CAPÍTULO VI: ROL DEL PROFESIONAL DE RECURSOS HUMANOS EN LAS
6.2. La planificación del profesional de recursos humanos dentro de las entidades u
Kinect provides a floor-clipping-plane vector
A B C D , where A, B, C and D are the coefficients , , ,
of an estimated floor-plane equation AxByCz . D 0
A B C, ,
is the unit vector correspond-ing to the normal of the floor plane. D is the distance from the floor plane to the origin of the coor- dinate system associated with the Kinect, i.e., the height of the camera from the floor (in meters). When the floor is not visible or detectable, the floor clipping plane is a zero vector. The tilt angle in radians is obtained using (4.1).
tilt angle = atan C B
(4.1)
The file extension used for the different data types are: “.kvid”, “.kir”, “.kdpt”, “.kbi”, and “.kpos”, for colour, IR, depth, body index, and 3-D body joint data, respectively. For each acquisition and data type, a new file is created every minute. The name of the file is the timestamp of the first frame followed by the “#” symbol and the number of the acquisition (it is possible to carry out several acquisitions per session). When using the Kv2, a file with information regarding coordinate mapping is also saved (extension “.kmap”), which can be used for offline mapping between depth and camera spaces (more details can be found in [176]).
4.2 Kinect Motion Analyser (KiMA) Application
To allow the review of the data acquired with KiT, as well as the indication of relevant move- ments for motion analysis, we developed a different application: Kinect Motion Analyser or KiMA. The main requirements defined for KiMA are the following:
2.1. Offline visualization of a data acquisition carried out in the KiT application, as a video, including the possibility of viewing two or more data types simultaneously;
2.2. Selection of the session and acquisition to be reviewed (based on the same session organization used in KiT);
2.3. Creation, edition and deletion of “labels” and “events” associated with a given instant and time interval, respectively;
2.4. Exportation of events.
These requirements corresponds to the use cases represented the diagram shown in Figure 4.5. These use cases are carried out by the clinician in the clinical scenario, but can also be carried out by another person in other scenarios (e.g., ambulatory monitoring, research).
Figure 4.5. Use case diagram associated with the KiMA application.
The GUI of KIMA’s main window is shown in Figure 4.6. The data acquired using KiT can be reviewed as a video in KiMA (requirement 2.1), by using the typical video manipulation controls (e.g., play, pause, stop, fast forward, rewind). It is also possible to define the speed of the video, and jump to the previous or next frame (when paused). The data are presented to the user in a similar way to KiT (i.e., up to three data types can be viewed at the same time).
To review an acquisition, it is firstly necessary to choose the folder where the session is stored (Options panel). Then, it is possible to select one of the sessions available for the indicated folder. Finally, the user can chose one of the acquisition performed in the selected session (require- ment 2.2). When an acquisition is chosen, the first frame of the video is shown. Only the data types available for that acquisition can be selected as primary and secondary sources. The video manipu- lation (play, stop, etc.) involves obtaining the data from the files that corresponds to the current frame number, ensuring the synchronization between the different data types.
KiMA also facilitates the creation, edition and deletion of tags (labels and events), which enable the identification of relevant movements that occur during an acquisition (requirement 2.3). A label has an associated frame number and timestamp, as well as a name and an optional description. An event is similar to a label, with the difference that it has a beginning and an ending instant.
The information associated with tags created by the user is saved to an XML file, which can the be used in other application to carry out motion analysis taking into account the created tags. The
4.3 Summary
Figure 4.6. KiMA’s main window GUI, including the display of depth and body data, as well as labels and events.
Tags can be created/deleted/edited by using the context menu shown in Figure 4.6, or the shortcuts indicted in the same menu. The tags are visually represented in the white bar above the video manipulation controls, and listed in the tree view below the Options panel (see Figure 4.6). The events can be exported as images or files (requirement 2.4), allowing to save only the data cor- responding to relevant movements and discard uneventful data. This feature is especially useful for long term monitoring (several hours or days corresponding to several gigabytes of data), in which movements of interest only occur sporadically.
4.3 Summary
With the main aim of developing a solution for automated gait analysis, we relied on an RGB-D camera to obtain quantitative information regarding the movements of a given subject. The selected camera was the Microsoft Kinect, which provides both colour (RGB) and/or IR images and depth information. Furthermore, it also provides the 3-D position of several body joints, which is essential for our gait analysis solution. Therefore, we developed the KiT (KinecTracker) software
application for enabling the online visualization and acquisition of the multimodal data provided by the Kinect (v1 or v2). The data acquired with KiT can be reviewed in our KiMA (Kinect Motion Analyser) application, which also allows the indication and/or exportation of instants or events of interest for further analysis.
KiT has been used in the last few years for acquiring data in different scenarios: gait analysis in healthy subjects, Parkinson’s disease (PD) patients and Transthyretin Familial Amyloid Polyneu- ropathy (TTR-FAP) patients; and seizure analysis in epileptic patients. KiMA was used to manually identify gait events in studies using the data acquired from PD patients and/or healthy subjects [79- 84] and TTR-FAP patients [85]. It was also used in studies with epilepsy patients to select data cor- responding to relatively short and unpredictable seizure events in data acquisitions lasting several hours or days [86, 87].