Load packages

## vector with package names
x <- c( "pbapply", "parallel", "ggplot2", "warbleR", "Rraven", "readxl", "Sim.DiffProc", "png", "baRulho", "viridis")

aa <- lapply(x, function(y) {
  
  # check if installed, if not then install 
  if (!y %in% installed.packages()[,"Package"]) 
    install.packages(y) 

  # load package
  try(require(y, character.only = T), silent = T)
})

knitr::opts_knit$set(root.dir = normalizePath(".."), warnings = FALSE, message = FALSE)
knitr::opts_chunk$set(warnings = FALSE, message = FALSE)

calibration.path <- "./data/raw/calibration"

# parameters for calibration
f = 48
reps <- 5
gap.dur <- 1

# for ggplots
pd <- position_dodge(0.1)

lseq <- function(from=1, to=100000, length.out=6) {
  # logarithmic spaced sequence
  # blatantly stolen from library("emdbook"), because need only this
  exp(seq(log(from), log(to), length.out = length.out))
}

decrease <- (1 / lseq(length.out = 11))[-11]

wav_dur <- wavdur

detec_saturation <- function(X, parallel = 1, bit = 16, max_amplitude = NULL){
  
  if (is.null(max_amplitude))
  max_amplitude <- max((2 ^ bit) / 2) - 1
  
  out <- pblapply(1:nrow(X), cl = parallel, function(x){
    
    wv <- read_wave(X, index = x)
  
    out_df <- data.frame(sound.files = X$sound.files[x], selec = X$selec[x], prop.saturated = sum(wv@left == max_amplitude) / length(wv@left))
    
    return(out_df) 
})

 sat_df <- do.call(rbind, out)
 
 
 if (any(sat_df$prop.saturated > 0.2))
 cat(crayon::magenta(paste(sum(sat_df$prop.saturated > 0.2), "selections look saturated (i.e. > 20% of amplitude samples reached the highest value)"))) 

  if (any(all(sat_df$prop.saturated < 0.2) & any(sat_df$prop.saturated > 0.05)))
 cat(crayon::magenta(paste(sum(sat_df$prop.saturated > 0.05), "selections with some degree of saturation (i.e. < 20% but  > 5% of amplitude samples reached the highest value)"))) 

  if (all(sat_df$prop.saturated < 0.05))
 cat(crayon::silver("no saturation detected (all selections < 5% of amplitude samples reached the highest value)")) 

 
  return(sat_df)

}

Measure frequency on real data

ad.test <- auto_detec(threshold = 25, bp = c(1, 9), ssmooth = 300, path = "./data/processed/recording_cuts", flist = c("397_SUR_02-Feb-2021_09.21_10.32_AAE-6.wav", "416_SUR_08-Feb-2021_07.36_10.47_A3C-4.wav", "417_SUR_27-Jan-2021_07.27_10.14_AED-1.wav", "438_SJA_12-Feb-2021_08.21_09.19_AAE-2.wav", "432_SJA_16-Feb-2021_08.21_11.01_ABB-3.wav", "437_SJA_10-Feb-2021_07.30_10.46_A1A-3.wav", "444_CCL_27_Feb-2021_08.05_08.32_AC8-1.wav", "446_CCL_27-Feb-2021_07.17_07.48_A3C-1.wav", "448_CCL_26-Feb-2021_00.00_00.00_AC8-2.wav"), thinning = 0.05, hold.time = 0.1)

sa <- spectro_analysis(ad.test,  path = "./data/processed/recording_cuts", bp = c(1, 9), ssmooth = 300)

write.csv(sa, file.path(calibration.path, "spectro_analysis_on_subset_for_peak_frequency_for_calibration.csv"), row.names = FALSE)

Amplitude calibration

source('~/Dropbox/R_package_testing/baRulho/baRulho/R/master_sound_file.R')
source('~/Dropbox/R_package_testing/warbleR/warbleR/R/sim_songs.R')
source('~/Dropbox/R_package_testing/warbleR/warbleR/R/fade_env_wrblr_int.R')

# remove file created in next line (it cannot be overwritten)
unlink(file.path(calibration.path, "tones_for_amplitude_test_soundmeter.wav"))

# simulate sounds
sm_sng_sm <- sim_songs(freq = 6.5, n = length(decrease), harms = 1, diff.fun = "pure.tone", bgn = 0.0001, fin = 0.2, fout = 0.2, shape = "cos", selec.table = TRUE, samp.rate = f, file.name = "tones_for_amplitude_test_soundmeter", path = calibration.path, durs = 5, gaps = gap.dur)

# normalize
sm_sng_sm$wave <- normalize(sm_sng_sm$wave, unit = "16")

# reduce amplitude 
for (i in 1:nrow(sm_sng_sm$selec.table))
sm_sng_sm$wave@left[(sm_sng_sm$selec.table$start[i]*f*1000):(sm_sng_sm$selec.table$end[i]*f*1000)] <- sm_sng_sm$wave@left[(sm_sng_sm$selec.table$start[i]*f*1000):(sm_sng_sm$selec.table$end[i]*f*1000)] * decrease[i]

