find_time_peaks

find_time_peaks(fpeak_proms, fpeak_freqs, fpeak_bandwidths, fpeak_bandwidth_bounds, stimes, varargin)

PEAK PROMINENCE CALCULATION OF PROMINENCE CURVE IN THE TIME DOMAIN

🔶

function used to calculate the peak prominence of the extracted prominence curve in the time domain. The prominence curve time series, ffreq, and fwidth can be calculated using the find_frequency_peaks.m function that uses findpeaks in the frequency domain.

Usage: [ tpeak_proms, tpeak_times, tpeak_widths, fwidth_at_tpeak, sd_ffreq, sd_fwidth, tpeak_intervals ] = …

find_time_peaks(prominence_curve, ffreq, fwidth, stimes, stages, stage_times, ‘<flag#1>’,<arg#1>…’<flag#n>’,<arg#n>);

##### Declared Inputs: [[[NEEDS UPDATES!!!]]]

The following variables are derived from find_frequency_peaks.m

• prominence_curve: This is the prominence of the peaks

  detected on the normalized spectrum at each time slice.

• ffreq: The frequency of the peaks detected on the

  normalized spectrum at each time slice.

• fwidth: The width of the peaks detected on the

  normalized spectrum at each time slice.

The following variable can be generated from the multitaper spectrogram function:

• stimes: The time axis vector as used in multitaper

  spectrogram estimation. This vector has the same length as prominence_curve, ffreq, and fwidth. Each sample is the time in seconds for the center of the window used to compute one time slice of the spectrogram. This vector should be in seconds.

Other Parameters:

• stages: Scored sleep stages as a vector. All motion

  artifacts should already be marked as periods of wake prior to calling this function. This is important for detecting spindles during only NREM segments.

• stage_times: The is the vector of times at which every

  new stage starts. This should be in seconds.

##### Optional Inputs:

• ‘smooth_sec’: The number of seconds to smooth the

  prominence curve by. This is implemented with movmean.m default: [0.3]

• ‘only_NREM’: A boolean specifying whether only to

  detect peaks in prominence curve during NREM sleep stages. If false, peaks are also detected during REM sleep. default: True

• ‘MinPeakWidth_sec’: The minimum peak width required for a

  peak to be detected by findpeaks. It is specified in seconds. Can be adjusted to prevent detecting tiny noisy peaks. A default value of 0.3 is chosen considering smoothing and the literature assumption of sleep spindles having durations in the range 0.3-3s. default: [0.3]

• ‘MinPeakDistance_sec’:The minimum distance between detected

  peaks in findpeaks. It is specified in seconds. Peak distance is measured from peak to peak. This corresponds to prior assumption about silence period between peaks. This parameter should match the step size and time window parameters in multitaper spectrogram. If a multitaper spectrogram is estimated with window lenth of 1s, then it is difficult to interpret two peaks that are less than 0.5s apart. However, default is set to be 0s here because usually this parameter doesn’t change results by much. default: [0]

• ‘report_width_scale’: This parameter determines the

  conversion from half height width of detected peaks on the prominence curve to start and end times of possible spindle candidates. The spindle start and end times are derived as

  tpeak_time - (report_width_scale*half_height_width)

  ~

  tpeak_time + (report_width_scale*half_height_width)

  thus, if report_width_scale is set to be 0.5, the width of detected peak will have the same width as the half height width. default = [0.5];

##### Outputs:

• tpeak_proms: The prominence values of peaks detected

  in the input prominence curve.

• tpeak_times: The time points of detected peaks.

  This will be the a 2 column vector in the format: [start_times, end_times]. The times are calculated based on the report_width_scale.

• tpeak_widths: The width of peaks detected in

  the input prominence curve. These values are reported in seconds.

• fwidth_at_tpeak: The width of maximal spectral peak at the

  time slice of a peak detected in the prominence curve.

• sd_ffreq: The standard deviation of the frequencies

  of spectral peaks in the time slices where a peak is detected in the prominence value. These values are computed using ffreqs from find_spectrogram_peaks.m

• sd_fwidth: The standard deviation of the width of

  spectral peaks in the time slices where a peak is detected in the prominence value. These values are computed using fwidth from find_spectrogram_peaks.m

• ffreq_at_tpeak: The frequency of maximal spectral peak at

  the time slice of a peak detected in the prominence curve.

• tpeak_intervals: Intervals between tpeaks, calculated as end

  time of previous peak to start time of next peak. This is the upper bound for peakdistance, which is calculated from peak to peak.