CASAS Smart Home dataset - scripted complex activities, activity scores, and cognitive diagnosis

We hypothesize cognitive impairment can be evident in everyday task performance. We also postulate that differences in task performance can be automatically detected between cognitively healthy individuals and those with dementia and mild cognitive impairment (MCI) using smart home and ubiquitous computing technologies. This dataset contains ambient sensor readings collected in the CASAS smart apartment testbed at Washington State University for 179 participants with corresponding cognitive diagnoses.

Data are collected continuously from ambient sensors while participants perform 24 scripted activities. Each sensor reading is reported on a separate line and is described by fields date, time, sensor, and message. The first 8 activities are performed without cues (task step reminders), the second 8 activities are performed with cues when needed, and the last 8 activities, part of a complex day out task, are interwoven in a natural manner without cues. The task list is included in the file activitylist.txt.

The file activityscores.txt provides numeric values for activities 1-8 that are based on experimenter assessment of task quality. The file also gives scores for the day out task. Finally, the file diagnosis.txt lists the diagnosis for each participant, coded as the following:

1 = dementia
2 = MCI
3 = middle age 45-59
4 = young-old 60-74
5 = old-old 75+
6 = other medical
7 = watch/at risk - follow longitudinally
8 = younger adult
9 = younger adult, English second language
10 = diagnosis not available

The sensors are categorized (and named) as:

• M01 - M51: PIR motion detectors (ON when detected motion starts and OFF when it stops)
• I01 - I10: item use sensors (PRESENT or ABSENT indicating item is on sensor or not)
• D01 - D019: door sensor on cabinets and doors (OPEN or CLOSE)
• P001 and P002: current eletricity consumption
• T001 - T006: ambient temperature sensors
• BATP and BATV: sensor battery levels