null
Nodes | |
Head orientation
Group
|
Orientation is coded using an external element in the frame of reference, namely that of the interlocutor (the person addressed in the communication). So the body part is coded as either facing the interlocutor, or faced away from the interlocutor. Orientation is coded continuously and is body part specific. |
Head posture
Group
|
General definition posture: A posture represents the general alignment of one or a set of articulators (head, trunk, arms) to a particular resting configuration, which shows periodic changes known as posture shifts. Posture shifts (labelled as posture transitions) are movements associated with positioning the body.
Postures are different from actions because a) postures are less subject to frequent change and thus have larger durations, b) postures are robust (small movements do not change or distort the posture), c) whereas actions may or may not be displayed, the body is continually in one or another postural alignment. This means that when a body part is not involved in an action, it is always in a particular posture (but not vice versa, see remark below). Head posture units (-PU) are defined for each direction (e.g. left, right) and anatomical articulation (e.g. turn, tilt) separately. Direction is coded using an anatomical frame of reference, which is independent of any external observer. The direction of the movement is respective to three orthogonal planes that are defined relative to the anatomical standard position of the body (following kinematic standards). The direction coded for postures (e.g. head to left) refers to the end position. Segmentation rules: A posture unit (-PU) is segmented in a transition and configuration phase. The posture transition phase (-PT) (posture shift) is the time segment where the articulator performs the movements for obtaining the end position. During the configuration phase (-PC) the end position of the coded articulator is maintained for the coded direction. This does not imply the obtained posture is static. The transition onset is the starting point of the movement needed to achieve the end position, or the starting frame of the video when the movement onset is cut off. The transition offset is the time point where the transition described in the particular category is ended, or the last frame of the video if the movement offset is cut from the video. The frame following the transition offset is coded as the onset of the posture configuration. The offset of the configuration phase is the time point when the coded end position is broken off by the beginning of a new positioning movement or by an action behavior of same body part. Note: the transition phase can overlap with the retraction of an action. Remark. Not all action behaviors interrupt an ongoing position. It is thus possible that a position of a body part (e.g. head forward) is not broken off by an action by that body part (lateral head shake). When the end of the posture is cut off by the video boundary, the last frame has to be coded as the offset. |
Trunk orientation
Group
|
Orientation is coded using an external element in the frame of reference, namely that of the interlocutor (the person addressed in the communication). So the body part is coded as either facing the interlocutor, or faced away from the interlocutor. Orientation is coded continuously and is body part specific. |
Trunk posture
Group
|
General definition posture: A posture represents the general alignment of one or a set of articulators (head, trunk, arms) to a particular resting configuration, which shows periodic changes known as posture shifts. Posture shifts (labelled as posture transitions) are movements associated with positioning the body.
Postures are different from actions because a) postures are less subject to frequent change and thus have larger durations, b) postures are robust (small movements do not change or distort the posture), c) whereas actions may or may not be displayed, the body is continually in one or another postural alignment. This means that when a body part is not involved in an action, it is always in a particular posture (but not vice versa, see remark below). Definition of trunk: The trunk, or torso, is the upper body excluding the head, neck and upper limbs. It includes the thorax (chest) and abdomen (belly). The thorax extends from the neck to the diaphragm, and does not include the upper limbs. The abdomen constitutes the part of the body between the thorax and pelvis (hip area). Remark: the shoulder belongs to the category of upper limbs (arms). Trunk posture units (-PU) are defined for each direction (e.g. left, right) and anatomical articulation (e.g. lean, rotation) separately. Direction is coded using an anatomical frame of reference, which is independent of any external observer. The direction of the movement is respective to three orthogonal planes that are defined relative to the anatomical standard position of the body (following kinematic standards). The direction coded for postures (e.g. trunk lean to left) refers to the end position. Segmentation rules: A posture unit (-PU) is segmented in a transition and configuration phase. The posture transition phase (-PT) (posture shift) is the time segment where the articulator performs the movements for obtaining the end position. During the configuration phase (-PC) the end position of the coded articulator is maintained for the coded direction. This does not imply the obtained posture is static. The transition onset is the starting point of the movement needed to achieve the end position, or the starting frame of the video when the movement onset is cut off. The transition offset is the time point where the transition described in the particular category is ended, or the last frame of the video if the movement offset is cut from the video. The frame following the transition offset is coded as the onset of the posture configuration. The offset of the configuration phase is the time point when the coded end position is broken off by the beginning of a new positioning movement or by an action behavior of same body part. Note: the transition phase can overlap with the retraction of an action. Remark. Not all action behaviors interrupt an ongoing position. It is thus possible that a position of a body part (e.g. head forward) is not broken off by an action by that body part (lateral head shake). When the end of the posture is cut off by the video boundary, the last frame has to be coded as the offset. |
Whole body posture
Group
|
General definition posture: A posture represents the general alignment of one or a set of articulators (head, trunk, arms) to a particular resting configuration, which shows periodic changes known as posture shifts. Posture shifts (labelled as posture transitions) are movements associated with positioning the body.
