genre-classifier-mfcc/calculate_MFCC.Rmd at master · jsmith13/genre-classifier-mfcc · GitHub

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
---
title: "Music Library Dataset Preparation"
author: Jake Smith
---

```{r message = FALSE}
# import required libaries
require(dplyr)
require(tuneR)
require(seewave)
require(doParallel)
set.seed(12481)
```

```{r}
# import the metadata information
music <- data.table::fread("./selected_music.csv")
```

```{r}
# start a parallel processing cluster
cl <- makeCluster(3, type = "SOCK")
registerDoParallel(cl)

# calculate MFCC descriptors

foreach (row = 1:dim(music)[1], .packages = c("tuneR", "seewave")) %dopar% {
  # load the wave
  # using mp3 function if mp3
  if (substr(music[row, "Filename"], nchar(music[row, "Filename"]), nchar(music[row, "Filename"])) == "3") {
    wave.file <- readMP3(paste(music[row, "Path"], music[row, "Filename"], sep = ""))
  }
  # using flac function if flac
  if (substr(music[row, "Filename"], nchar(music[row, "Filename"]), nchar(music[row, "Filename"])) == "c") {
    # convert flac to wave
    wav2flac(paste(music[row, "Path"], music[row, "Filename"], sep = ""), reverse = TRUE, path2exe = "C:/Users/Jake/Desktop/")

    # load wave, then delete the saved wave file
    wave.file <- readWave(paste(music[row, "Path"], sub("flac", "wav", music[row, "Filename"]), sep = ""))
    file.remove(paste(music[row, "Path"], sub("flac", "wav", music[row, "Filename"]), sep = ""))
  }

  # strip silence
  wave.file <- noSilence(wave.file)

  # cut three 10 s segments from the wave file centered on approximate quarter points
  cut.points <- floor(c(1, 2, 3) * 0.25 * length(wave.file))
  segments <- cbind(cut.points - 5*wave.file@samp.rate, cut.points + 5*wave.file@samp.rate)

  wave.start <- wave.file[segments[1, 1] : segments [1, 2]]
  wave.mid <- wave.file[segments[2, 1] : segments [2, 2]]
  wave.end <- wave.file[segments[3, 1] : segments [3, 2]]

  # calculate MFCC descriptors
  descriptors <- sapply(list(wave.start, wave.mid, wave.end), function(x) {melfcc(x, numcep = 64)}, simplify = "array")

  # save to file
  save(descriptors, file = paste("./data/", row, ".descriptors", sep = ""))
}

# stop the parallel cluster
stopCluster(cl)
```