Posit AI Weblog: Wavelet Remodel

Notice: Like a number of prior ones, this put up is an excerpt from the forthcoming ebook, Deep Studying and Scientific Computing with R torch. And like many excerpts, it’s a product of laborious trade-offs. For added depth and extra examples, I’ve to ask you to please seek the advice of the ebook.

Wavelets and the Wavelet Remodel

What are wavelets? Just like the Fourier foundation, they’re features; however they don’t prolong infinitely. As a substitute, they’re localized in time: Away from the middle, they rapidly decay to zero. Along with a location parameter, in addition they have a scale: At completely different scales, they seem squished or stretched. Squished, they are going to do higher at detecting excessive frequencies; the converse applies once they’re stretched out in time.

The essential operation concerned within the Wavelet Remodel is convolution – have the (flipped) wavelet slide over the info, computing a sequence of dot merchandise. This fashion, the wavelet is mainly in search of similarity.

As to the wavelet features themselves, there are various of them. In a sensible utility, we’d wish to experiment and decide the one which works greatest for the given information. In comparison with the DFT and spectrograms, extra experimentation tends to be concerned in wavelet evaluation.

The subject of wavelets could be very completely different from that of Fourier transforms in different respects, as nicely. Notably, there’s a lot much less standardization in terminology, use of symbols, and precise practices. On this introduction, I’m leaning closely on one particular exposition, the one in Arnt Vistnes’ very good ebook on waves (Vistnes 2018). In different phrases, each terminology and examples mirror the alternatives made in that ebook.

Introducing the Morlet wavelet

The Morlet, also called Gabor, wavelet is outlined like so:

[
Psi_{omega_{a},K,t_{k}}(t_n) = (e^{-i omega_{a} (t_n – t_k)} – e^{-K^2}) e^{- omega_a^2 (t_n – t_k )^2 /(2K )^2}
]

This formulation pertains to discretized information, the varieties of knowledge we work with in observe. Thus, (t_k) and (t_n) designate closing dates, or equivalently, particular person time-series samples.

This equation appears to be like daunting at first, however we are able to “tame” it a bit by analyzing its construction, and pointing to the primary actors. For concreteness, although, we first take a look at an instance wavelet.

We begin by implementing the above equation:

Evaluating code and mathematical formulation, we discover a distinction. The operate itself takes one argument, (t_n); its realization, 4 (omega, Okay, t_k, and t). It is because the torch code is vectorized: On the one hand, omega, Okay, and t_k, which, within the components, correspond to (omega_{a}), (Okay), and (t_k) , are scalars. (Within the equation, they’re assumed to be fastened.) t, then again, is a vector; it’s going to maintain the measurement occasions of the sequence to be analyzed.

We decide instance values for omega, Okay, and t_k, in addition to a variety of occasions to guage the wavelet on, and plot its values:

omega <- 6 * pi
Okay <- 6
t_k <- 5
 
sample_time <- torch_arange(3, 7, 0.0001)

create_wavelet_plot <- operate(omega, Okay, t_k, sample_time) {
  morlet <- morlet(omega, Okay, t_k, sample_time)
  df <- information.body(
    x = as.numeric(sample_time),
    actual = as.numeric(morlet$actual),
    imag = as.numeric(morlet$imag)
  ) %>%
    pivot_longer(-x, names_to = "half", values_to = "worth")
  ggplot(df, aes(x = x, y = worth, coloration = half)) +
    geom_line() +
    scale_colour_grey(begin = 0.8, finish = 0.4) +
    xlab("time") +
    ylab("wavelet worth") +
    ggtitle("Morlet wavelet",
      subtitle = paste0("ω_a = ", omega / pi, "π , Okay = ", Okay)
    ) +
    theme_minimal()
}

create_wavelet_plot(omega, Okay, t_k, sample_time)

What we see here’s a complicated sine curve – word the actual and imaginary elements, separated by a section shift of (pi/2) – that decays on each side of the middle. Trying again on the equation, we are able to establish the components answerable for each options. The primary time period within the equation, (e^{-i omega_{a} (t_n – t_k)}), generates the oscillation; the third, (e^{- omega_a^2 (t_n – t_k )^2 /(2K )^2}), causes the exponential decay away from the middle. (In case you’re questioning concerning the second time period, (e^{-Okay^2}): For given (Okay), it’s only a fixed.)

