3. Results

The Dependent variables in this experiment were a) object recognition for the first and second time subjects completed the experiment b) multiple choice comprehension for the first and second time subjects completed the experiment c) content analysis for sentences written by subjects describing the gist of the video clips and d)evaluation scales for the number and speed of videos (1-7 Likert scales).

3.1 Object Recognition, Multiple Choice Comprehension Questions

Nonparametric statistics test hypotheses about data that do not meet parametric test assumption that the underlying distribution of the data is normally distributed. Nonparametric tests employ alternatives such as sequentially ranking observations from all groups or comparing two groups observation by observation to test hypotheses. These are unlike parametric tests that estimate the parameters of normally distributed data, then perform a computation on these estimates. Because the data in the video c omprehension questions and object recognition questions are not normally distributed and do not meet the assumptions of the parametric tests (ANOVA), a nonparametric test (Kruskal-Wallis) is used for the analysis.

The Kruskal-Wallis test is a nonparametric equivalent of a one-way analysis of variance by ranks. In other words, it tests the hypothesis that two or more groups all come from the same distribution against the alternative that at least one of the group s comes from a different distribution. It is calculated as a regular ANOVA, but instead it uses the ranks of the data. Included with the test statistic is a table displaying the mean rank for each of the groups that can be used to assist in determining w hich group tends to have larger values than the others.

Using the Kruskal-Wallis Test, a significant difference was found between conditions 1 through 4 for object recognition after watching the scrolling key frames one time through. H(3)=15.96, p<0.001. The rank information is listed in Table 1. The mean r ank for condition 1 is listed as the highest indicating that condition 1 performed better on the object recognition tasks than conditions 2, 3, and 4. Conditions 2 and 3, having almost the same mean rank, both perform better than condition 4.

Table 1: Kruskal-Wallis Rank Information for Object Recognition (first time through). Condition number is listed on the left side of the table. Count is equal to 7 indicating n=7 for each condition.

For object recognition data for the second time through the experiment, the Kruskal-Wallis Test again showed a significant difference between the four conditions. H(3)=12.74, p<0.005. The rank information is listed in Table 2. The mean rank for conditi on 1 is listed as the highest indicating that condition 1 performed better on the object recognition tasks than conditions 2, 3, and 4. Conditions 2 and 3, having almost the same mean rank, both perform better than condition 4. Results are almost identical in performance for the first time through the experiment.

Table 2: Kruskal-Wallis Rank Information for Object Recognition (second time through). Condition number is listed on the left side of the table. Count is equal to 7 indicating n=7 for each condition.

There were no significant differences found between conditions for the multiple choice comprehension questions (first time through). Kruskal-Wallis revealed H(3)=3.88, p<0.25. The rank information is listed in Table 3. The mean rank for all conditions appear approximately the same indicating that subjects in all conditions performed equally well on the comprehension questions.

Table 3: Kruskal-Wallis Rank Information for Multiple Choice Comprehension (first time through). Condition number is listed on the left side of the table. Count is equal to 7 indicating n=7 for each condition.

There were also no significant differences found between conditions for the multiple choice comprehension questions (second time through). Kruskal-Wallis revealed H(3)=.591, p<0.90. The rank information is listed in Table 4. Results are almost identic al in performance for the first time through the experiment. The mean rank for all conditions are approximately the same indicating that subjects in all of the conditions performed equally well on the comprehension questions.

Table 4: Kruskal-Wallis Rank Information for Multiple Choice Comprehension (second time through). Condition number is listed on the left side of the table. Count is equal to 7 indicating n=7 for each condition.

3.2 Sentence Analysis

The following table, Table 5, contains data for the content analysis of sentences written by subjects describing what they perceived was the gist of the videos. The content of the sentences was broken into four different types of responses, each about a different aspect of the video that was mentioned: People, Objects, Actions/Concepts, and Places. The key words found in the list were determined by subject responses. It is possible for a subject to have listed multiple key words/concepts for each aspec t. All concepts/objects/people/actions/places mentioned by the subjects were included in the tables listed below. The analysis reveals that subjects in conditions 4 and 3 were not able to correctly grasp the meaning of the videos depicted through the use of the key frames with as much accuracy as those in conditions 1 and 2.

Table 5. Frequency Counts for Each Aspect Listed by Video Clip. For each condition, n=7. Key words extracted from the sentences are listed in the left-most column. Counts for each condition are listed separately.

No places were mentioned in Video Clip 4.

3.3 Evaluation

Means and standard deviations are given in Table 6, Figure 1 gives an overview.

The mean values for each question asked in the evaluation of the interface show that condition 1 rated the speed of the video key frames for the object recognition task was somewhere "in the middle" between too fast and too slow. Conditions 2, 3, and 4 rated the speed closer towards "too fast". Even though all the video key frames are displayed at the same rate, subjects perceive the videos as going "too fast" in the conditions with two, three and four videos simultaneously.

Results for the speed of videos for the comprehension task are similar to those found for the speed question for object recognition. Subjects in condition 1 perceived the speed of video key frames as somewhere between "the middle" and "too fast". Cond itions 2, 3, and 4 rated the speed closer towards "too fast". Subjects perceive the videos as going "too fast" in the conditions with two, three and four videos simultaneously because of the difficulty of comprehending several videos at once.

For both the object recognition tasks and the comprehension tasks, as the number of video screens shown increased, subjects rated the "number of videos" more towards the "imperceptible" end of the scale.

Table 6. Means and Standard Deviations for Evaluation Questions.

Figure 1. Results of evaluation questions. Error Bars: (+/-) 1 Standard deviation

All comments made by subjects can be found in Appendix B.

Continue
Return to Comprehension and Object Recognition Capabilities for Presentations of Simultaneous Video Key Frame Surrogates