Jordan J Bern
College of Information, University of Maryland, College Park MD 20740, USA.
Univers Access Hum Comput Interact (2025). 2025;15780:85-98. doi: 10.1007/978-3-031-93848-1_7. Epub 2025 Jun 3.
People with photosensitive epilepsy (PSE) may have seizures caused by flashing lights, particular video sequences, and bold standing patterns. To reduce the risk from video and other content, five international guidelines have been created that outline the characteristics of flashing that can cause harm when taken together: large areas, intense changes, and a range of flash rates. Automated tools can evaluate video sequences according to these standards. However, there are few reference videos available to validate and compare such automated PSE tools. Furthermore, the test videos that have been reported to date do not thoroughly test PSE algorithms' adherence to standards. This article describes some of the gaps in current testing paradigms and proposes additional validation tests for PSE hazard analysis tools. We have also created a new tool that can be used to generate ground-truth video sequences for validating and comparing PSE safety algorithms. This tool can create videos that extend beyond the simple alternating-frames test videos typically used to date. Additionally, more complicated test videos can be generated to test more nuanced characteristics of the standards or to test algorithms' susceptibility to confounding-but-harmless flashing.
患有光敏性癫痫(PSE)的人可能会因闪烁的灯光、特定的视频序列和鲜明的静止图案而引发癫痫发作。为降低视频及其他内容带来的风险,已制定了五项国际准则,概述了闪烁综合起来可能造成危害的特征:大面积、强烈变化以及一系列闪烁频率。自动化工具可根据这些标准评估视频序列。然而,用于验证和比较此类自动化PSE工具的参考视频很少。此外,迄今为止报告的测试视频并未全面测试PSE算法对标准的遵循情况。本文描述了当前测试范式中的一些差距,并为PSE危害分析工具提出了额外的验证测试。我们还创建了一种新工具,可用于生成用于验证和比较PSE安全算法的真实视频序列。该工具可以创建超越通常使用的简单交替帧测试视频的视频。此外,可以生成更复杂的测试视频,以测试标准中更细微的特征,或测试算法对混淆但无害闪烁的敏感性。