# seewave::spectro(sm_sng_sm$wave, scale = FALSE, osc = TRUE, heights = c(1, 1), grid = FALSE, flim = c(5, 8), collevels = seq(-100, 0 , 5), palette = reverse.gray.colors.1, ovlp = 0, fast.spec = TRUE)

saveRDS(sm_sng_sm, "./data/raw/calibration/tones_for_amplitude_test_soundmeter.RDS")

writeWave(sm_sng_sm$wave, "./data/raw/calibration/tones_for_amplitude_test_soundmeter.wav", extensible = FALSE)

# remove file created in next line (it cannot be overwritten)
unlink(file.path(calibration.path, "tones_for_amplitude_test_audiomoths.wav"))

# simulate sounds
sm_sng_ams <- sim_songs(freq = 6.5, n = length(decrease), harms = 1, diff.fun = "pure.tone", bgn = 0.0001, fin = 0.05, fout = 0.05, shape = "cos", selec.table = TRUE, samp.rate = f, file.name = "tones_for_amplitude_test_audiomoths", path = calibration.path)

# normalize
sm_sng_ams$wave <- normalize(sm_sng_ams$wave, unit = "16")

# reduce amplitude 
for (i in 1:nrow(sm_sng_ams$selec.table))
sm_sng_ams$wave@left[(sm_sng_ams$selec.table$start[i]*f*1000):(sm_sng_ams$selec.table$end[i]*f*1000)] <- sm_sng_ams$wave@left[(sm_sng_ams$selec.table$start[i]*f*1000):(sm_sng_ams$selec.table$end[i]*f*1000)] * decrease[i]

# seewave::spectro(sm_sng_ams$wave, scale = FALSE, osc = TRUE, heights = c(1, 1), grid = FALSE, flim = c(5, 8), collevels = seq(-100, 0 , 5), palette = reverse.gray.colors.1, ovlp = 0, fast.spec = TRUE)

saveRDS(sm_sng_ams, file.path(calibration.path, "tones_for_amplitude_test_audiomoths.RDS"))

writeWave(sm_sng_ams$wave, file.path(calibration.path, "tones_for_amplitude_test_audiomoths.wav"), extensible = FALSE)

sm_sng_sm <- readRDS(file.path(calibration.path, "tones_for_amplitude_test_soundmeter.RDS"))

sm_sng_ams <- readRDS(file.path(calibration.path, "tones_for_amplitude_test_audiomoths.RDS"))

sm_sng <- list()
sm_sng$wave <- pastew(sm_sng_ams$wave, sm_sng_sm$wave, output = "Wave", f = f * 1000)

writeWave(sm_sng$wave, file.path(calibration.path, "tones_for_amplitude_test_audiomoths_and_soundmeter.wav"), extensible = FALSE)

# add duration of tones for soundmeter to tones for audiomoth
# sm_sng_ams$selec.table$end <- sm_sng_ams$selec.table$end + duration(sm_sng_sm$wave)
# 
# sm_sng_ams$selec.table$start <- sm_sng_ams$selec.table$start + duration(sm_sng_sm$wave)

sm_sng$selec.table <- rbind(sm_sng_sm$selec.table, sm_sng_ams$selec.table)

# plot spectro
# seewave::spectro(clp, scale = FALSE, osc = TRUE, heights = c(1, 1), grid = FALSE, flim = c(5, 8), collevels = seq(-100, 0 , 5), palette = reverse.gray.colors.1)

## replicate audiomoths sound file
# extract selection table
st_ams <- sm_sng_ams$selec.table

st_ams2 <- st_ams[1, ]
st_ams2$end <- st_ams$end[nrow(st_ams)]

for(i in 1:reps)
st_ams2 <- rbind(st_ams2, st_ams2)
st_ams2$selec <- 1:nrow(st_ams2)


## replicate soundmeter sound file
# extract selection table
st_sm <- sm_sng$selec.table[sm_sng$selec.table$sound.files != "tones_for_amplitude_test_audiomoths.wav", ]

st_sm <- sm_sng_sm$selec.table

st_sm2 <- st_sm[1, ]
st_sm2$end <- st_sm$end[nrow(st_sm)]

for(i in 1:4)
st_sm2 <- rbind(st_sm2, st_sm2)

st2 <- rbind(st_sm2, st_ams2)

# add freq range (0.5 kHz)
st2$bottom.freq <- st2$bottom.freq - 0.3
st2$top.freq <- st2$top.freq + 0.3

st2$selec <- 1:nrow(st2)

# st2$sound.files <- "tones_for_amplitude_test_audiomoths_and_soundmeter.wav"

# make an extended selection table
synth.est <- selection_table(X = st2, extended = TRUE, pb = FALSE, confirm.extended = FALSE, path = calibration.path, by.song = "sound.files")

