This application claims the priority benefit of Taiwan application serial no. 102136045, filed on Oct. 4, 2013. The entirety of the above-mentioned patent application is hereby incorporated by reference herein and made a part of this specification.
The disclosure relates to a system and a method of multi-user coaching inside a tunable motion-sensing range.
Nowadays, people pay more attention to proper physical exercise. In recent years, the diversity of sports allow people not to work out on the equipment in general fitness gyms but also to do exercise for full-body fitness, such as dancing, aerobics, rhythmic gymnastics, etc. Thereby, people do not get bored and are able to persevere in exercising.
Most of the full-body exercises should be paid for requiring the coach at the fitness gym. However, the coach cannot well instruct individual users due to numerous students at the gym, the students neither learn effectively nor fully understand the correct posture.
When a player plays the commercial games via the Kinect on XBOX360 KINECT, his or her movement can be tracked by the KINECT sensor to experience the most real sport amusement by undergoing the professional one-by-one fitness program in a friendly virtual gym. However, the virtual private coach only guide one player in a one-by-one program to compare his or her similarity scores at certain time frames, and it also cannot compare similarity scores with multiple players through consecutive time frames nor evaluate the correctness of the posture or the motion of individual player. In addition, the view-angle range of the infrared Kinect sensor in the horizontal direction is only 57.5°, which will limit the detection range of the research, development, and even application. Moreover, to provide images with the expanding range, the image stitching technique is only applied to stitch images by multiple RGB cameras.
The disclosure provides a multi-user coaching system, and the system includes an image processing unit, a user motion capturing unit, a multithreading processing unit, a multi-user comparing unit, and a coaching database. The image processing unit receives a plurality of images in parallel and performs a depth image stitching process and a user skeleton capturing process on the plurality of images to obtain a plurality of user information. The user motion capturing unit generates a plurality of depth streaming motion images and a plurality of node information corresponding to each of the plurality of user information. The multithreading processing unit transmits in parallel the node information and the plurality of depth streaming motion images corresponding to the plurality of user information. The multi-user comparing unit is coupled to the multithreading processing unit, and receives and compares the plurality of depth streaming motion images and the node information with virtual coach information to generate a plurality of personalized comparing results. The coaching database stores a plurality of predetermined motion instructions and outputs the motion instructions corresponding to the personalized comparing results.
The disclosure provides a multi-user coaching method. In the method, a plurality of images is received in parallel, and a depth image stitching process and a user skeleton capturing process are performed on the plurality of images to obtain a plurality of user information. A plurality of depth streaming motion images and a plurality of node information are generated corresponding to each of the user information. The plurality of depth streaming motion images and each of the plurality of node information are comprated with virtual coach information to generate a plurality of personalized comparing results. A plurality of motion instructions are output corresponding to the personalized compared results.
Several exemplary embodiments accompanied with figures are described in detail below to further describe the disclosure in details.
The accompanying drawings are included to provide further understanding, and are incorporated in and constitute a part of this specification. The drawings illustrate exemplary embodiments and, together with the description, serve to explain the principles of the disclosure.
The existing motion-sensing system can merely be used by one user at one time; that is, the system can only perform calculations on motion similarity of one user, and trigger the motion instruction feedback function for that user, which may be exemplified by the U.S. patent application Ser. No. 13/849,535 (corresponding to Taiwan patent application No. 101147275 and the P.R. China patent application No. 201210568013.X) proposed by the inventors of this application and entitled “METHOD AND SYSTEM FOR MOTION COMPARISON”. The entirety of each of the above-mentioned patent applications is incorporated by reference herein and made a part of this specification. The disclosure is derived from said applications. According to an exemplary embodiment of the disclosure, skeletons of multiple users are input to the system, and the system performs calculations on motion similarity of each user and triggers the motion instruction feedback function for each user. Thereby, the multi-user coaching technology may be applied on various scenarios.
By using the system and the method of multi-user coaching inside a large tunable motion-sensing range, motions of users are compared at the same time to perform calculations on motion similarities and to trigger the motion instruction feedback function for individual user. Thereby, the system and the method described herein may be extensively applied to various scenario and product development, such as a virtual gymnasium, a virtual aerobics classroom, a virtual Budokan, and so on.
