or more somatosensory cameras 30.

The gesture action determining apparatus 10 includes a host 12 and a display unit 14 electrically connected to the host 12, an input unit 16, an image processor 121, and an analog-to-digital converter 125. The image processor 121 is electrically connected to the somatosensory camera 30, and receives the captured image result of the somatosensory camera 30, and outputs the result to the host 12 after image processing and determination.

The analog digital signal converter 125 is electrically connected to a plurality of sound receivers 22 of the sound receiving module 20, and receives each of the sound receivers 22 The radio result is output to the host 12 after processing such as filtering and amplification.

The display unit 14 and the input unit 16 respectively serve as user input and output interfaces of the posture motion determining device 10. The type of the display unit 14 is not limited, and is, for example, a liquid crystal flat panel display, an organic light emitting diode display, or the like.

The host 12 performs a walking type motion recognition method, and uses the images captured by each of the somatosensory cameras 30 and the received sound results of the respective sound receivers 22 to be compared with a database after analysis, thereby being used by the user. The walking action and the generated sound determine the user's possible identity.

The sound receiving module 20 of the embodiment includes a bottom plate 21 and a plurality of the sound receivers 22 mounted on the bottom plate 21. In use, a user 60 walks on the surface of the bottom plate 21, and each sound receiver 22 receives the sound receiving device 22 The user 60 walks on a sound spectrum 20A generated by the bottom plate 21. In order to reduce the recognition error caused by the difference in the spectral response of the sound generated by the material of the shoe worn by the user 60, the user 60 can be standardized to wear shoes of the same material or walk without shoes when walking on the bottom plate 21. When the user 60 walks on the bottom plate 21, the left and right feet 61 of the user 60 generate different sound spectrums 20A when the bottom plate 21 is stepped, as shown in the fourth figure. In order to prevent the sounds received by each of the sound receivers 22 from interfering with each other, the bottom plate 21 can be divided into local blocks that are spaced apart from each other, and one sound receiver 22 is disposed under each partial block, and the two adjacent portions are adjacent. The block has a soundproof construction.

The somatosensory camera 30 is disposed adjacent to the sound receiving module 20, and captures a corresponding direction of the bottom plate and captures an action posture image of the user 60 walking on the bottom plate, wherein the motion posture image includes one of the user outer Type size (head circumference ratio Hf/Wf, body ratio or size), one-step size (λ), a height change rate (ΔH), and a height change path (60A). Further, the embodiment may include a plurality of infrared cameras respectively electrically connected to the host 12, thereby assisting the host 12 in recognizing the position of the user and the characteristics of the extremities, thereby reducing the chance of misjudgment.

Further, in practice, a plurality of the somatosensory cameras 30 can be provided for capturing users 60 of different angles, thereby reducing the problem of misjudgment caused by a single camera camera 30 due to the shooting angle, and each of these The shooting results of the somatosensory camera 30 can also respectively store the identity data of the user 60, thereby providing a more accurate identification result. For example, in this embodiment, three of the somatosensory cameras 30 capture the front, the side, and the upper side of the user 60, and store the front, side, and upper dimensions of the head circumference and the body of the user 60 for comparison. It is judged that the user 60 is more precise in appearance.

The host 12 determines the identity of the user 60 based on the captured sound spectrum 20A and the motion gesture image and a database. The database may be configured to first assign each of the users 60 to walk through the bottom plate 21 to receive the sound spectrum 20A of each user 60 and the motion gesture image, and use the 60 The identity data of the subject is stored in the database, as shown in the fifth and seventh figures.

Each of the sound spectrum 20A unique to the use of the 60 walking is stored, and the steps of the fifth figure are as follows: receiving the aisle sound sensing signal: each sound receiver 22 receives the sound spectrum generated by the user 60 walking on the bottom plate 21. 60A.

Filter amplification processing: the host 12 will extract the sound spectrum 60A Zoom in after filtering out the noise.

The stepped ground sound wave characteristic signal is taken: a sound feature signal is taken out from the sound spectrum 60A. The sound feature signal includes a difference in sound caused by the left foot and the right foot, a length of the sound spectrum 60A generated by the foot contacting the bottom plate 21, a volume, a frequency distribution, and the like, and the user's 60-legged contact with the floor 21 is light and heavy, and the toe is heeled. Order and posture.

The storage establishment database: the host 12 matches the identity data of one of the users 60 to the sound feature signal and simultaneously stores the data in the database. Under the condition of the same material contact, the light weight and posture of each person's normal walking are the different influence factors of the generated sound spectrum 60A. Under the condition of not deliberately changing, each person is in the same material and approximate stride. In this case, the generated sound spectrum 60A has a high degree of approximation, and therefore, individual differences in the sound spectrum 60A generated by each user 60 can be found through recording and analysis. In this step, the analyzed audio spectrum 60A corresponds to the personal identification data of the user 60.