The third time period truly is a Gaussian, with location parameter (t_k) and scale (Okay). We’ll discuss (Okay) in nice element quickly, however what’s with (t_k)? (t_k) is the middle of the wavelet; for the Morlet wavelet, that is additionally the placement of most amplitude. As distance from the middle will increase, values rapidly method zero. That is what is supposed by wavelets being localized: They’re “energetic” solely on a brief vary of time.

The roles of (Okay) and (omega_a)

Now, we already mentioned that (Okay) is the size of the Gaussian; it thus determines how far the curve spreads out in time. However there’s additionally (omega_a). Trying again on the Gaussian time period, it, too, will impression the unfold.

First although, what’s (omega_a)? The subscript (a) stands for “evaluation”; thus, (omega_a) denotes a single frequency being probed.

Now, let’s first examine visually the respective impacts of (omega_a) and (Okay).

p1 <- create_wavelet_plot(6 * pi, 4, 5, sample_time)
p2 <- create_wavelet_plot(6 * pi, 6, 5, sample_time)
p3 <- create_wavelet_plot(6 * pi, 8, 5, sample_time)
p4 <- create_wavelet_plot(4 * pi, 6, 5, sample_time)
p5 <- create_wavelet_plot(6 * pi, 6, 5, sample_time)
p6 <- create_wavelet_plot(8 * pi, 6, 5, sample_time)

(p1 | p4) /
  (p2 | p5) /
  (p3 | p6)

Within the left column, we hold (omega_a) fixed, and range (Okay). On the proper, (omega_a) modifications, and (Okay) stays the identical.

Firstly, we observe that the upper (Okay), the extra the curve will get unfold out. In a wavelet evaluation, because of this extra closing dates will contribute to the rework’s output, leading to excessive precision as to frequency content material, however lack of decision in time. (We’ll return to this – central – trade-off quickly.)

As to (omega_a), its impression is twofold. On the one hand, within the Gaussian time period, it counteracts – precisely, even – the size parameter, (Okay). On the opposite, it determines the frequency, or equivalently, the interval, of the wave. To see this, check out the proper column. Akin to the completely different frequencies, we have now, within the interval between 4 and 6, 4, six, or eight peaks, respectively.

This double function of (omega_a) is the explanation why, all-in-all, it does make a distinction whether or not we shrink (Okay), protecting (omega_a) fixed, or enhance (omega_a), holding (Okay) fastened.

This state of issues sounds difficult, however is much less problematic than it may appear. In observe, understanding the function of (Okay) is essential, since we have to decide wise (Okay) values to strive. As to the (omega_a), then again, there will likely be a large number of them, similar to the vary of frequencies we analyze.

So we are able to perceive the impression of (Okay) in additional element, we have to take a primary take a look at the Wavelet Remodel.

Wavelet Remodel: A simple implementation

Whereas general, the subject of wavelets is extra multifaceted, and thus, could seem extra enigmatic than Fourier evaluation, the rework itself is less complicated to know. It’s a sequence of native convolutions between wavelet and sign. Right here is the components for particular scale parameter (Okay), evaluation frequency (omega_a), and wavelet location (t_k):

[
W_{K, omega_a, t_k} = sum_n x_n Psi_{omega_{a},K,t_{k}}^*(t_n)
]

That is only a dot product, computed between sign and complex-conjugated wavelet. (Right here complicated conjugation flips the wavelet in time, making this convolution, not correlation – a indisputable fact that issues lots, as you’ll see quickly.)

