RStudio AI Weblog: Audio classification with torch

Variations on a theme

Easy audio classification with Keras, Audio classification with Keras: Wanting nearer on the non-deep studying elements, Easy audio classification with torch: No, this isn’t the primary put up on this weblog that introduces speech classification utilizing deep studying. With two of these posts (the “utilized” ones) it shares the overall setup, the kind of deep-learning structure employed, and the dataset used. With the third, it has in frequent the curiosity within the concepts and ideas concerned. Every of those posts has a distinct focus – must you learn this one?

Nicely, in fact I can’t say “no” – all of the extra so as a result of, right here, you will have an abbreviated and condensed model of the chapter on this subject within the forthcoming ebook from CRC Press, Deep Studying and Scientific Computing with R torch. By the use of comparability with the earlier put up that used torch, written by the creator and maintainer of torchaudio, Athos Damiani, important developments have taken place within the torch ecosystem, the top outcome being that the code obtained rather a lot simpler (particularly within the mannequin coaching half). That mentioned, let’s finish the preamble already, and plunge into the subject!

Inspecting the info

We use the speech instructions dataset (Warden (2018)) that comes with torchaudio. The dataset holds recordings of thirty totally different one- or two-syllable phrases, uttered by totally different audio system. There are about 65,000 audio information total. Our process might be to foretell, from the audio solely, which of thirty potential phrases was pronounced.

library(torch)
library(torchaudio)
library(luz)

ds <- speechcommand_dataset(
  root = "~/.torch-datasets", 
  url = "speech_commands_v0.01",
  obtain = TRUE
)

We begin by inspecting the info.

[1]  "mattress"    "hen"   "cat"    "canine"    "down"   "eight"
[7]  "5"   "4"   "go"     "blissful"  "home"  "left"
[32] " marvin" "9"   "no"     "off"    "on"     "one"
[19] "proper"  "seven" "sheila" "six"    "cease"   "three"
[25]  "tree"   "two"    "up"     "wow"    "sure"    "zero" 

Selecting a pattern at random, we see that the data we’ll want is contained in 4 properties: waveform, sample_rate, label_index, and label.

The primary, waveform, might be our predictor.

pattern <- ds[2000]
dim(pattern$waveform)

[1]     1 16000

Particular person tensor values are centered at zero, and vary between -1 and 1. There are 16,000 of them, reflecting the truth that the recording lasted for one second, and was registered at (or has been transformed to, by the dataset creators) a price of 16,000 samples per second. The latter data is saved in pattern$sample_rate:

[1] 16000

All recordings have been sampled on the similar price. Their size nearly all the time equals one second; the – very – few sounds which are minimally longer we are able to safely truncate.

Lastly, the goal is saved, in integer kind, in pattern$label_index, the corresponding phrase being out there from pattern$label:

pattern$label
pattern$label_index

[1] "hen"
torch_tensor
2
[ CPULongType{} ]

How does this audio sign “look?”

library(ggplot2)

df <- information.body(
  x = 1:size(pattern$waveform[1]),
  y = as.numeric(pattern$waveform[1])
  )

ggplot(df, aes(x = x, y = y)) +
  geom_line(measurement = 0.3) +
  ggtitle(
    paste0(
      "The spoken phrase "", pattern$label, "": Sound wave"
    )
  ) +
  xlab("time") +
  ylab("amplitude") +
  theme_minimal()

What we see is a sequence of amplitudes, reflecting the sound wave produced by somebody saying “hen.” Put in another way, we’ve right here a time sequence of “loudness values.” Even for specialists, guessing which phrase resulted in these amplitudes is an unimaginable process. That is the place area information is available in. The knowledgeable might not have the ability to make a lot of the sign on this illustration; however they could know a method to extra meaningfully signify it.

Two equal representations

Think about that as an alternative of as a sequence of amplitudes over time, the above wave had been represented in a approach that had no details about time in any respect. Subsequent, think about we took that illustration and tried to get better the unique sign. For that to be potential, the brand new illustration would someway need to include “simply as a lot” data because the wave we began from. That “simply as a lot” is obtained from the Fourier Rework, and it consists of the magnitudes and section shifts of the totally different frequencies that make up the sign.