# create master sound file
synth.master.sf <- master_sound_file(X = synth.est, file.name = "calibration_sound_for_audiomoths_and_soundmeter", gap.duration = 1, dest.path = calibration.path, overwrite = TRUE, delay = 0, amp.marker = 1)


out <- pblapply(1:nrow(synth.master.sf), function(x){
  
  if(grepl("marker", synth.master.sf$orig.sound.file[x])){
    out <- synth.master.sf[x, ]
    out$label <- synth.master.sf$orig.sound.file[x]
    } else {
      
      if (grepl("soundmeter", synth.master.sf$orig.sound.file[x]))
       st_temp <- st_sm else st_temp <- st_ams 
      
       # make them start at 0
      st_temp$end <- st_temp$end - min(st_temp$start) 
      st_temp$start <- st_temp$start - min(st_temp$start) 
      st_temp$start <- st_temp$start + synth.master.sf$start[x]  
      st_temp$end <- st_temp$end + synth.master.sf$start[x]  
      
      st_temp$label <- round(decrease, 4)
      st_temp$orig.sound.file <- synth.master.sf$orig.sound.file[x]
      out <- st_temp
    }
  return(out)
} )

st3 <- do.call(rbind, out)

st3$bottom.freq <- st3$bottom.freq - 0.25
st3$top.freq <- st3$top.freq + 0.25

st3$selec <- 1:nrow(st3)

# st3$label <- c("start", rep(seq(1, 0.01, length.out = 10), (nrow(st3) - 2) / 10), "end")
st3$sound.files <- synth.master.sf$sound.files[1]

full_spectrograms(st3, path = calibration.path, fast.spec = TRUE, labels = "label", sxrow = 8, rows = 34, flim = c(3, 8), ovlp = 0)

write.csv(st3, file.path(calibration.path, "selection_table_audiomoths_and_soundmeter_playback.csv"), row.names = FALSE)

Sound pressure level from soundmeter for synthetic tones

Synthetic sounds were created as decreasing amplitude in a log scale

calb <- read_excel(file.path(calibration.path, "soundmeter_calibration.xls"))

plot(1:100, calb$`Noise(dB)`[1:100], xlab = "Time (s)", ylab = "SPL (dB)", col = viridis(10)[3], pch = 20, cex = 2)

# xys <- locator(1)

dbdf <- data.frame(dB =calb$`Noise(dB)`[rep(seq(1, 2000, 6) + 8, each = 3) + 0:2])
dbdf$label <- rep(rep(round(decrease, 5), each = 3), 40)[1:nrow(dbdf)]

# select those that match noise (the rest (> 450) don't match labels)
dbdf <- dbdf[1:450,]

agg_spl_sm <- aggregate(dB ~ label, data = dbdf, mean)

agg_spl_sm$sd <- aggregate(dB ~ label, data = dbdf, sd)$dB
agg_spl_sm$sound.files <- "soundmeter"
agg_spl_sm$label <- c(0.0000, 0.0010, 0.0032, 0.0100, 0.0316, 0.1000, 0.3162, 1.0000, 0.0001, 0.0003)

Mean (+/- sd) of sound pressure level from soundmeter for each synthetic tone

agg_spl_sm$label_cat <- factor(agg_spl_sm$label, levels = agg_spl_sm$label)

ggplot(agg_spl_sm, aes(x = label_cat, y = dB, color = sound.files)) +
    geom_errorbar(aes(ymin = dB - sd, ymax = dB + sd),width=.4, position=pd) +
    geom_line(position=pd) +
      scale_color_viridis_d(begin = 0.2, end =  0.8) +
    geom_point(position=pd, size = 2) +
  theme_classic() + labs(x = "Calibration sound decrease factor", y = "SPL (dB)")

sel_tab <- read.csv(file.path(calibration.path, "selection_table_audiomoths_and_soundmeter_playback.csv"))

frng_start <- freq_range(sel_tab[1, ], path = calibration.path)

# adjust start freq range
sel_tab$bottom.freq[1] <- frng_start$bottom.freq
sel_tab$top.freq[1] <- frng_start$top.freq

wavs <- list.files(path = calibration.path, pattern = "\\.WAV$")

found_starts <- search_templates(X = sel_tab, 
template.rows = which(sel_tab$orig.sound.file == "start_marker"), 
test.files = wavs, path = calibration.path, pb = TRUE, parallel = 1)

write.csv(found_starts, file.path(calibration.path, "starts_for_audiomoth_sync.csv"), row.names = FALSE)

found_starts <- read.csv(file.path(calibration.path, "starts_for_audiomoth_sync.csv"))

alg.tests_df <- align_test_files(X = sel_tab, Y = found_starts, path = calibration.path, by.song = TRUE, marker = "end", mar = 0.1, output = "data.frame")

exp_raven(X = alg.tests_df, path = calibration.path, sound.file.path = normalizePath(calibration.path), file.name = "audiomoth_calibration_sound_sync.txt")

