HoloLens & LED Screen-Based Handwashing Dataset: Quantitative Analysis and User Experience Data

Research Hypothesis, Data Insights, Notable Findings, and Interpretation
Research Hypothesis
The primary hypothesis of this study posits that a mixed reality (MR)-based handwashing activity recognition system, utilizing HoloLens 2 and guided by the World Health Organization (WHO) Handwashing Guidelines (2009), will enhance task performance and user satisfaction compared to traditional desktop-based systems. Specifically, we hypothesize that guided MR conditions will yield higher accuracy in completing the six WHO-defined handwashing steps (palm to palm, palm over dorsum, etc.) and lower cognitive load, as measured by NASA-TLX scores, due to real-time feedback. Additionally, we anticipate that qualitative feedback will reveal improved immersion and usability in MR settings, supported by Igroup Presence Questionnaire (IPQ) and System Usability Scale (SUS) scores.
Dataset Description and Collection
The dataset comprises quantitative and qualitative data from 21 subjects, collected over a 105-minute experimental protocol approved by Nazarbayev University IRB and aligned with the Declaration of Helsinki (2013). It is divided into two categories: a qualitative Excel file and quantitative logs in 21 subject-specific folders (1-21). The qualitative file captures subjective feedback, including SUS (usability), NASA-TLX (cognitive load), IPQ (immersion), and open-ended comments from guided/unguided tasks on desktop and HoloLens 2, organized in a table with columns for variables (e.g., subject ID, task type) and 21 rows. Quantitative logs, stored in "Desktop" and "MR" subfolders, contain time-stamped data (e.g., completion times, movement accuracy) from 4-minute tasks in a 2x2 factorial crossover design (groups A and B), saved as CSV/text files with 10-minute switches including questionnaire data. Data was gathered in four phases: Orientation (15 min for eligibility/consent), Training (20 min for WHO sensitization), Main Experiment (55 min for tasks), and Resting/Data Collection (15 min for surveys), with 30 FPS logging potential.