UN ERROR PROVIDENCIAL
6. NUAY Nº SIRVASE PASAR AL Nº 7
However to accurately render the different presentation rates on a standard display is a difficult task.
Maintaining accurate timing and hence an accurate presentation rate is critical for this investigation because then reduction in identification accuracy is highly significant as the presentation rate is increased beyond four tiles per second (see the graphs of this data by Potter1976, p. 514; Metzger and Antes1983, p. 271; Potter and Fox2009, p. 31).
Thestimuliwas presented on a monitor with a refresh rate of 60 Hz, meaning the tile durations can only be presented at 16.7 ms intervals. Therefore for the average tile duration to be accurate, the display duration of each tile was jittered with the pseudo code shown in listing5.1.
Listing 5.1– Pseudo C code showing the jitter algorithm which prints the number of fields for each tile to the console. This example demonstrates the tile frame durations for the case of segmentation degree.
1 float cumulative_error = 0.0f;
2 const float tile_error = 3.0f / 11.0f; // ratio time duration is base 3 const unsigned int tiles = 15;
4 const unsigned int base_fields = 232;
5
6 for (unsigned int tile = 1; tile <= tiles; tile++) { 7 cumulative_error += tile_error;
8
9 if (cumulative_error >= 1.0f) { 10
11 // Standard duration frame
12 fprintf(
13 stdout,
14 "Tile %d has length %d fields.\n",
15 tile,
16 base_fields
17 );
18
19 cumulative_error = cumulative_error - 1.0f;
20
21 } else {
22
23 // Extended duration frame
24 fprintf(
25 stdout,
26 "Tile %d has length %d fields.\n",
27 tile,
28 base_fields + 1
29 );
30
31 }
32 }
Walking through the pseudo code, the example forsegmentationdegree 1 is shown in listing5.1.
The first four lines are calculated for eachsegmentationdegree, and forsegmentationdegree 1, the number of fields to correctly display a tile for the required 3878 ms on a 60 Hz display (16.7 ms per field) is 3878/16.7 = 233, however only an integer number of fields can be displayed, therefore the result is between 232 and 233. Hence each tile is shown for 233 fields (3883 ms) which is 5 ms/tile too fast. Such an error would be compounded over many tiles and result is each sequence at a givensegmentationdegree lasting a different amount of time, in addition to the perceived duration being longer. The solution to this is to periodically apply a correction—
displaying a tile for 232 fields (3867 ms)— correcting the compounded error in tile duration while also distributing the correction through the tile-set.
The equations for calculating the tile error and base fields are shown in equations5.1–5.2, where
⌊x⌋ corresponds to the floor(x) operator and ⌊x⌉ means ‘round x to the nearest integer’. The value of 60 corresponds to the refresh rate of the display system, which for the equipmen t used, operates at 60 Hz.
tile error= 1 − ((60× sequence length
number of tiles ) − ⌊60× sequence length
number of tiles ⌋) (5.1) base fields= ⌊60× sequence length
number of tiles ⌉ (5.2)
The parameters shown in equations5.1–5.2are calculated for each segmentationdegree and continually adjusted during the presentation in the event of any missed fields during the rendering of the sequence.
5.3.3 Interface
By segmenting the image, the resulting tiles are smaller than the fixed-size monitor on which they are presented. Therefore the unused area of this interface is shaded grey to avoid both high-contrast artefacts on the boundary of the tiles and to limit large swings in display luminance that could hinder perception or even induce a blink reflex in subjects observing the rapidly changing visuals. The resulting interface is shown in figure5.2, demonstrating the trade-off between image dimensions and presentation time within the interface design. As with the previous investigation intoSVPin chapter4, imagery from six different fly-past terrain strips were each divided into six differentsegmentationdegrees, which were shown to the human operator at correspondingly faster or slower presentation rates, as shown below each screen-shot on figure 5.2. As before, the six strips of terrain contained 3–5 human beings (ortargets) in three distinct clusters per trial, located pseudo-randomly in both location and time. The original terrain strips weredesigned to ensure thattargetsdid not lie across tile edges.
Thesegmentationdegrees 1–6 successively divide the total image area into smaller and smaller image tiles (one complete screen: 1×1, a quarter of the screen: 2×2, a ninth of the screen: 3×3,1/16 of the screen: 4× 4,1/25of the screen: 5× 5, and finally1/36of the screen: 6× 6). The dimensions of the tiles at the chosensegmentationdegrees are shown in table5.2. In addition to the presentation rate for sixsegmentationdegrees encompassing the range fromSVPto a cognitively challenging display, the dimensions of the tiles range from the full size of the monitor (segmentationdegree 1) to an area approximating the size which would be projected on thefoveaof the subject’s eyes.
Six differentsegmentationdegrees were chosen to explore the dimensions surrounding two different presentation interfaces. The first consists ofsegmentationdegrees 1–3, where the the tile is displayed for almost half a second or longer in addition to the dimensions of the tile being larger than the para-foveal region of the subject’s eye. The second presentation interface consists of segmentationdegrees 3–6 where the presentation rate is of the order of 100 ms with tile dimensions that ensure that the imagery does not expand beyond the para-foveal region of the subject’s eye.
The six different degrees were also chosen for their applicability in a simple experiment design.
(a)1: 15 tiles, 3878 ms/tile (b)2: 60 tiles, 970 ms/tile
(c)3: 135 tiles, 431 ms/tile (d)4: 240 tiles, 242 ms/tile
(e)5: 375 tiles, 155 ms/tile (f)6: 540 tiles, 108 ms/tile
Figure 5.2– From the terrain strip, full-screen examples of the stimuli are shown, complete with the segmentation–presentation time tradeoff
Table 5.2– The tile image size, screen percentage covered and visual angle subtended for given segmentationdegrees.
Segmentation Degree Tiles
Tile Dimensions
Image Size (pixels) Screen (%)1 Visual Angle (○)2,3 Width Height Area Width Width Height
1 15 1024 768 100 100 22.7 17.0
2 60 512 384 25 50 11.3 8.51
3 135 3424 256 11 33 7.58 5.67
4 240 256 192 6 25 5.67 4.25
5 375 2054 1544 4 20 4.54 3.41
6 540 1714 128 3 17 3.788 2.836
1 Values rounded to the nearest percent for clarity.
2 Values shown to three decimal places.
3 Visual angle correct for a 15 inch 1024×768 pixel resolution screen (12×9 inch monitor at 85.3 pixels/inch, 0.298 mm dot pitch) viewed from a distance of approximately 76 cm (30 inches).
4 Values have been rounded up to the nearest integer so that fractions of a pixel are displayed and not hidden.