How, then, does the Fourier-transformed model of the “hen” sound wave look? We get hold of it by calling torch_fft_fft() (the place fft stands for Quick Fourier Rework):

dft <- torch_fft_fft(pattern$waveform)
dim(dft)

[1]     1 16000

The size of this tensor is identical; nevertheless, its values are usually not in chronological order. As an alternative, they signify the Fourier coefficients, comparable to the frequencies contained within the sign. The upper their magnitude, the extra they contribute to the sign:

magazine <- torch_abs(dft[1, ])

df <- information.body(
  x = 1:(size(pattern$waveform[1]) / 2),
  y = as.numeric(magazine[1:8000])
)

ggplot(df, aes(x = x, y = y)) +
  geom_line(measurement = 0.3) +
  ggtitle(
    paste0(
      "The spoken phrase "",
      pattern$label,
      "": Discrete Fourier Rework"
    )
  ) +
  xlab("frequency") +
  ylab("magnitude") +
  theme_minimal()

From this alternate illustration, we might return to the unique sound wave by taking the frequencies current within the sign, weighting them in keeping with their coefficients, and including them up. However in sound classification, timing data should certainly matter; we don’t actually wish to throw it away.

Combining representations: The spectrogram

The truth is, what actually would assist us is a synthesis of each representations; some type of “have your cake and eat it, too.” What if we might divide the sign into small chunks, and run the Fourier Rework on every of them? As you’ll have guessed from this lead-up, this certainly is one thing we are able to do; and the illustration it creates known as the spectrogram.

With a spectrogram, we nonetheless maintain some time-domain data – some, since there may be an unavoidable loss in granularity. Then again, for every of the time segments, we study their spectral composition. There’s an necessary level to be made, although. The resolutions we get in time versus in frequency, respectively, are inversely associated. If we break up up the alerts into many chunks (known as “home windows”), the frequency illustration per window is not going to be very fine-grained. Conversely, if we wish to get higher decision within the frequency area, we’ve to decide on longer home windows, thus shedding details about how spectral composition varies over time. What seems like an enormous drawback – and in lots of circumstances, might be – gained’t be one for us, although, as you’ll see very quickly.

First, although, let’s create and examine such a spectrogram for our instance sign. Within the following code snippet, the scale of the – overlapping – home windows is chosen in order to permit for affordable granularity in each the time and the frequency area. We’re left with sixty-three home windows, and, for every window, get hold of 200 fifty-seven coefficients:

fft_size <- 512
window_size <- 512
energy <- 0.5

spectrogram <- transform_spectrogram(
  n_fft = fft_size,
  win_length = window_size,
  normalized = TRUE,
  energy = energy
)

spec <- spectrogram(pattern$waveform)$squeeze()
dim(spec)

[1]   257 63

We will show the spectrogram visually:

bins <- 1:dim(spec)[1]
freqs <- bins / (fft_size / 2 + 1) * pattern$sample_rate 
log_freqs <- log10(freqs)

frames <- 1:(dim(spec)[2])
seconds <- (frames / dim(spec)[2]) *
  (dim(pattern$waveform$squeeze())[1] / pattern$sample_rate)

picture(x = as.numeric(seconds),
      y = log_freqs,
      z = t(as.matrix(spec)),
      ylab = 'log frequency [Hz]',
      xlab = 'time [s]',
      col = hcl.colours(12, palette = "viridis")
)
foremost <- paste0("Spectrogram, window measurement = ", window_size)
sub <- "Magnitude (sq. root)"
mtext(facet = 3, line = 2, at = 0, adj = 0, cex = 1.3, foremost)
mtext(facet = 3, line = 1, at = 0, adj = 0, cex = 1, sub)

We all know that we’ve misplaced some decision in each time and frequency. By displaying the sq. root of the coefficients’ magnitudes, although – and thus, enhancing sensitivity – we had been nonetheless capable of get hold of an inexpensive outcome. (With the viridis coloration scheme, long-wave shades point out higher-valued coefficients; short-wave ones, the alternative.)

Lastly, let’s get again to the essential query. If this illustration, by necessity, is a compromise – why, then, would we wish to make use of it? That is the place we take the deep-learning perspective. The spectrogram is a two-dimensional illustration: a picture. With photos, we’ve entry to a wealthy reservoir of strategies and architectures: Amongst all areas deep studying has been profitable in, picture recognition nonetheless stands out. Quickly, you’ll see that for this process, fancy architectures are usually not even wanted; a simple convnet will do an excellent job.