In an exemplary embodiment of the disclosure, a multi-user coaching system inside a tunable motion-sensing range is provided. A plurality of motion-sensing cameras are applied to capture images, and a depth image stitching module is applied to perform a depth image stitching process on the captured images to expand the motion-sensing range; thereby, a virtual environment for multi-user coaching is established, and each user can be coached individually by a one-to-multiple approach to improve his or her motions.
The sensor set 110 detects and obtains a plurality of images. The images at least include a plurality of user's images. As shown in
The depth streaming motion images and the node information of the users are transmitted to the multithreading processing unit 140 to process a plurality of threads in parallel at the same time. Multithreading is a kind of technology of executing the threads in parallel by means of software or hardware. The hardware support of the processor capable of performing the multithreading process allows the processor to execute more than one thread at the same time, and the overall processing performance is enhanced. The system possessing such capability includes a symmetric multiprocessor, a multi-core processor, a chip-level multithreading processor, and a simultaneous multithreading processor. As to software multithreading, even the processor merely executes one thread, the operation system is able to switch between different threads quickly. The time duration at which the operation system switches between different threads is so small as if the multiple threads are executed at the same time, which is referred to as the software multithreading.
The multithreading processing unit 140 respectively transmits the depth streaming motion images and the node information corresponding to the user information to the multi-user motion comparing unit 150 and the multi-user gravity comparing unit 160. In an exemplary embodiment, the multi-user motion comparing unit 150 includes a calculation unit 152 and a virtual coach motion information unit 154. The multi-user gravity comparing unit 160 includes a user gravity calculating unit 162, a calculation unit 164, and a virtual coach gravity calculating unit 166.
In the multi-user motion comparing unit 150, the virtual coach motion information unit 154 provides the pre-recorded virtual coach depth streaming motion images and the node information. The virtual coach motion information unit 154 can process the corresponding information according to the course or the program selected by the user. The calculation unit 152 compares and analyzes the trajectories of a plurality of user depth streaming motion images and the trajectory of the virtual coach depth streaming motion image to provide a plurality of personalized comparing results. For example, there are M users, and then the calculation unit 152 compares and analyzes the trajectories of the M users' depth streaming motion images and the trajectory of the corresponding virtual coach depth streaming motion image.
In the multi-user gravity comparing unit 160, the virtual coach gravity calculating unit 162 which is connected to the virtual coach motion information unit 154 calculates the gravity of the virtual coach depth streaming image. The user gravity calculating unit 162 calculates the gravity of the depth streaming motion images of multiple users at the same time, respectively. The calculation unit 164 compares and analyzes the individual users' gravity shifting trajectories and the virtual coach gravity shifting trajectory at the same time to provide the corresponding personalized gravity shifting comparing and analyzing results.
The coaching database 170 stores advices of the predetermined motion instructions, e.g., the speed of the motion, the location where the motion takes places, the speed and the direction of shifting the gravity, etc. The coaching database 170 searches and outputs appropriate advices of the motion instructions according to the personalized comparing results and the personalized gravity shifting comparing results. The output of the multi-user coaching system 100 inside a tunable motion-sensing range is transmitted to the monitor 180 to display relative information. The monitor 180 displays a plurality of advices of the personalized motion instructions for the users to correct the motions.
By means of the multi-user coaching system inside a tunable motion-sensing range as shown in
Please refer to
The image data captured by the M motion-sensing cameras are transmitted to a host computer 220. A multi-user coaching system 222 inside a tunable motion-sensing range provided in the exemplary embodiment is configured in the host computer 220. The multi-user coaching system 222 can be executed due to the hardware architecture within the host computer 220 or any independent hardware architecture, which does not depart from the scope of the disclosure. According to the images captured by the motion-sensing cameras 210-214, a virtual environment of the multi-user coaching system 222 is established, and the information can be output and displayed to the monitor 230. Each user can be coached individually by a one-to-multiple approach to improve his or her motions at the same time. As shown in
During exercise, N users' depth streaming motion images and the skeleton node information can be captured at the same time from a new stitched depth image input by a user motion capturing unit. Additionally, the virtual coach motion information unit provides the recorded virtual coach depth streaming motion images and the skeleton node information. The multithreading processing unit processes the N users' skeleton inputted to one single calculation unit to process motion similarity comparisons in continuous time frames. The individual motion is compared and analyzed according to the motion trajectory of the visual coach and the motion trajectories of the N users in terms of the direction, the speed, and the distance. On the other hand, the multithreading processing unit also processes the depth images of N individual users inputted to the user gravity calculating unit, and the virtual coach motion information unit provides the depth image inputted to the virtual coach gravity calculating unit. Then the calculation unit compares the gravity shifting analyzing result at the same time; according to the personalized comparing results and the personalized gravity shifting comparing and analyzing results, a coaching database which stores the advices of the motion instructions is searched, and N appropriate personalized motion instructions are output to and displayed on a monitor for the users.