Correspondingly, simple implementation ends in a sequence of dot merchandise, every similar to a distinct alignment of wavelet and sign. Beneath, in wavelet_transform(), arguments omega and Okay are scalars, whereas x, the sign, is a vector. The result’s the wavelet-transformed sign, for some particular Okay and omega of curiosity.

wavelet_transform <- operate(x, omega, Okay) {
  n_samples <- dim(x)[1]
  W <- torch_complex(
    torch_zeros(n_samples), torch_zeros(n_samples)
  )
  for (i in 1:n_samples) {
    # transfer middle of wavelet
    t_k <- x[i, 1]
    m <- morlet(omega, Okay, t_k, x[, 1])
    # compute native dot product
    # word wavelet is conjugated
    dot <- torch_matmul(
      m$conj()$unsqueeze(1),
      x[, 2]$to(dtype = torch_cfloat())
    )
    W[i] <- dot
  }
  W
}

To check this, we generate a easy sine wave that has a frequency of 100 Hertz in its first half, and double that within the second.

gencos <- operate(amp, freq, section, fs, length) {
  x <- torch_arange(0, length, 1 / fs)[1:-2]$unsqueeze(2)
  y <- amp * torch_cos(2 * pi * freq * x + section)
  torch_cat(listing(x, y), dim = 2)
}

# sampling frequency
fs <- 8000

f1 <- 100
f2 <- 200
section <- 0
length <- 0.25

s1 <- gencos(1, f1, section, fs, length)
s2 <- gencos(1, f2, section, fs, length)

s3 <- torch_cat(listing(s1, s2), dim = 1)
s3[(dim(s1)[1] + 1):(dim(s1)[1] * 2), 1] <-
  s3[(dim(s1)[1] + 1):(dim(s1)[1] * 2), 1] + length

df <- information.body(
  x = as.numeric(s3[, 1]),
  y = as.numeric(s3[, 2])
)
ggplot(df, aes(x = x, y = y)) +
  geom_line() +
  xlab("time") +
  ylab("amplitude") +
  theme_minimal()

Now, we run the Wavelet Remodel on this sign, for an evaluation frequency of 100 Hertz, and with a Okay parameter of two, discovered by means of fast experimentation:

Okay <- 2
omega <- 2 * pi * f1

res <- wavelet_transform(x = s3, omega, Okay)
df <- information.body(
  x = as.numeric(s3[, 1]),
  y = as.numeric(res$abs())
)

ggplot(df, aes(x = x, y = y)) +
  geom_line() +
  xlab("time") +
  ylab("Wavelet Remodel") +
  theme_minimal()

The rework accurately picks out the a part of the sign that matches the evaluation frequency. In case you really feel like, you would possibly wish to double-check what occurs for an evaluation frequency of 200 Hertz.

Now, in actuality we’ll wish to run this evaluation not for a single frequency, however a variety of frequencies we’re involved in. And we’ll wish to strive completely different scales Okay. Now, in the event you executed the code above, you may be frightened that this might take a lot of time.

Properly, it by necessity takes longer to compute than its Fourier analogue, the spectrogram. For one, that’s as a result of with spectrograms, the evaluation is “simply” two-dimensional, the axes being time and frequency. With wavelets there are, as well as, completely different scales to be explored. And secondly, spectrograms function on entire home windows (with configurable overlap); a wavelet, then again, slides over the sign in unit steps.

Nonetheless, the scenario isn’t as grave because it sounds. The Wavelet Remodel being a convolution, we are able to implement it within the Fourier area as a substitute. We’ll do this very quickly, however first, as promised, let’s revisit the subject of various Okay.

Decision in time versus in frequency

We already noticed that the upper Okay, the extra spread-out the wavelet. We are able to use our first, maximally simple, instance, to research one quick consequence. What, for instance, occurs for Okay set to twenty?

Okay <- 20

res <- wavelet_transform(x = s3, omega, Okay)
df <- information.body(
  x = as.numeric(s3[, 1]),
  y = as.numeric(res$abs())
)

ggplot(df, aes(x = x, y = y)) +
  geom_line() +
  xlab("time") +
  ylab("Wavelet Remodel") +
  theme_minimal()

The Wavelet Remodel nonetheless picks out the right area of the sign – however now, as a substitute of a rectangle-like outcome, we get a considerably smoothed model that doesn’t sharply separate the 2 areas.

