---
title: "spectrogram"
output: rmarkdown::html_vignette
vignette: >
  %\VignetteIndexEntry{spectrogram}
  %\VignetteEngine{knitr::rmarkdown}
  %\VignetteEncoding{UTF-8}
---

```{r, include = FALSE}
knitr::opts_chunk$set(
  collapse = TRUE,
  comment = "#>"
  )

## To silence warnings
require(gsignal, warn.conflicts = F)
require(tuneR, warn.conflicts = F)
```

```{r setup}
library(ggplot2)
library(dplyr, warn.conflicts = F)
library(retimer)
data(mm1)
```

Various packages implement a spectrogram method including `seewave`, `tuneR`, `phonTools`, and `gsignal`. The problem is that they don't return spectrograms in a standard format. With `retimer` you can use the `spectrogram()` wrapper to create a spectrogram with any of these packages and either return a `list` or a `tibble`.

For example, we can use the `tuneR::powerspec()` method like so:

```{r tuner-example}
tuner_spec <- spectrogram(mm1, method = 'tuner', output = 'list')

```

and then plot it in base R:

```{r tuner-plot, fig.width = 7, fig.height = 4, out.width = '100%'}
with(tuner_spec, image(time, freq, amp, col = hcl.colors(20)))

```

Or we could output as a `tibble` (the default output) like so:

```{r phontools-example}
phontools_spec <- spectrogram(mm1, method = 'phontools')

```

and then plot it with `ggplot2`:

```{r phontools-plot, fig.width = 7, fig.height = 3, out.width = '100%'}
phontools_spec |>
    ggplot() +
    geom_raster(aes(t, f, fill = amp), show.legend = F) +
    scale_fill_viridis_c()

```

We should get similar results with different methods:

```{r all-example, fig.width = 7, fig.height = 10, out.width = '100%', warning = FALSE}
list(seewave = 'seewave', phontools = 'phontools', tuner = 'tuner', gsignal = 'gsignal') |>
    lapply(\(method) spectrogram(mm1, method = method)) |>
    bind_rows(.id = 'method') |>
    mutate(method = factor(method, levels = c('seewave', 'phontools', 'tuner', 'gsignal'))) |>
    group_by(method) |>
    mutate(amp = amp - min(amp),
           amp = amp / max(amp)) |>
    ungroup() |>
    ggplot() +
    geom_raster(aes(t, f, fill = amp), show.legend = F) +
    facet_grid(rows = vars(method)) +
    scale_fill_viridis_c()

```