Postures are different from actions because a) postures are less subject to frequent change and thus have larger durations, b) postures are robust (small movements do not change or distort the posture), c) whereas actions may or may not be displayed, the body is continually in one or another postural alignment. This means that when a body part is not involved in an action, it is always in a particular posture (but not vice versa, see remark below). Whole body posture units (-PU) represent postural movement of the whole body caused by articulation of the lower body, but the articulators that are responsible for the movement, i.e. hips, leg and/or feet, are not specified. Whereas only the upper body moves in trunk articulation (leaning, rotation), these categories involve leaning behavior or actual displacement of the body caused by the lower limbs. Whole body posture units are defined for each direction (e.g. left, right) separately. Direction is coded using an anatomical frame of reference, which is independent of any external observer. The direction of the movement is respective to three orthogonal planes that are defined relative to the anatomical standard position of the body (following kinematic standards). The direction coded for postures (e.g. whole body lean to left) refers to the end position. <#comment>#comment> Segmentation rules: A posture unit (-PU) is segmented in a transition and configuration phase. The posture transition phase (-PT) (posture shift) is the time segment where the articulator performs the movements for obtaining the end position. During the configuration phase (-PC) the end position of the coded articulator is maintained for the coded direction. This does not imply the obtained posture is static. The transition onset is the starting point of the movement needed to achieve the end position, or the starting frame of the video when the movement onset is cut off. The transition offset is the time point where the transition described in the particular category is ended, or the last frame of the video if the movement offset is cut from the video. The frame following the transition offset is coded as the onset of the posture configuration. The offset of the configuration phase is the time point when the coded end position is broken off by the beginning of a new positioning movement or by an action behavior of same body part. Note: the transition phase can overlap with the retraction of an action. Remark. Not all action behaviors interrupt an ongoing position. It is thus possible that a position of a body part (e.g. head forward) is not broken off by an action by that body part (lateral head shake). When the end of the posture is cut off by the video boundary, the last frame has to be coded as the offset. |
Arms posture
Group
|
General definition posture: A posture represents the general alignment of one or a set of articulators (head, trunk, arms) to a particular resting configuration, which shows periodic changes known as posture shifts. Posture shifts (labelled as posture transitions) are movements associated with positioning the body.