fix_time <- data.frame(sound.files = c("calibration_LM_A1A.WAV", "calibration_LM_A3C.WAV", "calibration_LM_A41A.WAV", "calibration_LM_A61.WAV", "calibration_LM_AAE.WAV", "calibration_LM_ABB.WAV", "calibration_LM_AC8.WAV", "calibration_LM_AED.WAV", "calibration_M_A1A.WAV", "calibration_M_A3C.WAV", "calibration_M_A41A.WAV", "calibration_M_A61.WAV", "calibration_M_AAE.WAV", "calibration_M_AC8.WAV", "calibration_LM_A56.WAV"), 
           fix.start = c(1094.0962 - 1094.0655, 2218.2000 - 2218.1660, 3401.6597 - 3401.6116, 4658.9416 - 4658.8791, 5862.1449 - 5862.1145, 6988.0653 - 6988.0220, 8245.7745 - 8245.7395, 9359.3046 - 9359.2705, 10626.2795 - 10626.2553, 11749.9104 - 11749.8747, 12865.9948 - 12865.9454, 13977.6981 - 13977.6395, 15106.0178 - 15105.9875, 16285.0662 - 16285.0320, 4660.2555 - 4660.2103), 
           old.selec = c(472, 472, 472, 472, 472, 472, 472, 472, 472, 472, 472, 472, 472, 472, 472))

alg.tests_fix_delay <- alg.tests_df

for(i in 1:nrow(fix_time)){
      
  delay <- seq(0, fix_time$fix.start[i], length.out = fix_time$old.selec[i])
  
  delay <- c(delay, delay[2:(1 + sum(alg.tests_df$sound.files == fix_time$sound.files[i]) - fix_time$old.selec[i])] + delay[length(delay)])
  
  alg.tests_fix_delay$start[alg.tests_fix_delay$sound.files == fix_time$sound.files[i]] <- alg.tests_fix_delay$start[alg.tests_fix_delay$sound.files == fix_time$sound.files[i]] + delay  

  alg.tests_fix_delay$end[alg.tests_fix_delay$sound.files == fix_time$sound.files[i]] <- alg.tests_fix_delay$end[alg.tests_fix_delay$sound.files == fix_time$sound.files[i]] + delay  

  }

  # only for plotting and double-cheching
alg.tests_fix_delay$bottom.freq <- 3.2
alg.tests_fix_delay$top.freq <- 7.8

exp_raven(X = alg.tests_fix_delay, path = calibration.path, sound.file.path = normalizePath(calibration.path), file.name = "audiomoth_calibration_sound_sync_delay_fixed.txt")

# full_spectrograms(alg.tests_fix_delay, path = calibration.path, fast.spec = TRUE, labels = "label", sxrow = 8, rows = 60, flim = c(3, 8), ovlp = 0)

alg.tests_fix_delay$label <- sel_tab$label

alg.tests <- selection_table(X = alg.tests_fix_delay, extended = TRUE, path = calibration.path, by.song = "sound.files", mar = 0.1)

alg.tests <- rename_est_waves(alg.tests, new.sound.files = sapply(alg.tests$sound.files, function(x) strsplit(x, split = ".WAV")[[1]][1]))

saveRDS(alg.tests, file.path(calibration.path, "sel_tab_audiomoth_calibration.RDS"))

alg.tests <- readRDS(file.path(calibration.path, "/sel_tab_audiomoth_calibration.RDS"))

sat_sels <- detec_saturation(X = alg.tests, parallel = 20, bit = 16)

# agg_sat <- aggregate(prop.saturated ~ sound.files + selec, data = sat_sels, mean)
# 
# agg_sat <- agg_sat[order(agg_sat$sound.files), ]
# 
# agg_sat[agg_sat$prop.saturated > 0.05, ]
# 
# agg_sat <- merge(agg_sat, saturated_df)
# 
# agg_sat[order(agg_sat$prop.saturated), ]

sum(sat_sels$prop.saturated < 0.01) / nrow(sat_sels)

unsat_sels <- alg.tests[sat_sels$prop.saturated < 0.01, ]

saveRDS(unsat_sels, file.path(calibration.path, "unsaturated_sel_tab_audiomoth_calibration.RDS"))

alg.tests <- readRDS(file.path(calibration.path, "unsaturated_sel_tab_audiomoth_calibration.RDS"))

# alg.tests <- alg.tests[grep("marker", alg.tests$template, invert = TRUE), ]

alg.tests$spl_snr <- pbsapply(1:nrow(alg.tests), cl = 20, function(x) {

  signal <- read_wave(alg.tests, index = x, path = calibration.path)
  
  noise1 <- read_wave(alg.tests, index = x, from = alg.tests$end[alg.tests$sound.files == alg.tests$sound.files[x]][1] + 0.1, to = alg.tests$end[alg.tests$sound.files == alg.tests$sound.files[x]][1] + 0.9, path = calibration.path)
    
  signal <- ffilter(signal, f = signal@samp.rate, from = bp[1] * 1000, to = bp[2] * 1000, bandpass = TRUE, output = "Wave")
  
  noise1 <- ffilter(noise1, f = noise1@samp.rate, from = bp[1] * 1000, to = bp[2] * 1000, bandpass = TRUE, output = "Wave")
    