Notably, the primary 0.05 seconds, too, present appreciable smoothing. The bigger a wavelet, the extra element-wise merchandise will likely be misplaced on the finish and the start. It is because transforms are computed aligning the wavelet in any respect sign positions, from the very first to the final. Concretely, after we compute the dot product at location t_k = 1, only a single pattern of the sign is taken into account.

Other than presumably introducing unreliability on the boundaries, how does wavelet scale have an effect on the evaluation? Properly, since we’re correlating (convolving, technically; however on this case, the impact, in the long run, is identical) the wavelet with the sign, point-wise similarity is what issues. Concretely, assume the sign is a pure sine wave, the wavelet we’re utilizing is a windowed sinusoid just like the Morlet, and that we’ve discovered an optimum Okay that properly captures the sign’s frequency. Then every other Okay, be it bigger or smaller, will lead to much less point-wise overlap.

Performing the Wavelet Remodel within the Fourier area

Quickly, we’ll run the Wavelet Remodel on an extended sign. Thus, it’s time to pace up computation. We already mentioned that right here, we profit from time-domain convolution being equal to multiplication within the Fourier area. The general course of then is that this: First, compute the DFT of each sign and wavelet; second, multiply the outcomes; third, inverse-transform again to the time area.

The DFT of the sign is rapidly computed:

F <- torch_fft_fft(s3[ , 2])

With the Morlet wavelet, we don’t even should run the FFT: Its Fourier-domain illustration may be said in closed kind. We’ll simply make use of that formulation from the outset. Right here it’s:

morlet_fourier <- operate(Okay, omega_a, omega) {
  2 * (torch_exp(-torch_square(
    Okay * (omega - omega_a) / omega_a
  )) -
    torch_exp(-torch_square(Okay)) *
      torch_exp(-torch_square(Okay * omega / omega_a)))
}

Evaluating this assertion of the wavelet to the time-domain one, we see that – as anticipated – as a substitute of parameters t and t_k it now takes omega and omega_a. The latter, omega_a, is the evaluation frequency, the one we’re probing for, a scalar; the previous, omega, the vary of frequencies that seem within the DFT of the sign.

In instantiating the wavelet, there’s one factor we have to pay particular consideration to. In FFT-think, the frequencies are bins; their quantity is decided by the size of the sign (a size that, for its half, straight depends upon sampling frequency). Our wavelet, then again, works with frequencies in Hertz (properly, from a person’s perspective; since this unit is significant to us). What this implies is that to morlet_fourier, as omega_a we have to move not the worth in Hertz, however the corresponding FFT bin. Conversion is completed relating the variety of bins, dim(x)[1], to the sampling frequency of the sign, fs:

# once more search for 100Hz elements
omega <- 2 * pi * f1

# want the bin similar to some frequency in Hz
omega_bin <- f1/fs * dim(s3)[1]

We instantiate the wavelet, carry out the Fourier-domain multiplication, and inverse-transform the outcome:

Okay <- 3

m <- morlet_fourier(Okay, omega_bin, 1:dim(s3)[1])
prod <- F * m
remodeled <- torch_fft_ifft(prod)

Placing collectively wavelet instantiation and the steps concerned within the evaluation, we have now the next. (Notice the way to wavelet_transform_fourier, we now, conveniently, move within the frequency worth in Hertz.)

wavelet_transform_fourier <- operate(x, omega_a, Okay, fs) {
  N <- dim(x)[1]
  omega_bin <- omega_a / fs * N
  m <- morlet_fourier(Okay, omega_bin, 1:N)
  x_fft <- torch_fft_fft(x)
  prod <- x_fft * m
  w <- torch_fft_ifft(prod)
  w
}

We’ve already made vital progress. We’re prepared for the ultimate step: automating evaluation over a variety of frequencies of curiosity. It will lead to a three-dimensional illustration, the wavelet diagram.

