Skip to contents

Per-Sensor Statistics Over Time

Source: R/sensor-map.R
pr_sensor_map.Rd

Reduces a recording along the time axis: one number per sensor, with the sensor's grid coordinates attached. This is the counterpart of pr_frame_metrics(), which reduces along the sensor axis instead.

Usage

pr_sensor_map(
  trial,
  statistic = c("mean", "max", "sd", "pti", "pct_zero"),
  threshold = 0
)

Arguments

trial

A pr_trial object.

statistic

Character. One of "mean", "max", "sd", "pti" or "pct_zero". See Details.

threshold

Numeric. Cells at or below this value count as unloaded. Default 0.

Value

A tibble::tibble with one row per sensor and columns sensor (integer), row, col (integer grid indices) and value (numeric).

Details

The available statistics, for a sensor read at every frame:

  • "mean" — mean pressure, unloaded frames included in the denominator.

  • "max" — the maximum pressure picture (MPP): the peak this sensor ever saw. pr_sensor_map(trial, "max")$value is the map pr_plot_heatmap() draws.

  • "sd" — standard deviation over frames (NA for a single-frame trial), i.e. how much this sensor's load fluctuated.

  • "pti" — pressure-time integral, trapezoidal over the trial's own timestamps; identical to pr_calc_pti() at the default threshold.

  • "pct_zero" — fraction of frames in which the sensor was at or below threshold. 1 means never loaded, 0 means always loaded; it is the per-sensor duty cycle, and 1 - pct_zero is the fraction of the recording the sensor was in contact.

threshold treats cells at or below it as unloaded, exactly as in pr_frame_metrics(): they are zeroed before mean, sd and pti are taken, they cannot raise max, and they are what pct_zero counts. At the default threshold = 0 on non-negative data no masking is needed and none is done.

The sensor column is the pressure column index, taken positionally from trial$layout$coords_mm, not the device channel number. For a layout built by pr_layout_from_index_map() the two differ, and pr_channel_order(trial$layout)[sensor] recovers the device channel. Rows come back in that same column-major grid order, so $value can be used as a feature vector directly.

See also

pr_frame_metrics() for the per-frame reduction, pr_profile_matrix() to stack this across a dataset.

Other sensor aggregation functions: pr_batch_frame_summary(), pr_profile_matrix()

Examples

trial <- pr_example_trial("saddle_horse")
mpp <- pr_sensor_map(trial, "max")
mpp[which.max(mpp$value), ]
#> # A tibble: 1 × 4
#>   sensor   row   col value
#>    <int> <int> <int> <dbl>
#> 1     68     4     5  18.4

# The mean map is never above the peak map, sensor for sensor:
avg <- pr_sensor_map(trial, "mean")
all(avg$value <= mpp$value + 1e-9)
#> [1] TRUE

# Fraction of the recording each sensor spent in contact:
duty <- 1 - pr_sensor_map(trial, "pct_zero")$value
round(range(duty), 3)
#> [1] 0.462 1.000