  # spectro(signal, flim = c(2, 10), palette = reverse.gray.colors.1, collevels = seq(-100, 0, 5))
  
  sigamp <- seewave::rms(seewave::env(signal, envt = "abs", plot = FALSE))
  noisamp <- seewave::rms(seewave::env(noise1, envt = "abs", plot = FALSE))

  signaldb <- 20 * log10(sigamp)
  noisdb <- 20 * log10(noisamp)

  spl <- 10 * log10((10 ^ (signaldb/ 10)) - (10 ^ (noisdb / 10)))

    return(spl)
  })

alg.tests$spl <- pbsapply(1:nrow(alg.tests), cl = 20, function(x) {

  signal <- read_wave(alg.tests, index = x, path = calibration.path)
  
   # signal <- ffilter(signal, f = signal@samp.rate, from = bp[1] * 1000, to = bp[2] * 1000, bandpass = TRUE, output = "Wave")
   
  sigamp <- seewave::rms(seewave::env(signal, envt = "abs", plot = FALSE))
  signaldb <- 20 * log10(sigamp)
  
    return(signaldb)
  })

# get mean and sd
agg_spl <- aggregate(spl ~ sound.files + label, data = alg.tests, meandB)
agg_spl$sd <- aggregate(spl ~ sound.files + label, data = alg.tests, sddB)$spl

# difference among H for 2 audiomoths
agg_spl$label <- as.numeric(agg_spl$label)

write.csv(agg_spl, file.path(calibration.path, "spl_by_audiomoth_calibration_sounds.csv"), row.names = FALSE)

Raw amplitude values (before calibration)

Split in 3 plots to better see specific audiomoths

agg_spl <- read.csv(file.path(calibration.path, "spl_by_audiomoth_calibration_sounds.csv"))

names(agg_spl)[3] <- "dB"
agg_spl_sm$label_cat <- NULL

agg_spl_all <- rbind(agg_spl_sm, agg_spl)

agg_spl_all$orig.dB <- factor(agg_spl_all$label, levels = unique(agg_spl_all$label))

ggplot(agg_spl_all[agg_spl_all$sound.files %in% c("soundmeter", unique(agg_spl_all$sound.files)[2:6]), ], aes(x = orig.dB, y = dB, group = sound.files, color = sound.files)) +
    geom_errorbar(aes(ymin=dB - sd, ymax = dB + sd), colour="black", width=.2, position=pd) +
  scale_color_viridis_d() +
    geom_line(position=pd) +
    geom_point(position=pd, size=1) +
  theme_classic() + labs(x = "Calibration sound")

ggplot(agg_spl_all[agg_spl_all$sound.files %in% c("soundmeter", unique(agg_spl_all$sound.files)[7:10]), ], aes(x = orig.dB, y = dB, group = sound.files, color = sound.files)) +
    geom_errorbar(aes(ymin=dB - sd, ymax = dB + sd), colour="black", width=.2, position=pd) +
  scale_color_viridis_d() +
    geom_line(position=pd) +
    geom_point(position=pd, size=1) +
  theme_classic() + labs(x = "Calibration sound")

ggplot(agg_spl_all[agg_spl_all$sound.files %in% c("soundmeter", unique(agg_spl_all$sound.files)[11:16]), ], aes(x = orig.dB, y = dB, group = sound.files, color = sound.files)) +
    geom_errorbar(aes(ymin=dB - sd, ymax = dB + sd), colour="black", width=.2, position=pd) +
  scale_color_viridis_d() +
    geom_line(position=pd) +
    geom_point(position=pd, size=1) +
  theme_classic() + labs(x = "Calibration sound")

Calibrating audiomoths using proportional adjustment

Note that this way of calibrating has high and non-randomly distributed error

agg_spl <- read.csv(file.path(calibration.path, "spl_by_audiomoth_calibration_sounds.csv"))

names(agg_spl)[3] <- "dB"
agg_spl_sm$sound.files <- "soundmeter"
agg_spl_sm$label <- sort(as.numeric(unique(agg_spl$label)))

agg_spl_all <- rbind(agg_spl_sm, agg_spl)

agg_spl_all$orig.dB <- factor(agg_spl_all$label, levels = unique(agg_spl_all$label))

agg_spl_all$adj_dB <- agg_spl_all$dB

agg_spl_all <- agg_spl_all[agg_spl_all$label > 0.01, ]