Creating the wavelet diagram

Within the Fourier Remodel, the variety of coefficients we acquire depends upon sign size, and successfully reduces to half the sampling frequency. With its wavelet analogue, since anyway we’re doing a loop over frequencies, we’d as nicely resolve which frequencies to research.

Firstly, the vary of frequencies of curiosity may be decided operating the DFT. The subsequent query, then, is about granularity. Right here, I’ll be following the advice given in Vistnes’ ebook, which is predicated on the relation between present frequency worth and wavelet scale, Okay.

Iteration over frequencies is then applied as a loop:

wavelet_grid <- operate(x, Okay, f_start, f_end, fs) {
  # downsample evaluation frequency vary
  # as per Vistnes, eq. 14.17
  num_freqs <- 1 + log(f_end / f_start)/ log(1 + 1/(8 * Okay))
  freqs <- seq(f_start, f_end, size.out = flooring(num_freqs))
  
  remodeled <- torch_zeros(
    num_freqs, dim(x)[1],
    dtype = torch_cfloat()
    )
  for(i in 1:num_freqs) {
    w <- wavelet_transform_fourier(x, freqs[i], Okay, fs)
    remodeled[i, ] <- w
  }
  listing(remodeled, freqs)
}

Calling wavelet_grid() will give us the evaluation frequencies used, along with the respective outputs from the Wavelet Remodel.

Subsequent, we create a utility operate that visualizes the outcome. By default, plot_wavelet_diagram() shows the magnitude of the wavelet-transformed sequence; it may possibly, nonetheless, plot the squared magnitudes, too, in addition to their sq. root, a way a lot really useful by Vistnes whose effectiveness we’ll quickly have alternative to witness.

The operate deserves a couple of additional feedback.

Firstly, similar as we did with the evaluation frequencies, we down-sample the sign itself, avoiding to counsel a decision that’s not truly current. The components, once more, is taken from Vistnes’ ebook.

Then, we use interpolation to acquire a brand new time-frequency grid. This step might even be needed if we hold the unique grid, since when distances between grid factors are very small, R’s picture() might refuse to just accept axes as evenly spaced.

Lastly, word how frequencies are organized on a log scale. This results in rather more helpful visualizations.

plot_wavelet_diagram <- operate(x,
                                 freqs,
                                 grid,
                                 Okay,
                                 fs,
                                 f_end,
                                 kind = "magnitude") {
  grid <- swap(kind,
    magnitude = grid$abs(),
    magnitude_squared = torch_square(grid$abs()),
    magnitude_sqrt = torch_sqrt(grid$abs())
  )

  # downsample time sequence
  # as per Vistnes, eq. 14.9
  new_x_take_every <- max(Okay / 24 * fs / f_end, 1)
  new_x_length <- flooring(dim(grid)[2] / new_x_take_every)
  new_x <- torch_arange(
    x[1],
    x[dim(x)[1]],
    step = x[dim(x)[1]] / new_x_length
  )
  
  # interpolate grid
  new_grid <- nnf_interpolate(
    grid$view(c(1, 1, dim(grid)[1], dim(grid)[2])),
    c(dim(grid)[1], new_x_length)
  )$squeeze()
  out <- as.matrix(new_grid)

  # plot log frequencies
  freqs <- log10(freqs)
  
  picture(
    x = as.numeric(new_x),
    y = freqs,
    z = t(out),
    ylab = "log frequency [Hz]",
    xlab = "time [s]",
    col = hcl.colours(12, palette = "Mild grays")
  )
  most important <- paste0("Wavelet Remodel, Okay = ", Okay)
  sub <- swap(kind,
    magnitude = "Magnitude",
    magnitude_squared = "Magnitude squared",
    magnitude_sqrt = "Magnitude (sq. root)"
  )

  mtext(aspect = 3, line = 2, at = 0, adj = 0, cex = 1.3, most important)
  mtext(aspect = 3, line = 1, at = 0, adj = 0, cex = 1, sub)
}

Let’s use this on a real-world instance.