Postures are different from actions because a) postures are less subject to frequent change and thus have larger durations, b) postures are robust (small movements do not change or distort the posture), c) whereas actions may or may not be displayed, the body is continually in one or another postural alignment. This means that when a body part is not involved in an action, it is always in a particular posture (but not vice versa, see remark below). Arms and hand posture units (-PU) are coded in a number of general categories and exclude the particular hand-fingers configuration. In addition, shoulder posture units (-PU) are defined for each direction (e.g. left, right) separately. Direction is coded using an anatomical frame of reference, which is independent of any external observer. The direction of the movement is respective to three orthogonal planes that are defined relative to the anatomical standard position of the body (following kinematic standards). The coded postures (e.g. arms crossed, shoulder up) always refer to the end position. Finally, the configuration of the resting position of the arms (including shoulders) is coded as symmetrical or asymmetrical. Segmentation rules: A posture unit (-PU) is segmented in a transition and configuration phase. The posture transition phase (-PT) (posture shift) is the time segment where the articulator performs the movements for obtaining the end position. During the configuration phase (-PC) the end position of the coded articulator is maintained for the coded direction. This does not imply the obtained posture is static. The transition onset is the starting point of the movement needed to achieve the end position, or the starting frame of the video when the movement onset is cut off. The transition offset is the time point where the transition described in the particular category is ended, or the last frame of the video if the movement offset is cut from the video. The frame following the transition offset is coded as the onset of the posture configuration. The offset of the configuration phase is the time point when the coded end position is broken off by the beginning of a new positioning movement or by an action behavior of same body part. Note: the transition phase can overlap with the retraction of an action. Remark. Not all action behaviors interrupt an ongoing posture. It is thus possible that a posture of a body part (e.g. one arm holds other in front) is not broken off by an action by that body part. When the end of the posture is cut off by the video boundary, the last frame has to be coded as the offset. |
Gaze
Group
|
Gaze direction is described from an external reference frame, namely the interlocutor standing in front. Gaze is coded continuously. |
Action form
Group
|
Actions are discrete units of body movement which are not part of body positioning. A body action unit (-AU) is a local excursion of one or a set of articulators (head, trunk, arm, hand) outside a previously held posture configuration and always returns to that or another posture configuration (e.g., head shake, pointing arm gesture). Contrary to posture units, action units occur and change more frequently, and have a very discrete onset (start point), a relatively short duration and a distinct offset (end point). These body actions are performed by the head, shoulders, trunk, arms and legs and involve activities such as nodding, shrugging, gesturing, scratching , kicking etc. The specific form and dynamics of an action indicate that the movement cannot be considered as the taking of a new posture (form: direction, path, manner; dynamics: no holding of a particular configuration). To code a head, trunk or arm movement as an action, the dynamics and form of the movement (e.g. head nodding, arm gesticulation, scratching) should thus exclude the possibility of simple posture changing. When this necessary condition for action behavior is not fulfilled, the movement should be coded as a postural movement. Segmentation rules: The onset of an action unit is the time-point when the articulator moves away from the current resting position. The offset of an action unit is the time-point when the articulator has returned to a (resting) position (the initial or a new position). Remark: The returning phase (retraction) may overlap with postural movement. Action units are segmented into a sequence of discrete subunits according to differences in the form properties of the movement. The boundary between two subunits is defined by a discontinuous change (a break or a hold) of the movement direction as specified in the attributes (based on Kita, 1998). When the action shows no discontinuous change in any of the attributes , the onset is simply the time when the head moves away from its position and the offset is the time when the head has returned to a position (the initial or a new position). The direction of an action subunit refers to the movement itself and is coded from its starting point or current position, not relative to the anatomical reference frame. Repetitive movement is not segmented or coded for each repeated cycle separately. |
Lower limbs
Group
|
--- |
Action function
Group
|
Actions are discrete units of body movement which are not part of body positioning. A body action unit (-AU) is a local excursion of one or a set of articulators (head, trunk, arm, hand) outside a previously held posture configuration and always returns to that or another posture configuration (e.g., head shake, pointing arm gesture). Contrary to posture units, action units occur and change more frequently, and have a very discrete onset (start point), a relatively short duration and a distinct offset (end point). These body actions are performed by the head, shoulders, trunk, arms and legs and involve activities such as nodding, shrugging, gesturing, scratching , kicking etc. Actions are coded on a functional level that is parallel to the anatomical and form description level. Functional BAP codes are the three behavioral classes identified by Ekman and Friesen (1972): emblem, illustrator (including two subtypes, deictic and beat) and manipulator. An action can be multifunctional (eg. iconicity combined with deixis) and can thus belong to several categories in different degrees. Each functional unit is therefore coded on an ordinal (5-point) saliency scale in terms of duration, location and execution. The salience scale ranges from 5 as very pronounced or salient and 1 as very subtle (0 is in practice never scored because in this case the function is absent and thus not coded). For example, a functional unit is coded as very subtle (“1”) when it is executed with minor articulation (e.g. deictic pointing with a partly extended finger), for a very short period of time (e.g. few repetitions of a beat), or located in the peripheral movement space of the person (e.g. below the hips). On the other extreme, a functional unit is coded as very pronounced (“5”) when it is supported by maximum articulation (e.g. deicitic pointing with fully extended finger and arm), stretched in time over several seconds (many repetitions of a beat), or located in the central movement space of the person (e.g. near the face). The onset is coded when the movement starts to perform the coded function, the offset is coded when the movement stops performing that function. |