for(i in unique(agg_spl_all$sound.files[agg_spl_all$sound.files != "soundmeter"]))
agg_spl_all$adj_dB[agg_spl_all$sound.files == i] <- agg_spl_all$adj_dB[agg_spl_all$sound.files == i] *
  (mean(agg_spl_all$dB[agg_spl_all$sound.files == "soundmeter"])
/ mean(agg_spl_all$dB[agg_spl_all$sound.files == i]) )


ggplot(agg_spl_all[agg_spl_all$sound.files %in% c("soundmeter", unique(agg_spl_all$sound.files)[2:6]), ], aes(x = orig.dB, y = adj_dB, group = sound.files, color = sound.files)) +
    geom_errorbar(aes(ymin=adj_dB - sd, ymax = adj_dB + sd, color = sound.files),width=.2, position=pd) +
    geom_line(position=pd) +
    scale_color_viridis_d() +
    geom_point(position=pd, size=1) +
  theme_classic() + labs(x = "Calibration sound decrease factor")

ggplot(agg_spl_all[agg_spl_all$sound.files %in% c("soundmeter", unique(agg_spl_all$sound.files)[7:10]), ], aes(x = orig.dB, y = adj_dB, group = sound.files, color = sound.files)) +
    geom_errorbar(aes(ymin=adj_dB - sd, ymax = adj_dB + sd, color = sound.files),width=.2, position=pd) +
    geom_line(position=pd) +
    scale_color_viridis_d() +
    geom_point(position=pd, size=1) +
  theme_classic() + labs(x = "Calibration sound decrease factor")

ggplot(agg_spl_all[agg_spl_all$sound.files %in% c("soundmeter", unique(agg_spl_all$sound.files)[11:16]), ], aes(x = orig.dB, y = adj_dB, group = sound.files, color = sound.files)) +
    geom_errorbar(aes(ymin=adj_dB - sd, ymax = adj_dB + sd, color = sound.files),width=.2, position=pd) +
    geom_line(position=pd) +
    scale_color_viridis_d() +
    geom_point(position=pd, size=1) +
  theme_classic() + labs(x = "Calibration sound decrease factor")

Calibrating audiomoths using addition of mean amplitude difference compared to soundmeter

These plots show the SPL estimated by each audiomoths after calibration

agg_spl <- read.csv(file.path(calibration.path, "spl_by_audiomoth_calibration_sounds.csv"))

names(agg_spl)[3] <- "dB"
agg_spl_sm$sound.files <- "soundmeter"
agg_spl_sm$label <- sort(as.numeric(unique(agg_spl$label)))

agg_spl_all <- rbind(agg_spl_sm, agg_spl)

agg_spl_all$orig.dB <- factor(agg_spl_all$label, levels = unique(agg_spl_all$label))

agg_spl_all$adj_dB <- agg_spl_all$dB

agg_spl_all <- agg_spl_all[agg_spl_all$label > 0.01, ]

# remove those with 0.1 not working
agg_spl_all <- agg_spl_all[!(agg_spl_all$label <= 0.1 & agg_spl_all$sound.files == "calibration_LM_A41A"), ]

agg_spl_all <- agg_spl_all[!(agg_spl_all$label < 0.04 & agg_spl_all$sound.files == "calibration_LM_AAE"), ]

agg_spl_all <- agg_spl_all[!(agg_spl_all$label < 0.04 & agg_spl_all$sound.files == "calibration_M_A41A"), ]


for(i in unique(agg_spl_all$sound.files[agg_spl_all$sound.files != "soundmeter"]))
  agg_spl_all$adj_dB[agg_spl_all$sound.files == i] <- agg_spl_all$adj_dB[agg_spl_all$sound.files == i] +
mean(agg_spl_all$dB[agg_spl_all$sound.files == "soundmeter" & agg_spl_all$label %in% agg_spl_all$label[agg_spl_all$sound.files == i]]) -
  mean(agg_spl_all$dB[agg_spl_all$sound.files == i])


ggplot(agg_spl_all[agg_spl_all$sound.files %in% c("soundmeter", "calibration_LM_A1A", "calibration_LM_A3C", "calibration_LM_A56", "calibration_LM_A61",  "calibration_LM_AAE"), ], aes(x = orig.dB, y = adj_dB, group = sound.files, color = sound.files)) +
    geom_errorbar(aes(ymin=adj_dB - sd, ymax = adj_dB + sd, color = sound.files),width=.2, position=pd) +
    geom_line(position=pd) +
    scale_color_viridis_d() +
    geom_point(position=pd, size=1) +
  theme_classic() + labs(x = "Calibration sound decrease factor")

ggplot(agg_spl_all[agg_spl_all$sound.files %in% c("soundmeter", "calibration_LM_ABB", "calibration_LM_AC8", "calibration_LM_AED", "calibration_LM_A41A"), ], aes(x = orig.dB, y = adj_dB, group = sound.files, color = sound.files)) +
    geom_errorbar(aes(ymin=adj_dB - sd, ymax = adj_dB + sd, color = sound.files),width=.2, position=pd) +
    geom_line(position=pd) +
    scale_color_viridis_d() +
    geom_point(position=pd, size=1) +
  theme_classic() + labs(x = "Calibration sound decrease factor")