An actual-world instance: Chaffinch’s track

For the case examine, I’ve chosen what, to me, was essentially the most spectacular wavelet evaluation proven in Vistnes’ ebook. It’s a pattern of a chaffinch’s singing, and it’s obtainable on Vistnes’ web site.

url <- "http://www.physics.uio.no/pow/wavbirds/chaffinch.wav"

obtain.file(
 file.path(url),
 destfile = "/tmp/chaffinch.wav"
)

We use torchaudio to load the file, and convert from stereo to mono utilizing tuneR’s appropriately named mono(). (For the sort of evaluation we’re doing, there isn’t a level in protecting two channels round.)

library(torchaudio)
library(tuneR)

wav <- tuneR_loader("/tmp/chaffinch.wav")
wav <- mono(wav, "each")
wav

Wave Object
    Variety of Samples:      1864548
    Length (seconds):     42.28
    Samplingrate (Hertz):   44100
    Channels (Mono/Stereo): Mono
    PCM (integer format):   TRUE
    Bit (8/16/24/32/64):    16 

For evaluation, we don’t want the entire sequence. Helpfully, Vistnes additionally printed a advice as to which vary of samples to research.

waveform_and_sample_rate <- transform_to_tensor(wav)
x <- waveform_and_sample_rate[[1]]$squeeze()
fs <- waveform_and_sample_rate[[2]]

# http://www.physics.uio.no/pow/wavbirds/chaffinchInfo.txt
begin <- 34000
N <- 1024 * 128
finish <- begin + N - 1
x <- x[start:end]

dim(x)

[1] 131072

How does this look within the time area? (Don’t miss out on the event to really pay attention to it, in your laptop computer.)

df <- information.body(x = 1:dim(x)[1], y = as.numeric(x))
ggplot(df, aes(x = x, y = y)) +
  geom_line() +
  xlab("pattern") +
  ylab("amplitude") +
  theme_minimal()

Now, we have to decide an inexpensive vary of study frequencies. To that finish, we run the FFT:

On the x-axis, we plot frequencies, not pattern numbers, and for higher visibility, we zoom in a bit.

bins <- 1:dim(F)[1]
freqs <- bins / N * fs

# the bin, not the frequency
cutoff <- N/4

df <- information.body(
  x = freqs[1:cutoff],
  y = as.numeric(F$abs())[1:cutoff]
)
ggplot(df, aes(x = x, y = y)) +
  geom_col() +
  xlab("frequency (Hz)") +
  ylab("magnitude") +
  theme_minimal()

Based mostly on this distribution, we are able to safely prohibit the vary of study frequencies to between, roughly, 1800 and 8500 Hertz. (That is additionally the vary really useful by Vistnes.)

First, although, let’s anchor expectations by making a spectrogram for this sign. Appropriate values for FFT dimension and window dimension have been discovered experimentally. And although, in spectrograms, you don’t see this completed typically, I discovered that displaying sq. roots of coefficient magnitudes yielded essentially the most informative output.

fft_size <- 1024
window_size <- 1024
energy <- 0.5

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

spec <- spectrogram(x)
dim(spec)

[1] 513 257

Like we do with wavelet diagrams, we plot frequencies on a log scale.

bins <- 1:dim(spec)[1]
freqs <- bins * fs / fft_size
log_freqs <- log10(freqs)

frames <- 1:(dim(spec)[2])
seconds <- (frames / dim(spec)[2])  * (dim(x)[1] / fs)

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

The spectrogram already reveals a particular sample. Let’s see what may be completed with wavelet evaluation. Having experimented with a couple of completely different Okay, I agree with Vistnes that Okay = 48 makes for a wonderful alternative:

f_start <- 1800
f_end <- 8500

Okay <- 48
c(grid, freqs) %<-% wavelet_grid(x, Okay, f_start, f_end, fs)
plot_wavelet_diagram(
  torch_tensor(1:dim(grid)[2]),
  freqs, grid, Okay, fs, f_end,
  kind = "magnitude_sqrt"
)

The achieve in decision, on each the time and the frequency axis, is completely spectacular.

Thanks for studying!

Photograph by Vlad Panov on Unsplash