annotation_overlap: calculate overlap between two binary annotation sets
In gobbios/avutils: various utilities for DiViMe tools and audio processing

Description Usage Arguments Details Value Examples

calculate overlap between two binary annotation sets

1 2	annotation_overlap(x1, x2, seg_length = NULL, overlap_res = 100, return_raw = FALSE)

`x1`	data.frame or matrix with annotations, needs columns named `"start"` and `"end"`
`x2`	data.frame or matrix with annotations, needs columns named `"start"` and `"end"`
`seg_length`	numeric, total duration of the data (default is `NULL`, i.e. it's guessed from the data), see details
`overlap_res`	numeric, resolution for overlap (default is `100`), see details
`return_raw`	logical, by default `FALSE`, i.e. return the proportion. Otherwise return a 0/1 vector where 1 indicates a congruent comparison.

The function works like this. The total duration is segmented into overlap_res segments. The midpoints of these segments serve as test values. For example, if the resolution is 100 and the duration is 60 seconds, we will obtain 100 segments of 0.6 seconds each. The test values lay in the center of each segment, i.e. at 0.3, 0.9, 1.2 etc. Now, for each of these test values, check in x1 and x2 whether an annotation exists.

Also, this function is more of a 'helper' function, that primarily is used inside the evaluate_sad function.

a numeric value of overlap proportion, or a 0/1 vector.

x1 <- data.frame(start = c(1, 4, 7), end = c(2.3, 4.2, 8))
x2 <- data.frame(start = c(0, 5), end = c(4, 6))
annotation_overlap(x1, x2)
annotation_overlap(x1, x2, seg_length = 60)
# one annotation set is empty
x3 <- x2[0, ]
annotation_overlap(x1, x3)
# with 'real' data
x1 <- read_rttm(system.file("noisemesSAD_synthetic_speech.rttm", package = "avutils"))
x2 <- read_elan(system.file("synthetic_speech.eaf", package = "avutils"))
annotation_overlap(x1, x2, overlap_res = 10)
annotation_overlap(x1, x2, overlap_res = 10, return_raw = TRUE)