Generate a synthetic 1D signal with known ground truth

Builds a synthetic spectrum (a smooth baseline plus a number of peaks plus noise) with a known true baseline and known individual peak areas, useful for testing baseline-estimation code or trying out psalsa() without real data. This is what tune_psalsa() uses internally to build its synthetic tuning pool, matched to the characteristics of a real example spectrum; called directly, it is a standalone generator.

Usage

gen_synthetic_1d(
  n = 1000,
  density = 0.02,
  fwhm_range = c(10, 30),
  csnr = 0.03,
  peak_shape = c("gaussian", "gex"),
  A = 1,
  seed = 1,
  cap_density = TRUE,
  min_spacing_mult = 2
)

Arguments

n

Number of points.

density

Peaks per point; the number of peaks is max(3, round(density * n)) (further reduced by cap_density).

fwhm_range

c(lo, hi): each peak's full width at half maximum (in points) is drawn uniformly from this range.

csnr

Noise scale, roughly the noise standard deviation as a fraction of the signal amplitude A. 0 gives an exactly noise-free signal.

peak_shape

Either "gaussian" or "gex" (an asymmetric, exponentially-modified peak shape typical of chromatography).

A

Amplitude scale for both the baseline and the peaks.

seed

Random seed, for reproducibility.

cap_density

If TRUE (the default), reduces the peak count so peaks have room to stay non-overlapping (density and fwhm_range themselves are left untouched – only how many peaks are actually placed). Peak centers are still placed uniformly at random, so this lowers the odds of overlap rather than guaranteeing it; set to FALSE to allow arbitrarily dense, overlapping peaks.

min_spacing_mult

Only used when cap_density = TRUE: the peak count is capped so the average spacing between peaks is at least min_spacing_mult times the widest FWHM in fwhm_range.

Value

A list with:

y

the synthetic signal, baseline + peaks + noise.

baseline, peaks, noise

the individual components of y.

sigma

the (possibly position-dependent) noise standard deviation used to generate noise.

peak_info

a data frame with one row per individual peak: height, fwhm, center, the index window lo:hi where that peak alone exceeds 0.1% of its own height, and area (its exact true area, sum() of that peak's own values over the full signal).

Examples

sig <- AlpsNMR:::gen_synthetic_1d(n = 500, seed = 1)
plot(sig$y, type = "l")
lines(sig$baseline, col = "red")

result <- AlpsNMR:::psalsa(sig$y, lambda = 1e6)
lines(result$baseline, col = "blue", lty = 2)