Please refer to
The comparison and calculation in one of several exemplary embodiments can refer to comparison of gravity. After the user information of N users is obtained, e.g., the depth image information and the skeleton node information, gravity comparison can be performed. First, the gravity of the depth streaming image of virtual coach is calculated, and the gravity of the depth streaming image of each individual user is calculated at the same time. The individual users' gravity shifting trajectories are analyzed and compared with the visual coach gravity shifting trajectory at the same time, and the corresponding personalized gravity shifting comparing and analyzing result is provided. In another embodiment, the motion and the gravity can be compared in parallel at the same time according to the obtained user information.
Please refer to
Then, components in the multi-user coaching system inside a tunable motion-sensing range are explained below.
Please refer to
Please refer to
In one of several exemplary embodiments in this disclosure, the foregoing calibration function is a function of calibrating the view angles of the motion-sensing cameras and includes following step. First, a feature extraction method is performed to detect a plurality of feature points of RGB images in every two motion-sensing cameras, and the nearest neighbors search method is applied to obtain the correlations between feature points in every two of the RGB images. Since 1) the overlapping rate of every two images is not high because the depth values of the feature points for every dual image are too close, 2) the feature points on the left pixel of left image and the right pixel of right image in the horizontal (x-axis) direction is invalid due to the low overlapping rate, 3) the difference in every two feature points in dual RGB images in a vertical (y-axis) direction should not be significant, 4) the position distribution of every two corresponding feature points is unified, and 5) the depth values along with feature points are converted into the real word coordinates, by converting the depth values along with feature points to world coordinates, the x-axis difference in every two corresponding feature points can be obtained, the x-axis difference in every two corresponding feature points is sorted, and pairs of the corresponding feature points are obtained. These pairs of the corresponding points are characterized by the high correction rate. The foregoing process is stopped until the corresponding points with the obtained high correction rate, and whether the high correction rate can be determined by a threshold value.
Every two corresponding points obtained by performing said step can be divided into two groups. The first group is the feature points of image A, and the second group is the feature points of image B. With respect to the feature points of the first group as the basis, the feature points of the second group are rotated along with the horizontal (x-axis) and vertical (y-axis) directions, respectively. By calculating the angle of the nearest distance of the two groups of feature points in the real world coordinate, the view angle of image B can be calibrated, and the different diagrams illustrating can be completed by comparing before and after the calibrating view angles of every two images. After the calibration and the stitching steps are performed on every two of the images, the final depth images 638 are obtained. The final depth images 638 are, in form of meta data, transmitted via the depth image configuration generator 640 to the user skeleton generator 650, so as to obtain the user information and track the user skeleton. Therefore, the depth image and skeleton of the user can be tracked and captured from the final stitched depth image 638.
From the foregoing, the limitation of the horizontal view angle (57.5°) for one single motion-sensing camera no longer exits, and the motion-sensing range can be expanded and dynamically adjusted. As to the existing issue that the motion-sensing system is merely applicable to one single user, skeletons of multiple users are input to a motion similarity comparing element, and the motion similarity and the motion instructions of the individual user are calculated as shown in
Please refer to
In step S722, each of the motion-sensing cameras respectively captures the depth images (including a plurality of environmental images and the images of N users) in its view angle range, and the individual depth images are converted into the real world coordinate. In step S724, if the view angles of the motion-sensing cameras are not aligned, these view angles can be calibrated and aligned by the calibration function. In step S726, the plurality of the corresponding points of the real world coordinate can be determined from the location of the motion-sensing cameras, and the corresponding points are aligned and stitched to generate the new depth expanded images. Note that this step includes the hidden surface removal step. In step S728, the new depth expanded image in form of meta data is transmitted to the core architecture of the motion-sensing unit, so as to obtain the user information. In step S730, the depth streaming motion images and the skeleton node information of the N users are detected.
