Seewave signal to noise ratio ($r) incorrect after looping through a few audio files (correct for first few)

[marywestwood]

Using seewave, I have sequential 1 minute audio files which I need to loop through to determine the signal:noise ratio. The ratio is correct for the first few files, then starts producing the wrong value for all subsequent files. I know this because when I run the files individually the correct ratio is produced.

E.g.

test<-readMP3("Audio/4_28NOV19_0442_C1 246.mp3")
# create an amplitude over time plot, with a threshold in which anything above the amplitude threshold is signal and anything below is noise
chirptest<-timer(test, f=44100, threshold=80, envt="hil", msmooth=c(100,0), 
             colval = "blue", plot=TRUE, xlab = "nochirp")
chirptest$r

= 1.5E-04

However, the output df from the loop says that the ratio of "Audio/4_28NOV19_0442_C1 246.mp3" is 3.7E-05.

Strangely, the incorrect values are repetitive, e.g.:

Here's my code, essentially I've written a loop that:

1. imports the file and creates a blank df
2. uses the seewave
   Timer function to create an amplitude threshold
3. pull the ratio ($r) of signal:noise from Timer and populate that in a new df
4. create a new df with relevant info & saves as csv

#import audio files and sort in correct order
files <- list.files("Audio", pattern="*.mp3", full.names=TRUE)
files <- mixedsort(files)

#create an empty dataframe to populate with data
df <- data.frame(chirp_r = NULL,filename=NULL)

#Create list of chirp values
for (i in files){ #Look inside your list of files (see above, called 'files')
  print(i) #take each file one by one, show on screen what it's working on
  data_inside_loop <- readMP3(i) #run readmp3 on each file one by one
  chirp <- timer(data_inside_loop, f=44100, threshold=80, envt="hil", msmooth=c(100,0), colval = "blue", plot=FALSE) 
  df <- rbind(df, data.frame(chirp_r=chirp$r, filename=i))} #populate df with the new data, one line at a time 

df$filename <- as.character(df$filename) #turn the file name into characters
df$filename_without_mp3 <- substr(df$filename, 1, nchar(df$filename)-4) #remove last 4 characters from the name (i.e. remove ".mp3)

#this code gives you each of the numbers in the character string
str_extract_all(df$filename_without_mp3,"\\d+\\)?") #see the starting time is always the 4th number in the filename

#str_extract_all finds all the numbers and puts them in a list
#the sapply part takes out the 4th number from that list
#that fourth number is the starting time 
df$starting_time <- sapply(str_extract_all(df$filename_without_mp3,"\\d+\\)?"), "[[", 4) 

#convert it to time so R understands what it is
df$starting_time <- format(as.POSIXct(strptime(df$starting_time,"%H%M",tz="")), format = "%H:%M")

#use filename to construct a full correct date and time
df$t <- paste0(sapply(str_extract_all(df$filename_without_mp3,"\\d+\\)?"), "[[", 2), #2nd number from filename is day
               "/11/", #November
               sapply(str_extract_all(df$filename_without_mp3,"\\d+\\)?"), "[[", 3), #3rd number from filename is year
               " ", #add a space
               df$starting_time) #add time

#convert time string into a format R understands as time
df$t <- as.POSIXct(df$t, format="%d/%m/%y %H:%M", tz="GMT") #yay now R understands our date and time

#the last number from the filename is the minutes elapsed
df$elapsed_min <- sapply(str_extract_all(df$filename_without_mp3,"\\d+\\)?"), "[[", 6) 

#sapply always saves stuff as characters so need to tell r that elapsed_min is a number
df$elapsed_min <- as.numeric(df$elapsed_min)

#add the elapsed time column (as minutes) to the starting time
df$time <- df$t + minutes(df$elapsed_min)

#new df
df <- df %>% dplyr::select("filename","time","chirp_r")

#write csv
write.csv(df_padded,file = paste0("Output/df_padded.csv"))

Can anyone please advise why this is happening and how I correct for it? If of interest, the goal here is that I run this code through sequential audio files of crickets chirping, to determine time-of-day of calling. I've broken down my weeklong audio files into sequential 1 minute files for workability in R.

*Update, I've just re-ran the loop on my massive list of audio files and it's now only produced the correct ratio for the first file, then file 2 is incorrect, it then repeats the same incorrect ratio for a lot of subsequent files and continues being incorrect.