Coaching a neural community on spectrograms

We begin by making a torch::dataset() that, ranging from the unique speechcommand_dataset(), computes a spectrogram for each pattern.

spectrogram_dataset <- dataset(
  inherit = speechcommand_dataset,
  initialize = operate(...,
                        pad_to = 16000,
                        sampling_rate = 16000,
                        n_fft = 512,
                        window_size_seconds = 0.03,
                        window_stride_seconds = 0.01,
                        energy = 2) {
    self$pad_to <- pad_to
    self$window_size_samples <- sampling_rate *
      window_size_seconds
    self$window_stride_samples <- sampling_rate *
      window_stride_seconds
    self$energy <- energy
    self$spectrogram <- transform_spectrogram(
        n_fft = n_fft,
        win_length = self$window_size_samples,
        hop_length = self$window_stride_samples,
        normalized = TRUE,
        energy = self$energy
      )
    tremendous$initialize(...)
  },
  .getitem = operate(i) {
    merchandise <- tremendous$.getitem(i)

    x <- merchandise$waveform
    # be certain that all samples have the identical size (57)
    # shorter ones might be padded,
    # longer ones might be truncated
    x <- nnf_pad(x, pad = c(0, self$pad_to - dim(x)[2]))
    x <- x %>% self$spectrogram()

    if (is.null(self$energy)) {
      # on this case, there may be a further dimension, in place 4,
      # that we wish to seem in entrance
      # (as a second channel)
      x <- x$squeeze()$permute(c(3, 1, 2))
    }

    y <- merchandise$label_index
    record(x = x, y = y)
  }
)

Within the parameter record to spectrogram_dataset(), word energy, with a default worth of two. That is the worth that, except instructed in any other case, torch’s transform_spectrogram() will assume that energy ought to have. Underneath these circumstances, the values that make up the spectrogram are the squared magnitudes of the Fourier coefficients. Utilizing energy, you possibly can change the default, and specify, for instance, that’d you’d like absolute values (energy = 1), some other optimistic worth (corresponding to 0.5, the one we used above to show a concrete instance) – or each the actual and imaginary elements of the coefficients (energy = NULL).

Show-wise, in fact, the complete advanced illustration is inconvenient; the spectrogram plot would wish a further dimension. However we might nicely wonder if a neural community might revenue from the extra data contained within the “complete” advanced quantity. In any case, when decreasing to magnitudes we lose the section shifts for the person coefficients, which could include usable data. The truth is, my exams confirmed that it did; use of the advanced values resulted in enhanced classification accuracy.

Let’s see what we get from spectrogram_dataset():

ds <- spectrogram_dataset(
  root = "~/.torch-datasets",
  url = "speech_commands_v0.01",
  obtain = TRUE,
  energy = NULL
)

dim(ds[1]$x)

[1]   2 257 101

We have now 257 coefficients for 101 home windows; and every coefficient is represented by each its actual and imaginary elements.

Subsequent, we break up up the info, and instantiate the dataset() and dataloader() objects.

train_ids <- pattern(
  1:size(ds),
  measurement = 0.6 * size(ds)
)
valid_ids <- pattern(
  setdiff(
    1:size(ds),
    train_ids
  ),
  measurement = 0.2 * size(ds)
)
test_ids <- setdiff(
  1:size(ds),
  union(train_ids, valid_ids)
)

batch_size <- 128

train_ds <- dataset_subset(ds, indices = train_ids)
train_dl <- dataloader(
  train_ds,
  batch_size = batch_size, shuffle = TRUE
)

valid_ds <- dataset_subset(ds, indices = valid_ids)
valid_dl <- dataloader(
  valid_ds,
  batch_size = batch_size
)

test_ds <- dataset_subset(ds, indices = test_ids)
test_dl <- dataloader(test_ds, batch_size = 64)

b <- train_dl %>%
  dataloader_make_iter() %>%
  dataloader_next()

dim(b$x)

[1] 128   2 257 101

The mannequin is a simple convnet, with dropout and batch normalization. The actual and imaginary elements of the Fourier coefficients are handed to the mannequin’s preliminary nn_conv2d() as two separate channels.