The action gesture image unique to each user 60 is stored, and the step of receiving the captured image may be as follows: the host 12 drives the somatosensory camera 30 to capture an action gesture image of the user 60 walking on the bottom plate 21.

Binarization and cutting image: The host 12 binarizes the motion gesture image, and cuts the motion gesture image for capturing a specific part of the user 60, thereby analyzing the size of the specific part; for example, the user 60 Head size, aspect ratio, height, specific body shape width, stride size, etc.

Capture image feature vector: the host 12 analyzes the motion gesture image A particular feature vector is retrieved, such as the feature vector of the height change path 60A of the head of the user 60. Each feature vector is related to the aforementioned head circumference size, aspect ratio, height, specific body shape width, stride size and the like.

Obtaining the size, stride size, height change rate and height change path of the subject: the host 12 obtains the feature vector and the partial image analysis, and defines the partial size, stride size and height change of the user 60. Rate and height change path.

The storage establishment database: the host 12 stores the information of each person of the user 60 and the analysis or calculation result described above to the database.

Please refer to the sixth figure and the eighth figure and the ninth figure. When the user 60 walks through the bottom plate 21 again, after the host 12 has the image or sound data of the somatosensory camera 30 and the sound receivers 22, the host 12 is based on the host computer 12 The following steps determine and output the personal data of the user 60, and the steps include: Receiving aisle sound sensing signal: The host 12 receives the sound spectrum 60A that the user 60 walks through the backplane 21 again.

Filter amplification processing: The host 12 filters the sound spectrum 60A to remove noise and then amplifies.

The sound signal of the footstep is extracted: the host 12 takes out the sound feature signal of the sound spectrum 60A, and the sound feature signal includes a length, a frequency, an amplitude, and the like.

Correct (judgement): The host 12 compares the sound feature signal with the database to find the user 60 corresponding to the same or similar voice feature signal. For example, the user 10 who has the first 10 walking signals and the sound signal is more than 80% is selected.

Output identification result: The host 12 outputs and displays the found user 60.

According to the user 60 again passing the bottom plate 21, the somatosensory camera 30 captures an image again, and the host computer 12 executes a dynamic posture image recognition program to determine the possible identity of the user 60 by the captured dynamic posture image, and the steps include : Receiving a captured image: The somatosensory camera 30 captures an image of the user 60 passing through the bottom plate 21.

Binarization and cutting: The host computer 12 binarizes and cuts the image to select an image of a particular portion of the user 60 or a particular feature.

The image feature vector is captured: the host 12 takes out the feature vector of the local or specific feature after the image analysis, and determines the head circumference size, aspect ratio, height, specific body shape width, stride size, etc. of the user 60. The status of the parameter.

Perform image recognition and comparison of height change rate, high step change path, size and stride size: compare the head circumference size, aspect ratio, height, specific part width, stride size and other parameters of the user 60, The database may select, from the database, the possible or approximate identity data of the user 60 that the parameters meet or approximate.

Correct (judgment): the host 12 determines whether the user's head circumference size, aspect ratio, height, specific body shape width, stride size, and the like are found in the database to approximate or the same corresponding use. The identity of 60.

Output image recognition result: The host 12 outputs and controls the display unit 14 to generate a determination result.

In order to reduce the error that may occur after using the sound or image recognition, the sound and image can be combined with the image to identify the user 60 more accurately. The steps include: performing the voiceprint identification and comparison of the footstep: The host 12 takes the sensing results of the respective sound receivers 22 and finds a list of similar or identical users 60 in the database based on the aforementioned analysis and comparison. When there is no approximate user in the database, the display unit 14 is driven to display "the database has no such person" or a similar warning message.

Performing image recognition and comparison: the host 12 takes the action posture image result of the somatosensory camera 30, and judges the selected user 60 list by the sound, and further selects the user 60 whose walking voiceprint or image feature determination result is consistent. And displayed on the display 14. On the other hand, when there is no approximate user in the database, the display unit 14 is driven to display "the database has no such person" or a similar warning message.

Based on the foregoing description, the present invention can determine the movements of the human body and the sounds that may be generated by walking, and store the movements of the limbs, the generated sounds, the walking postures and the personal user data in a database; when the user passes again, Based on this, the identity is determined to achieve the technical effect of identity recognition. The identification of the present invention is suitable for applications such as airports that require rapid identification of identity, and the present invention can be matched with devices such as facial feature recognition, and the risk of identity forgery can be greatly reduced.

10‧‧‧ attitude motion judging device

12‧‧‧Host

14‧‧‧Display unit

16‧‧‧Input unit

121‧‧‧Image Processor

125‧‧‧ analog digital signal converter

20‧‧‧Sound Receiver Module

20A‧‧‧Sound spectrum

21‧‧‧floor

22‧‧‧Sound Receiver

30‧‧‧Sports camera

60‧‧‧Users

60A‧‧‧Highly variable path

The first figure is a schematic diagram of a system in accordance with a preferred embodiment of the present invention.

