A spectrogram displaying time-frequency representation of a signal as a heatmap. It shows how the frequency content of a signal changes over time, with color intensity representing the amplitude or power at each time-frequency point. Essential for analyzing non-stationary signals where frequency characteristics vary, revealing patterns invisible in time-domain or frequency-domain views alone.
#' anyplot.ai
#' spectrogram-basic: Spectrogram Time-Frequency Heatmap
#' Library: ggplot2 3.5.1 | R 4.4.1
#' Quality: 87/100 | Created: 2026-09-09
library(ggplot2)
library(ragg)
set.seed(42)
# --- Theme tokens -----------------------------------------------------------
THEME <- Sys.getenv("ANYPLOT_THEME", "light")
PAGE_BG <- if (THEME == "light") "#FAF8F1" else "#1A1A17"
ELEVATED_BG <- if (THEME == "light") "#FFFDF6" else "#242420"
INK <- if (THEME == "light") "#1A1A17" else "#F0EFE8"
INK_SOFT <- if (THEME == "light") "#4A4A44" else "#B8B7B0"
# --- Data: synthetic linear chirp signal ------------------------------------
sample_rate <- 4000
duration <- 4
n_samples <- sample_rate * duration
t <- seq_len(n_samples) / sample_rate
freq_start <- 100
freq_end <- 800
instantaneous_phase <- 2 * pi * (freq_start * t + (freq_end - freq_start) * t^2 / (2 * duration))
signal <- sin(instantaneous_phase) + 0.15 * rnorm(n_samples)
# --- Short-time Fourier transform (Hann window, 75% overlap) ---------------
window_size <- 512
hop <- 128
hann_window <- 0.5 - 0.5 * cos(2 * pi * (0:(window_size - 1)) / (window_size - 1))
n_windows <- floor((n_samples - window_size) / hop) + 1
n_freq_bins <- window_size %/% 2 + 1
freqs <- (0:(n_freq_bins - 1)) * sample_rate / window_size
times <- ((0:(n_windows - 1)) * hop + window_size / 2) / sample_rate
power_db <- matrix(0, nrow = n_windows, ncol = n_freq_bins)
for (i in seq_len(n_windows)) {
start <- (i - 1) * hop + 1
segment <- signal[start:(start + window_size - 1)] * hann_window
spectrum <- fft(segment)[1:n_freq_bins]
power_db[i, ] <- 20 * log10(Mod(spectrum) + 1e-6)
}
# Light separable 3-tap smoothing (time then frequency) quiets salt-and-pepper
# noise-floor speckle without blurring the ridge, which spans many bins.
smooth_1d <- function(v) {
n <- length(v)
c(v[1], (v[1:(n - 2)] + v[2:(n - 1)] + v[3:n]) / 3, v[n])
}
power_db <- t(apply(power_db, 1, smooth_1d))
power_db <- apply(power_db, 2, smooth_1d)
dynamic_range_db <- 48
power_db <- pmax(power_db, max(power_db) - dynamic_range_db)
max_freq_display <- 1000
freq_display_idx <- which(freqs <= max_freq_display)
spec_df <- data.frame(
time = rep(times, times = length(freq_display_idx)),
frequency = rep(freqs[freq_display_idx], each = n_windows),
power = as.vector(power_db[, freq_display_idx])
)
# Ridge trace: the peak-power frequency at each time step, from the real STFT
# data (not a fitted curve) — a second ggplot2 layer beyond the raster.
ridge_df <- data.frame(
time = times,
frequency = freqs[freq_display_idx][apply(power_db[, freq_display_idx, drop = FALSE], 1, which.max)]
)
overlap_pct <- round((window_size - hop) / window_size * 100)
# --- Plot ---------------------------------------------------------------
title_str <- "spectrogram-basic · r · ggplot2 · anyplot.ai"
title_fontsize <- round(12 * min(1.0, 67 / nchar(title_str)))
subtitle_str <- sprintf(
"Hann window, %d samples · %d-sample hop (%d%% overlap) · %d dB dynamic range",
window_size, hop, overlap_pct, dynamic_range_db
)
p <- ggplot(spec_df, aes(x = time, y = frequency, fill = power)) +
geom_raster() +
geom_line(
data = ridge_df, aes(x = time, y = frequency),
inherit.aes = FALSE, color = INK, linewidth = 0.35, alpha = 0.55
) +
scale_fill_gradient(
low = "#009E73", high = "#4467A3",
name = "Power (dB)",
guide = guide_colorbar(barwidth = unit(0.35, "cm"), barheight = unit(6, "cm"))
) +
scale_x_continuous(expand = c(0, 0), breaks = seq(0, duration, 1)) +
scale_y_continuous(expand = c(0, 0), breaks = seq(0, max_freq_display, 200)) +
labs(x = "Time (s)", y = "Frequency (Hz)", title = title_str, subtitle = subtitle_str) +
theme_minimal(base_size = 8) +
theme(
plot.background = element_rect(fill = PAGE_BG, color = PAGE_BG),
panel.background = element_rect(fill = PAGE_BG, color = NA),
panel.grid = element_blank(),
axis.ticks = element_blank(),
axis.title = element_text(color = INK, size = 10),
axis.text = element_text(color = INK_SOFT, size = 8),
plot.title = element_text(color = INK, size = title_fontsize),
plot.subtitle = element_text(color = INK_SOFT, size = 8),
legend.background = element_rect(fill = ELEVATED_BG, color = INK_SOFT),
legend.text = element_text(color = INK_SOFT, size = 8),
legend.title = element_text(color = INK, size = 10),
plot.margin = margin(10, 10, 10, 10)
)
# --- Save -------------------------------------------------------------------
ggsave(
filename = sprintf("plot-%s.png", THEME),
plot = p,
device = ragg::agg_png,
width = 8,
height = 4.5,
units = "in",
dpi = 400
)
Runnable source as JSON, for any HTTP client: https://api.anyplot.ai/specs/spectrogram-basic/ggplot2/code. Any spec id and library id listed in llms-full.txt fit the same URL shape; every URL below is complete and callable.
{
"spec_id": "spectrogram-basic",
"language": "r",
"library": "ggplot2",
"page": "https://anyplot.ai/spectrogram-basic/r/ggplot2",
"hub": "https://anyplot.ai/spectrogram-basic",
"code_json": "https://api.anyplot.ai/specs/spectrogram-basic/ggplot2/code",
"spec_json": "https://api.anyplot.ai/specs/spectrogram-basic",
"render_light_png": "https://storage.googleapis.com/anyplot-images/plots/spectrogram-basic/r/ggplot2/plot-light.png",
"render_dark_png": "https://storage.googleapis.com/anyplot-images/plots/spectrogram-basic/r/ggplot2/plot-dark.png",
"quality_score": 87.0,
"license": "MIT",
"guide": "https://anyplot.ai/llms.txt"
}Part of Spectrogram Time-Frequency Heatmap on anyplot.ai.