Short-term scene analysis

This page describes an example analysis sequence for a specific scene over a relatively short period of time.

Example 1: I want to graph the number of steps taken on a weekday.

Generally, most people spend their weekdays in the following cycle: get up, go to work, work, go home, free time, go to bed. Let’s graph the number of steps taken for those who are concerned about lack of exercise.

Data type

Although both 15-second and 60-second intervals can be used, we will use 60-second interval data that has been summarized to some extent.

Preprocessing

  • wea_s >= 30 (out of 60 seconds, 30 seconds is true for the attachment flag, sometimes the flag is raised at rest, so filter around 50%)
  • tl_xav > -45, tl_xav < 45, tl_yav > -45, tl_yav < 90 (if the angle is not the angle at which the glasses are normally worn, filter because they may be stored without being worn)

Grouping

  • Period to summarize: 15 minute intervals for detailed recall of actions, 60 minute intervals for ease of viewing
  • Aggregate function: sum(stp_fst + stp_mid + stp_slw + stp_vsl)

Example 2: I want to visualize the degree of concentration during a workday.

We often have subjective impressions such as “I got a lot done today,” or “I couldn’t concentrate for some reason”. Let’s look back the day objectively by looking at the actual data.

Data type

We recommend using 15-second interval data for indicators such as concentration.

Preprocessing

  • isl == false (when in the wearing state)
  • vld == true (use only data when blink-related measurement accuracy is sufficient)
  • tl_xav > -45, tl_xav < 45, tl_yav > -45, tl_yav < 90 (If the angle is not the angle at which the spectacles are normally worn, filter the data as they may be stored in a bag or something)

Grouping

  • Period to summarize: 10-minute intervals for mental indicators, because the data can be difficult to understand if the swings of increase/decrease are too large and too frequent.
  • Aggregation function: mean(sc_fcs), or sum(sc_fcs >= 60 ? 1 : 0)/sum(sc_fcs ! = null ? 1 : 0) (take the mean or the rate of the high interval)

Example 3: I want to compare the pitch (number of steps per minute) of various people.

There is an legend that “a person who walks fast is a good worker”. To confirm this legend, let’s calculate the average pitch of the person.

Data type

Although both 15-second and 60-second intervals can be used, we will use the 60-second interval data that has been summarized to some extent.

Preprocessing

In order to keep only the data from the continuous walking, we will exclude other conditions.

  • stp_s >= 50 (we want to extract only the person’s normal walking condition (we want to remove the data from the person who just wandered around the room for a bit), so only the data where the person walked for more than 50 seconds)
  • tl_xav > -45, tl_xav < 45, tl_yav > -45, tl_yav < 45 (If the angle is not the angle at which the person normally wears glasses, we filter out the data because it may be stored without being worn)

Grouping

  • Period to summarize: per user
  • Aggregate function: mean((stp_fst + stp_mid + stp_slw + stp_vsl) * 60 / stp_s) (multiply the total number of steps in those 60 seconds by the rate of seconds walked)