Delta F DCT Normalization

Usage

norm_dct_deltaF(
  .data,
  ...,
  .token_id_col,
  .by = NULL,
  .param_col = NULL,
  .drop_orig = FALSE,
  .names = "{.formant}_df",
  .silent = FALSE
)

Arguments

.data: A data frame containing vowel formant data
...: <tidy-select> One or more unquoted expressions separated by commas. These should target the vowel formant data columns.
.token_id_col: <data-masking> A column that identifies token ids.
.by: <tidy-select> A selection of columns to group by. Typically a column of speaker IDs.
.param_col: A column identifying the DCT parameter number.
.drop_orig: Should the originally targeted columns be dropped.
.names: A glue::glue() expression for naming the normalized data columns. The "{.formant}" portion corresponds to the name of the original formant columns.
.silent: Whether or not the informational message should be printed.

Value

A data frame of Delta F normalized DCT coefficients.

Details

$$ \hat{F}_{ij} = \frac{F_{ij}}{S} $$ $$ S = \frac{1}{MN}\sum_{i=1}^M\sum_{j=1}^N \frac{F_{ij}}{i-0.5} $$

Where

$\hat{F}$ is the normalized formant
$i$ is the formant number
$j$ is the token number

References

Johnson, K. (2020). The $\Delta$F method of vocal tract length normalization for vowels. Laboratory Phonology: Journal of the Association for Laboratory Phonology, 11(1), Article 1. doi:10.5334/labphon.196

Examples

library(tidynorm)
library(dplyr)
ggplot2_inst <- require(ggplot2)

speaker_dct <- speaker_tracks |>
  reframe_with_dct(
    F1:F3,
    .by = speaker,
    .token_id_col = id,
    .time_col = t
  )

# Normalize DCT coefficients
speaker_dct_norm <- speaker_dct |>
  norm_dct_deltaF(
    F1:F3,
    .by = speaker,
    .token_id_col = id,
    .param_col = .param
  )
#> Normalization info
#> • normalized with `tidynorm::norm_dct_deltaF()`
#> • normalized `F1`, `F2`, and `F3`
#> • normalized values in `F1_df`, `F2_df`, and `F3_df`
#> • token id column: `id`
#> • DCT parameter column: `.param`
#> • grouped by `speaker`
#> • within formant: FALSE
#> • (.formant - 0)/mean(.formant/(.formant_num - 0.5), na.rm = T)
#> 

# Apply average and apply inverse dct
# to plot tracks
track_norm_means <- speaker_dct_norm |>
  summarise(
    .by = c(speaker, vowel, .param),
    across(
      ends_with("_df"),
      mean
    )
  ) |>
  reframe_with_idct(
    ends_with("_df"),
    .by = speaker,
    .token_id_col = vowel,
    .param_col = .param
  )


if (ggplot2_inst) {
  track_norm_means |>
    ggplot(
      aes(F2_df, F1_df, color = speaker)
    ) +
    geom_path(
      aes(
        group = interaction(speaker, vowel)
      )
    ) +
    scale_x_reverse() +
    scale_y_reverse() +
    scale_color_brewer(palette = "Dark2") +
    coord_fixed()
}