TF_peak_selection

TF_peak_selection(candidate_signals, candidate_times, varargin)

TF_PEAK_SELECTION Separate TF peaks from noise peaks among candidate prominence-curve peaks

Usage:

[TFpeak_times, noise_peak_times, clustering_idx, clustering_prom_order, lowbw_TFpeaks, clustering_centroids] = …

TF_peak_selection(candidate_signals, candidate_times, ‘Name’, Value, …)

Inputs:

candidate_signals : NxM double - feature matrix (rows = candidates, cols = features) – required candidate_times : Nx2 double - [start_time, end_time] for each candidate (s) – required

Name-Value Pairs:

‘detection_method’ : char - ‘kmeans’ or ‘threshold’ (default: ‘kmeans’) ‘kmeans_class’ : integer - number of k-means clusters (default: 2) ‘prominence_column’ : integer - column of candidate_signals holding log-prominence (default: 1) ‘threshold_percentile’ : double - percentile threshold for ‘threshold’ method, 0-100 (default: 75) ‘lowbw_TFpeaks’ : Nx2 double - pre-flagged low-bandwidth peaks (default: []) ‘verbose’ : logical - print diagnostics (default: true)

Outputs:

TFpeak_times : Nx2 double - [start, end] times of accepted TF peaks (s) noise_peak_times : Nx2 double - [start, end] times of rejected noise peaks (s) clustering_idx : Nx1 double - cluster assignment per candidate clustering_prom_order : 1xK integer - cluster ranking by mean prominence (highest first) lowbw_TFpeaks : Nx2 double - low-bandwidth TFpeaks (pass-through) clustering_centroids : KxM double - cluster centroids (k-means only)

Notes

The k-means path labels the cluster with the highest mean prominence as TF peaks. The threshold path cuts at the requested prominence percentile. TODO: support stage-restricted clustering.

See also: TF_peak_detection, select_signal_TFpeaks, kmeans

∿∿∿ Prerau Laboratory MATLAB Codebase · sleepEEG.org ∿∿∿