In step S740, the skeleton node information of the N users are processed in parallel by multithreading. In step S752, a recorded virtual coach depth streaming motion and the skeleton node information are provided. In step S750, a motion comparing and coaching method and a gravity comparing method are provided. The motion trajectories of the N users and that of the virtual coach are compared and analyzed, and the gravity shifting comparison and analysis is also performed. The comparing and analyzing results are output to the N users in real-time. In step S760, personalized instructions are provided to the N users to correct the motions.
It will be apparent to those skilled in the art that various modifications and variations can be made to the structure of the disclosed embodiments without departing from the scope or spirit of the disclosure. In view of the foregoing, it is intended that the disclosure cover modifications and variations of this disclosure provided they fall within the scope of the following claims and their equivalents.
Number | Date | Country | Kind |
---|---|---|---|
102136045 A | Oct 2013 | TW | national |
Number | Name | Date | Kind |
---|---|---|---|
7961174 | Markovic et al. | Jun 2011 | B1 |
20100195867 | Kipman | Aug 2010 | A1 |
20100277411 | Yee et al. | Nov 2010 | A1 |
20100306685 | Giaimo, III | Dec 2010 | A1 |
20100306712 | Snook | Dec 2010 | A1 |
20100306714 | Latta | Dec 2010 | A1 |
20110035666 | Geisner | Feb 2011 | A1 |
20110175809 | Markovic et al. | Jul 2011 | A1 |
20110228976 | Fitzgibbon | Sep 2011 | A1 |
20110304632 | Evertt | Dec 2011 | A1 |
20110306396 | Flury | Dec 2011 | A1 |
20110316963 | Li | Dec 2011 | A1 |
20120113092 | Bar-Zeev | May 2012 | A1 |
20120190505 | Shavit | Jul 2012 | A1 |
20120293518 | Geisner | Nov 2012 | A1 |
20130034264 | Lin | Feb 2013 | A1 |
20130083064 | Geisner | Apr 2013 | A1 |
20130286004 | McCulloch | Oct 2013 | A1 |
20140073383 | Lu | Mar 2014 | A1 |
20140147820 | Snow | May 2014 | A1 |
Number | Date | Country |
---|---|---|
102221886 | Oct 2011 | CN |
I286717 | Sep 2007 | TW |
201111014 | Apr 2011 | TW |
201242641 | Nov 2012 | TW |
201308253 | Feb 2013 | TW |
Entry |
---|
“Office Action of Taiwan Counterpart Application”, issued on Nov. 3, 2015, p. 1-p. 9, in which the listed reference was cited. |
Albaina et al., “Flowie: A Persuasive Virtual Coach to Motivate Elderly Individuals to Walk,” PervasiveHealth 2009 3rd International Conference on Pervasive Computing Technologies for Healthcare, Apr. 1-3, 2009, pp. 1-7. |
Tim Wortmann, “Automatic Stitching of Micrographs using Local Features,” 2010 International Symposium on Optomechatronic Technologies (ISOT), Oct. 25-27, 2010, pp. 1-6. |
Lu et al., “A Motion-Sensing Enabled Personalized Exercise System for Cardiac Rehabilitation,” 2012 IEEE 14th International Conference on e-Health Networking, Applications and Services (Healthcom), Oct. 10-13, 2012, pp. 167-171. |
Faion et al, “Intelligent Sensor-Scheduling for Multi-Kinect-Tracking,” 2012 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), Oct. 7-12, 2012, pp. 3993-3999. |
Lou et al., “A Multi-User Interaction System Based on Kinect and Wii Remote,” 2012 IEEE International Conference on Multimedia and Expo Workshops (ICMEW), Jul. 9-13, 2012, pp. 667. |
Number | Date | Country | |
---|---|---|---|
20150098625 A1 | Apr 2015 | US |