mannequin <- nn_module(
  initialize = operate() {
    self$options <- nn_sequential(
      nn_conv2d(2, 32, kernel_size = 3),
      nn_batch_norm2d(32),
      nn_relu(),
      nn_max_pool2d(kernel_size = 2),
      nn_dropout2d(p = 0.2),
      nn_conv2d(32, 64, kernel_size = 3),
      nn_batch_norm2d(64),
      nn_relu(),
      nn_max_pool2d(kernel_size = 2),
      nn_dropout2d(p = 0.2),
      nn_conv2d(64, 128, kernel_size = 3),
      nn_batch_norm2d(128),
      nn_relu(),
      nn_max_pool2d(kernel_size = 2),
      nn_dropout2d(p = 0.2),
      nn_conv2d(128, 256, kernel_size = 3),
      nn_batch_norm2d(256),
      nn_relu(),
      nn_max_pool2d(kernel_size = 2),
      nn_dropout2d(p = 0.2),
      nn_conv2d(256, 512, kernel_size = 3),
      nn_batch_norm2d(512),
      nn_relu(),
      nn_adaptive_avg_pool2d(c(1, 1)),
      nn_dropout2d(p = 0.2)
    )

    self$classifier <- nn_sequential(
      nn_linear(512, 512),
      nn_batch_norm1d(512),
      nn_relu(),
      nn_dropout(p = 0.5),
      nn_linear(512, 30)
    )
  },
  ahead = operate(x) {
    x <- self$options(x)$squeeze()
    x <- self$classifier(x)
    x
  }
)

We subsequent decide an acceptable studying price:

mannequin <- mannequin %>%
  setup(
    loss = nn_cross_entropy_loss(),
    optimizer = optim_adam,
    metrics = record(luz_metric_accuracy())
  )

rates_and_losses <- mannequin %>%
  lr_finder(train_dl)
rates_and_losses %>% plot()

Primarily based on the plot, I made a decision to make use of 0.01 as a maximal studying price. Coaching went on for forty epochs.

fitted <- mannequin %>%
  match(train_dl,
    epochs = 50, valid_data = valid_dl,
    callbacks = record(
      luz_callback_early_stopping(persistence = 3),
      luz_callback_lr_scheduler(
        lr_one_cycle,
        max_lr = 1e-2,
        epochs = 50,
        steps_per_epoch = size(train_dl),
        call_on = "on_batch_end"
      ),
      luz_callback_model_checkpoint(path = "models_complex/"),
      luz_callback_csv_logger("logs_complex.csv")
    ),
    verbose = TRUE
  )

plot(fitted)

Let’s test precise accuracies.

"epoch","set","loss","acc"
1,"prepare",3.09768574611813,0.12396992171405
1,"legitimate",2.52993751740923,0.284378862793572
2,"prepare",2.26747255972008,0.333642356819118
2,"legitimate",1.66693911248562,0.540791100123609
3,"prepare",1.62294889937818,0.518464153275649
3,"legitimate",1.11740599192825,0.704882571075402
...
...
38,"prepare",0.18717994078312,0.943809229501442
38,"legitimate",0.23587799138006,0.936418417799753
39,"prepare",0.19338578602993,0.942882159044087
39,"legitimate",0.230597475945365,0.939431396786156
40,"prepare",0.190593419024368,0.942727647301195
40,"legitimate",0.243536252455384,0.936186650185414

With thirty courses to differentiate between, a ultimate validation-set accuracy of ~0.94 appears like a really respectable outcome!

We will affirm this on the check set:

consider(fitted, test_dl)

loss: 0.2373
acc: 0.9324

An fascinating query is which phrases get confused most frequently. (After all, much more fascinating is how error possibilities are associated to options of the spectrograms – however this, we’ve to go away to the true area specialists. A pleasant approach of displaying the confusion matrix is to create an alluvial plot. We see the predictions, on the left, “movement into” the goal slots. (Goal-prediction pairs much less frequent than a thousandth of check set cardinality are hidden.)

Wrapup

That’s it for as we speak! Within the upcoming weeks, anticipate extra posts drawing on content material from the soon-to-appear CRC ebook, Deep Studying and Scientific Computing with R torch. Thanks for studying!

Photograph by alex lauzon on Unsplash