Adding Annotations
adding_annotations.RmdTypically you have raw onset data when loading files with
onsetsync which might not any other information within them
(beat subdivisions, etc.). Here is an example of how you can add
reference beat subdivisions and calculate reference beats such as mean
of all onsets in a sub-division.
Load raw onsets from OSF
library(onsetsync)
require(httr)
CSS_Song2_Onset <- get_OSF_csv('8a347') # Get onsets
print(head(CSS_Song2_Onset[,1:8,]))## Piece Label.SD SD Clave_. Section Isochronous.SD.Time Clave Bass
## 1 Song_2 1|1 1 1 Son 5.037333 NA NA
## 2 Song_2 1|2 2 N Son 5.260063 NA NA
## 3 Song_2 1|3 3 N Son 5.482792 NA NA
## 4 Song_2 1|4 4 2 Son 5.705521 NA 5.714555
## 5 Song_2 1|5 5 N Son 5.928250 NA 5.927078
## 6 Song_2 1|6 6 N Son 6.150979 NA NAThe data frame shows a portion of the onset data structure where the
last three columns refer to the onset times of the specific instruments
(Clave, Bass, and Guitar). Note
that we only show the first eight columns and 6 rows. The first five
columns are meta-data, referring to the piece, beats
and cycles and sections. Label.SD combines cycle and beat
information (first integer refers to cycle and the second integer refers
to beat sub-division). SD represents the
sub-division explicitly. In this music there are 16
subdivisions per cycle (which can be felt as 4 beats \(\times\) 4 subdivisions).
Section indicates the specific part of the piece. Finally,
in this example we have Clave_ column, which is a reference
to a clave pattern: strokes of the clave (woodblocks), whether sounded
or not, are noted here; there are normally 5 strokes per cycle. The
onset times for the instruments are recorded in seconds, although we
prefer to use milliseconds (ms) in the analyses of synchrony to avoid
reporting unnecessarily many digits. The onset times of the instruments
are the result of an automatic onset detection (that has been subjected
manual check) and the exact technique may depend on instrument used and
source of data.
Add annotations for cycles
CSS_Song2_Metre <- get_OSF_csv('4cdpr') # Annotations for cycles
print(head(CSS_Song2_Metre))## Cycle Time
## 1 1 5.037333
## 2 2 8.601000
## 3 3 12.123000
## 4 4 15.672000
## 5 5 19.174500
## 6 6 22.686000Let’s simplify this example (select only few columns).
CSS_Song2_Onset <- dplyr::select(CSS_Song2_Onset,
Label.SD,SD,Clave,Bass,Guitar,Tres) In the second data frame (CSS_Song2_Metre), shown in
Table 2, we have annotations, namely the cycle times, which are the
manually-estimated times when the cycles start (containing all
sub-divisions of the beat). For instance, here the first cycle starts at
5.037333, the second cycle at 8.601000) and so on. This is important
information for some type of analyses and often we want to combine the
automatically detected onset times and the annotated cycles. It is
important to realise that the information about the cycles comes from an
annotation of the music by an expert (or in some cases by several
experts). To get an annotation such as this, software such Sonic
Visualiser can be used to tap with the perceived beats of the audio
track to obtain the beat times and then the first beat of the cycle is
manually labelled to the data. The sub-divisions are then inferred from
the cycle beginnings. This annotation process requires expertise with
the musical style in question as the same patterns of onsets could
potentially give rise to multiple interpretations of the metre.
Combine onsets and annotations
As the onsets and annotations are in different files, we first
combine the onsets and annotations with onsetsync. Here we
first add annotations about the cycles into the onset data and then we
also add isochronous beat times to the data frame, since are useful
reference points for future calculations. The metre annotations have
already informed the allocation of onsets to beat positions; here we add
more information from the metre annotations to the dataframe to allow
further analysis options.
# Add annotations about the cycle to the data frame
CSS_Song2 <- add_annotation(df = CSS_Song2_Onset,
annotation = CSS_Song2_Metre$Cycle,
time = CSS_Song2_Metre$Time,
reference = 'Label.SD')Add isochronous beat subdivisions based on annotation
We add an isochronous reference subdivision of beat subdivisions based on annotated cycle start times.
# Add isochronous beats to the data frame
CSS_Song2 <- add_isobeats(df = CSS_Song2,
instr = 'CycleTime',
beat = 'SD')
print(head(CSS_Song2[,2:9]))## SD Clave Bass Guitar Tres CycleTime Cycle Iso.Time
## 1 1 NA NA NA NA 5.037333 1 5.037333
## 2 2 NA NA 5.281932 NA NA 1 5.260062
## 3 3 NA NA 5.480643 NA NA 1 5.482792
## 4 4 NA 5.714555 5.707537 5.730943 NA 1 5.705521
## 5 5 NA 5.927078 5.939071 5.917083 NA 1 5.928250
## 6 6 NA NA 6.153243 6.144901 NA 1 6.150979At this point we have the basic elements for most analyses. In the
data frame CSS_Song2 we have the onset data (coded under
instrument names, here Clave, Bass, and
Tres) and information about the beat cycles
(CycleTime and Cycle), and we also have timing
information for isochronous beat divisions (Iso.Time).
These annotation-based virtual beat structures can be systematically
off by a certain amount, so perhaps some it may be advantageous for some
analyses to infer the timing information related to the beat structure
from the existing onsets themselves and not rely on external reference.
add_isobeats function can accomplish this if given the
instruments from which it will calculate the mean onset time for the
first beat of the cycle, which will then be used to set the isochronous
beat timings for the rest of the beats within the cycle.
Add isochronous beat subdivisions based on mean onset times
# Add isochronous beats based on mean timing of guitar, tres, and clave
CSS_Song2 <- add_isobeats(df = CSS_Song2,
instr = c('Guitar','Tres','Clave'),
beat = 'SD')
# Show the newly calculated isochronous beat times
# from the second cycle onwards.
print(head(CSS_Song2[17:22,c(2,6:10)]))## SD Tres CycleTime Cycle Iso.Time Mean.Time
## 17 1 8.571167 8.601 2 8.601000 8.592649
## 18 2 NA NA 2 8.821125 8.803476
## 19 3 NA NA 2 9.041250 9.031986
## 20 4 9.240644 NA 2 9.261375 9.257808
## 21 5 NA NA 2 9.481500 9.478042
## 22 6 9.669264 NA 2 9.701625 9.670687These operations allow many to create a number of variant reference points to be used in the calculation of synchrony.