The second figure is a block diagram of a system according to a preferred embodiment of the present invention.

The third figure is a schematic diagram of a captured image according to a preferred embodiment of the present invention.

The fourth figure is a schematic diagram of sound extraction in accordance with a preferred embodiment of the present invention.

The fifth to ninth figures are flowcharts of a preferred embodiment of the present invention.

20‧‧‧Sound Receiver Module

60‧‧‧Users

60A‧‧‧Highly variable path

Claims (9)

A human body posture motion recognition system includes an attitude motion determining device and a sound receiving module, wherein: the sound receiving module comprises a bottom plate and a plurality of sound receivers mounted under the bottom plate, each of the sound receivers Receiving a sound spectrum generated by a user walking on the bottom plate, and outputting each sound spectrum to the posture motion determining device; the posture motion determining device analyzes the frequency and volume of the sound spectrum to determine the light weight of the user walking on the bottom plate The sound difference and posture caused by the toe landing sequence, the left foot and the right foot, and the posture motion judging device is compared with a database to select an approximate or identical user's possible corresponding identity, and the database includes a plurality of uses. The identity data corresponds to the user's weight, the toe and the heel landing sequence and posture; and the somatosensory camera is disposed adjacent to the sound receiving module, which captures the corresponding direction of the bottom plate and draws the user to walk on the bottom plate An action gesture image that includes an exterior size of the user, a stride size, a height change rate, and a height change path. After analyzing the local image feature and the specific feature of the action pose image, the posture action judging device searches for and outputs the approximate or identical outer size and step in the database. User identity information for size, height rate of change, and height change path data. The human body posture motion recognition system according to claim 1, wherein the posture motion determining device comprises a host and a display unit electrically connected to the host, an input unit, an image processor and an analog digital signal. a converter, the image processor and the somatosensory electromechanical connection Receiving, receiving the captured image result of the somatosensory camera and outputting to the host; the analog digital signal converter is electrically connected to a plurality of sound receivers of each sound receiving module, and receiving the sound of each of the sound receivers As a result, the processing is output to the host after filtering, amplification, etc.; the display unit and the input unit respectively serve as user input and output interfaces of the attitude action determining device. The human body posture motion recognition system according to claim 1 or 2, wherein the plurality of infrared cameras are electrically connected to the host, respectively, to assist the host in recognizing the position of the user and the appearance characteristics thereof. For example, in the human body posture motion recognition system described in claim 3, the host system determines the light weight, the toe heel landing sequence and posture of the user walking on the bottom plate, and the approximate user identity: receiving the user by the following steps: Passing through the sound spectrum of the bottom plate; filtering the sound spectrum to remove noise; and extracting the sound characteristic signal of the sound spectrum, the sound characteristic signal includes length, frequency, amplitude; taking the sound characteristic signal and the database The comparison finds the user corresponding to the same or similar voice feature signal and displays the identity information of the identified user. For example, in the human body posture motion recognition system described in claim 3, the host performs a dynamic gesture image identity determination program to determine the user's possible identity with the captured dynamic gesture image, and the steps include: the somatosensory camera Taking the image of the user through the bottom plate; the host binarizes and cuts the image to select an image of a specific part or specific feature of the specific part of the user; The host takes out the feature vector of the local or specific feature after the image analysis, determines the head circumference size, aspect ratio, height, specific body shape width, and stride size parameter state of the user; compares the user's The size of the head circumference, the aspect ratio, the height, the width of the specific part, the stride size, and the like, and the database selects the possible or approximate identity data of the user whose parameters meet or are approximate and the display generates the user. Identity information. The human body posture motion recognition system according to claim 4, wherein the host performs a dynamic gesture image identity determination program to determine the user's possible identity with the captured dynamic gesture image, the step comprising: the somatosensory camera Taking the image of the user through the bottom plate; the host binarizes and cuts the image to select an image of a specific part or specific feature of the specific part of the user; and the host takes out the feature of the local or specific feature after the image analysis The vector determines the head circumference size, aspect ratio, height, specific body shape width, and stride size parameter of the user; compares the head circumference, aspect ratio, height, and specific body of the user Parameters such as shape width and stride size, and the database selects the possible or approximate identity data of the user whose parameters meet or are approximated and displays the user identity information. The human body posture motion recognition system according to claim 4, wherein the database is configured such that each user walks the bottom plate and captures sound and image features, and stores the identity information of each user accordingly. . Human body posture motion recognition system as described in claim 6 The database is configured such that each user walks the bottom panel and captures sound and image features, and stores the identity information of each user accordingly. The human body posture motion recognition system according to claim 8, wherein the bottom plate comprises a plurality of mutually spaced partial blocks, and each of the partial blocks is provided with a sound receiver, two adjacent local regions. The block is provided with a soundproof construction.