ggplot(agg_spl_all[agg_spl_all$sound.files %in% c("soundmeter", "calibration_M_A1A", "calibration_M_A3C", "calibration_M_A41A", "calibration_M_A61", "calibration_M_AAE", "calibration_M_AC8"), ], aes(x = orig.dB, y = adj_dB, group = sound.files, color = sound.files)) +
    geom_errorbar(aes(ymin=adj_dB - sd, ymax = adj_dB + sd, color = sound.files),width=.2, position=pd) +
    geom_line(position=pd) +
    scale_color_viridis_d() +
    geom_point(position=pd, size=1) +
  theme_classic() + labs(x = "Calibration sound decrease factor")

Error margin after adjusting SPL from audiomoths

agg_spl_all <- agg_spl_all[order(agg_spl_all$sound.files, agg_spl_all$label), ]

mean_diff_dB_l <- lapply(unique(agg_spl_all$sound.files), function(x){  
  diffs <- agg_spl_all$dB[agg_spl_all$sound.files == "soundmeter" & agg_spl_all$label %in% agg_spl_all$label[agg_spl_all$sound.files == x]] - agg_spl_all$dB[agg_spl_all$sound.files == x] 
    
  out_df <- data.frame(sound.files = x, mean.diff = mean(diffs), sd.diff = sd(diffs))

  return(out_df)
})

mean_diff_dB <- do.call(rbind, mean_diff_dB_l)

mean_diff_dB$sd.diff[16] <- max(agg_spl_sm$sd)


mean_diff_dB <- mean_diff_dB[order(mean_diff_dB$sd.diff), ]

mean_diff_dB$sound.files <- factor(mean_diff_dB$sound.files, levels = mean_diff_dB$sound.files[order(mean_diff_dB$mean.diff)])

ggplot(mean_diff_dB, aes(x = sound.files, y = mean.diff, group = sound.files, color = sound.files)) +
    geom_errorbar(aes(ymin=mean.diff - sd.diff, ymax = mean.diff + sd.diff, color = sound.files),width=.2, position=pd) +
    geom_line(position=pd) +
    scale_color_viridis_d(end = 0.8) +
    geom_point(position=pd, size=1) +
  theme_classic() + labs(x = "Calibration sound decrease factor") +
  theme(axis.text.x = element_text(angle = 45, vjust = 0.5))

SPL adjustment for amplitude calibration

knitr::kable(mean_diff_dB)

	sound.files	mean.diff	sd.diff
15	calibration_M_AC8	12.00263	0.0345494
13	calibration_M_A61	14.17373	0.2095014
10	calibration_M_A1A	14.65209	0.2330067
14	calibration_M_AAE	15.53208	0.3143449
11	calibration_M_A3C	24.49568	0.6802857
8	calibration_LM_AC8	25.97500	0.7463357
4	calibration_LM_A56	22.43438	0.7670819
2	calibration_LM_A3C	20.90164	0.8571268
1	calibration_LM_A1A	22.34474	0.8788016
9	calibration_LM_AED	23.92264	0.9021281
5	calibration_LM_A61	25.97472	0.9079486
12	calibration_M_A41A	40.93480	0.9969448
7	calibration_LM_ABB	35.83943	1.0862459
3	calibration_LM_A41A	43.36072	1.1764828
6	calibration_LM_AAE	39.44369	1.2066030
16	soundmeter	0.00000	1.8214130

write.csv(mean_diff_dB, "./output/spl_constant_for_audiomoth_calibration.csv", row.names = FALSE)

saturated_df <- data.frame(sound.files = c("calibration_LM_A1A.WAV", "calibration_LM_A1A.WAV", "calibration_LM_A3C.WAV", "calibration_LM_A3C.WAV", "calibration_LM_A61.WAV", "calibration_LM_AED.WAV", "calibration_LM_AED.WAV", "calibration_M_A1A.WAV", "calibration_M_A1A.WAV", "calibration_M_A1A.WAV", "calibration_M_A1A.WAV", "calibration_M_A61.WAV", "calibration_M_A61.WAV", "calibration_M_A61.WAV", "calibration_M_A61.WAV", "calibration_M_AAE.WAV", "calibration_M_AAE.WAV", "calibration_M_AAE.WAV", "calibration_M_AAE.WAV", "calibration_M_AC8.WAV", "calibration_M_AC8.WAV", "calibration_M_AC8.WAV", "calibration_M_AC8.WAV", "calibration_M_AC8.WAV", "calibration_M_AC8.WAV", "calibration_M_AC8.WAV", "calibration_M_AC8.WAV"),
                           selec = c(432, 472, 2, 12, 2, 2, 32, 42, 82, 172, 352, 12, 52, 222, 302, 12, 82, 222, 262, 32, 62, 252, 282, 33, 63, 253, 283),
                           type = c("fine", "fine", "partial_saturation", "fine", "fine", "partial_saturation", "fine", "saturated", "saturated", "saturated",  "saturated",  "saturated",  "saturated", "saturated",  "saturated", "saturated", "saturated", "saturated", "saturated", "saturated", "saturated", "saturated", "saturated", "fine", "fine", "fine", "fine")) 

