spectrochempy.find_peaks

find_peaks(dataset, height=None, window_length=3, threshold=None, distance=None, prominence=None, width=None, wlen=None, rel_height=0.5, plateau_size=None, use_coord=True, as_result=False)[source]

Find and analyze peaks in spectroscopic data with advanced filtering options.

This function extends scipy.signal.find_peaks by adding spectroscopy-specific features like coordinate system awareness and unit handling. It performs peak detection through local maxima analysis and supports various filtering criteria to identify significant peaks.

Parameters:

  • dataset (NDDataset) – Input dataset containing spectral data. Must be 1D or 2D with len(X.y) == 1.

  • height (float or array-like, optional) – Minimum and/or maximum peak height criteria. Can be specified as: - Single value for minimum height - Tuple (min, max) for height range - Array matching x for position-dependent criteria

  • window_length (int, default: 3) – Window size for peak interpolation. Must be odd. Larger values provide smoother interpolation but may miss narrow peaks.

  • threshold (float or array-like, optional) – Minimum height difference between peak and neighboring points. Useful for filtering out noise-related peaks.

  • distance (float, optional) – Minimum separation between peaks. Peaks closer than this are filtered based on their prominence.

  • prominence (float or array-like, optional) – Required prominence (height above surrounding baseline) of peaks.

  • width (float or array-like, optional) – Required width of peaks. Interpreted as coordinate units if use_coord=True, otherwise as number of points.

  • wlen (int or float, optional) – Window length for prominence calculation. Affects computation speed for large datasets.

  • rel_height (float, default: 0.5) – Relative height for width calculation (0-1 range).

  • plateau_size (float or array-like, optional) – Required size of peak plateau (flat top).

  • use_coord (bool, default: True) – Whether to use coordinate system units instead of array indices.

  • as_result (bool, default: False) – If True, return a PeakFindingResult object. If False, preserve the historical (peaks, properties) tuple return.

Returns:

  • peaks (NDDataset or None) – Dataset containing identified peaks with interpolated positions and heights. Returned when as_result=False.

  • properties (dict or None) – Peak properties including heights, widths, prominences, and more. All values use appropriate units when use_coord=True. Returned when as_result=False.

  • result (PeakFindingResult) – Structured peak finding result. Returned when as_result=True.

Notes

  • Peak positions are refined using quadratic interpolation when window_length > 1

  • The function handles units automatically when use_coord=True

  • For noisy data, consider preprocessing with smoothing functions

Examples

Basic peak finding with a synthetic spectrum: >>> x = np.linspace(0.0, 10.0, 501) >>> y = np.exp(-((x - 3.0) ** 2) / 0.08) + 0.8 * np.exp(-((x - 7.0) ** 2) / 0.12) >>> ds = scp.NDDataset(y, coordset=[scp.Coord(x, title=”x”, units=”cm^-1”)]) >>> peaks, props = ds.find_peaks(height=0.5) >>> len(peaks) 2

Physical-unit spacing constraints are accepted when coordinates carry units: >>> peaks, props = ds.find_peaks(distance=”1 cm^-1”, width=0.2) >>> len(peaks) 2

Return a structured result when a tabular/export representation is useful: >>> result = ds.find_peaks(height=0.5, as_result=True) >>> rows = result.to_dict() >>> len(rows) 2