As mentioned in section 7, the annotations made from the motion capture modality includes the head position of the child, the distance between the faces of the mother and the child and nally the child's physical energy level. All these annotations have been executed on a frame-by-frame basis for each dyad.
The following three sections presents the proles obtained using the three au-tomatic methods for the mocap annotations.
9.3 Motion Capture Annotations 115
9.3.1 Child's Head Position
The angular prole between the mother and the child, as a representation of the child's head orientation, is shown for dyad 002 in gure 9.18(a), and for dyad 010 in gure 9.18(b)for the time interval 0-75 seconds.
(a) (b)
Figure 9.18: Angular proles for the rst 75 seconds of(a)dyad 002, and(b) dyad 010. It is seen that for dyad 010 there are several miss-ing angles around 5 seconds, reectmiss-ing the not-identied mocap markers in Qualisys. Note that the y-axis is not the same in the two gures.
What is seen from gure9.18, showing the angle between the mother and the child, is that in the interval 0-75 seconds, the angle is equal to 0 only once in (a), at around 70 seconds, and several times in(b). The zero angle, represents the time instants where the mother and the child are facing each other.
To relate the raw angles to the scheme of categories introduced in table7.1, the following gure 9.19 shows the above illustrated angle proles in a histogram where each bar represents one of the four categories.
From the gures it is clear that by far the most of the angles belong to the En face category. This represents the angles of 0-30 degrees. In only few of the frames the child's head is averted more than 30 degrees from the mother's which is clear from both 9.19(a) and 9.19(b). This is also the general image when analysing the angle proles.
As explained in chapter 7, the manual codings of the angles determined at Babylab are carried out using a given reference point in the room and not with
(a) (b)
Figure 9.19: Histogram representing the distribution of the 36000 (frame-by-frame) calculated angles between the mother and child with re-spect to the four categories, En face, Minor avert, Major avert and arch. (a) illustrate that of dyad 002, and (b) that of dyad 010. Note that the histogram has been normalized.
respect to the position of the mother as the approach used in this thesis. It has therefore not been possible to validate the angles obtained automatically with the manually coded angles.
9.3.2 Distance Between Faces
The distance between the mother's and child's faces for dyad 002 and 010 are illustrated in gure 9.20for 0-75 seconds.
As for the angle prole, it can be observed for dyad 010, in9.20(b), that some not-identied markers are present around 5 seconds.
As mentioned in the introduction of chapter7the distance between the faces of the mother and the child is calculated in Excel by the psychologists at Babylab.
No validation of the distance proles obtained in this thesis with respect to the ones calculated by Babylab has been carried out. The reason for this is that the distance proles are unique, meaning that they only have one solution. The only dierence is the actual distance which will dier from the ones in this thesis to those of Babylab, because they at Babylab use the back markers MheadB and CheadB, see gure3.1, for the calculations. In this study the MheadM and CheadM are used, see gure7.2, since this is thought of as a good representation of the head positions. Despite this dierence, the relation between the distance of each frame remains constant.
9.3 Motion Capture Annotations 117
(a) (b)
Figure 9.20: Distance between the head of the mother and child for the rst 75 seconds of(a)dyad 002, and(b)dyad 010. It is seen that for dyad 010 there are several missing distances around 5 seconds, reecting the not-identied mocap markers in Qualisys. Note again the dierence of the y-axis between the two gures.
9.3.3 Child's Physical Energy Level
Figure 9.21shows the distance prole for dyads 002 and 010 for 0-75 seconds, based on the method of calculation shown in section7.3.
It can be observed for dyad 010, in9.20(b), that many not-identied markers are present around 15 seconds, from 50-55 seconds, around 60 seconds and again from 65 seconds and up to 75 seconds.
As was the case for the calculations of the head distance above, it was men-tioned in the introduction of chapter 7that the child's physical energy level is calculated in Excel by the psychologists at Babylab. The energy proles ob-tained in this thesis has not been validated with respect to the ones calculated by Babylab with the same reasoning as above.
9.3.4 Summary
The motion capture annotations of the head position of the child, the distance between the faces of the mother and the child and nally the child's physical
(a) (b)
Figure 9.21: The child's physical energy level for the rst 75 seconds of (a) dyad 002, and (b)dyad 010. It is seen that for dyad 010 there are several missing distances around 15 seconds, from 50-55 sec-onds, around 60 seconds and again from 65 seconds and up to 75 seconds. These time intervals reect the not-identied mo-cap markers in Qualisys. Note again the dierence of the y-axis between the two gures. This reects the dierence in physical energy level of the child in the two recording sessions.
energy level have all been calculated and illustrated in this minor section. De-spite their own calculation methods, the psychologists at Babylab have shown interest in the annotation methods obtained in this thesis, which could be be-cause of the more automated, and thereby less time-consuming, approach used here.
The child's head position is calculated with respect to the mother, as explained in7.1, as opposed to a reference point in the room as in the approach of Babylab.
Therefore it has not seemed appropriate to compare the resulting annotations of the two methods.
The resulting distance proles as well as those of the child's physical energy level have not been validated either, due to the unique solutions of these when calculating them from the marker coordinates.