saturated_df$sound.files <- gsub("\\.WAV$", "", saturated_df$sound.files)
saturated_df$sf.selec <- paste(saturated_df$sound.files, saturated_df$selec, sep = "-")

alg.tests <- readRDS(file.path(calibration.path, "/sel_tab_audiomoth_calibration.RDS"))

alg.tests$sf.selec <- paste(alg.tests$sound.files, alg.tests$selec, sep = "-")

# extract examples for detecting saturation
sat_est <- alg.tests[alg.tests$sf.selec %in% saturated_df$sf.selec, ]

alg.tests$sf.selec <- saturated_df$sf.selec <- NULL

sat_est <- merge(sat_est, saturated_df, by = c("sound.files", "selec"))

sat_est <- fix_extended_selection_table(X = sat_est, Y = alg.tests)

sat <- detec_saturation(X = sat_est, parallel = 20)

sat[order(sat$type), ]

Session information

sessionInfo()

## R version 4.0.3 (2020-10-10)
## Platform: x86_64-pc-linux-gnu (64-bit)
## Running under: Ubuntu 20.04.1 LTS
## 
## Matrix products: default
## BLAS:   /usr/lib/x86_64-linux-gnu/openblas-pthread/libblas.so.3
## LAPACK: /usr/lib/x86_64-linux-gnu/openblas-pthread/liblapack.so.3
## 
## locale:
##  [1] LC_CTYPE=pt_PT.UTF-8       LC_NUMERIC=C              
##  [3] LC_TIME=es_CR.UTF-8        LC_COLLATE=pt_PT.UTF-8    
##  [5] LC_MONETARY=es_CR.UTF-8    LC_MESSAGES=pt_PT.UTF-8   
##  [7] LC_PAPER=es_CR.UTF-8       LC_NAME=C                 
##  [9] LC_ADDRESS=C               LC_TELEPHONE=C            
## [11] LC_MEASUREMENT=es_CR.UTF-8 LC_IDENTIFICATION=C       
## 
## attached base packages:
## [1] parallel  stats     graphics  grDevices utils     datasets  methods  
## [8] base     
## 
## other attached packages:
##  [1] viridis_0.5.1      viridisLite_0.3.0  baRulho_1.0.4      png_0.1-7         
##  [5] Sim.DiffProc_4.8   readxl_1.3.1       Rraven_1.0.12      warbleR_1.1.26    
##  [9] NatureSounds_1.0.3 knitr_1.31         seewave_2.1.6      tuneR_1.3.3       
## [13] ggplot2_3.3.3      pbapply_1.4-3     
## 
## loaded via a namespace (and not attached):
##  [1] tidyselect_1.1.0  xfun_0.21         purrr_0.3.4       colorspace_2.0-0 
##  [5] vctrs_0.3.6       generics_0.1.0    htmltools_0.5.1.1 yaml_2.2.1       
##  [9] rlang_0.4.10      pillar_1.4.6      glue_1.4.2        withr_2.4.1      
## [13] lifecycle_0.2.0   stringr_1.4.0     munsell_0.5.0     gtable_0.3.0     
## [17] cellranger_1.1.0  evaluate_0.14     labeling_0.4.2    fftw_1.0-6       
## [21] highr_0.8         Rcpp_1.0.6        scales_1.1.1      farver_2.1.0     
## [25] Deriv_4.1.3       gridExtra_2.3     rjson_0.2.20      digest_0.6.27    
## [29] stringi_1.5.3     dplyr_1.0.2       dtw_1.22-3        grid_4.0.3       
## [33] tools_4.0.3       bitops_1.0-6      magrittr_2.0.1    RCurl_1.98-1.2   
## [37] proxy_0.4-25      tibble_3.0.4      crayon_1.4.1      pkgconfig_2.0.3  
## [41] MASS_7.3-53       ellipsis_0.3.1    rmarkdown_2.4     R6_2.5.0         
## [45] signal_0.7-6      compiler_4.0.3

Calibration

Long-billed hermit song amplitude

Marcelo Araya-Salas, PhD & Melanie Talavera

26-03-2021

Load packages

Measure frequency on real data

Amplitude calibration

Sound pressure level from soundmeter for synthetic tones

Mean (+/- sd) of sound pressure level from soundmeter for each synthetic tone

Raw amplitude values (before calibration)

Calibrating audiomoths using proportional adjustment

Calibrating audiomoths using addition of mean amplitude difference compared to soundmeter

Error margin after adjusting SPL from audiomoths

SPL adjustment for amplitude calibration

Session information