Pitch-based sonification of quantitative data increases the accessibility of data visualizations that are otherwise inaccessible for blind and low-vision (BLV) individuals. We argue that, although pitch representations can reveal the coarse-grained information of data, such as data trend and value comparison, they cannot effectively convey the fine-grained details like the sign and exact value of individual data points. Informed by existing sound perception research, we propose a spatial audio-based approach by representing data values as the sound direction in the azimuth plane to achieve accessible fine-grained data representation. We conducted a user study with 26 participants (including 10 BLV participants) on four data perception tasks. The results show our approach significantly outperforms pitch representation on fine-grained data perception tasks like recognizing data signs and exact values, and performs similarly on data trend identification, despite its inferior accuracy on data value comparison.
Figures & tables
Figure 1: Illustration of data encoding in SpaudioData . The encoding space forms a semicircle in front of the participant, spanning 180∘ . Sounds from the left, straight ahead, and right represent data -10, 0, and 10 respectively. Negative values are encoded as leftward spatial audio, while positive values are encoded as rightward. The angular interval is 9∘ .
Figure 2: Pilot Study Results. The mean error for the 9∘ interval angle was smaller than that observed for the 3∘ configuration. For the 9∘ condition, errors for the one-playback and three-playback conditions were similar and lower than those in the five-playback condition.
ID
Gender
Age
Vision Level
Education
Exp.
P1
Male
26
Sighted
Master
None
P2
Female
23
Sighted
Bachelor
None
P3
Male
25
Sighted
Bachelor
None
P4
Male
28
Sighted
Master
Train
P5
Female
28
Sighted
Master
None
P6
Female
24
Sighted
Bachelor
Train
Table 1: Details of Participants (“Exp.” refers to the participant’s musical training experience).
Figure 3: Results of four tasks for BLV participants. Statistical significance is indicated by stars, where ∗∗∗ denotes p<0.001 , ∗∗ denotes p<0.01 , and ∗ denotes p<0.05 , based on the Wilcoxon rank-sum test.
Figure 4: Performance comparison between SpaudioData and the pitch representations for sighted participants across four tasks, showing consistent results with the overall participant group, with notably high accuracy and low variance in the data trend recognition tasks using SpaudioData .
Figure 5: Distribution of accuracy by different gaps for data value comparison task. For the Comparison, Trend, Sign, and Exact Match tasks, higher values indicate better performance, whereas for the Value Difference task, lower values indicate better performance.
Figure 6: Distribution of accuracy by different gaps for data trend recognition task.
Figure 7: Accuracy distribution across ground truth values for BLV participants.
Figure 8: Accuracy distribution across ground truth values for sighted participants.
Augmented Human Lab, Dept. of Computer Science, National University of Singapore · Augmented Human Lab, National University of Singapore · Saarland University, Saarland Informatics Campus Saarbrücken, Germany +2