Conversation
### Summary Adds a peak-detection and amplitude-analysis utility for real-time signals. The object uses a normalized `Detection signal` to track four event types — rising trigger crossing (`PEAK_RISING`), falling trigger crossing (`PEAK_FALLING`), local maxima (`PEAK_MAX`), and local minima (`PEAK_MIN`) — based on configurable trigger, reload, and fallback parameters. Amplitude is measured from a separate `Raw signal`, with two available modes: `PeakLocked` or `CycleExtrema`. Adapted from Sofian Audry’s [Plaquette implementation](https://plaquette.org/PeakDetector.html), with additional raw-amplitude tracking. ### Functionality Inputs are: - Detection signal (`float`, ideally normalized and scaled 0 to 1) - Raw signal (`float`, original signal used for amplitude measurement) Parameters are: - Trigger threshold (`float` between 0 and 1. Level used for rising and falling trigger crossings) - Reload threshold (`float` between 0 and 1. Level used to reset / re-arm the detector) - Fallback tolerance (`float` between 0 and 1. Drop percentage between trigger and apex required to confirm maxima and minima) - Amplitude mode (`enum`) — `PeakLocked` or `CycleExtrema` - P2P update (`enum`) — updates peak-to-peak on max, min, or both Outputs are: - Peak max (`boolean`. True when a local maximum is confirmed on the detection signal) - Peak min (`boolean`. True when a local minimum is confirmed on the detection signal) - Peak rising (`boolean`. True when the detection signal crosses upward through the trigger threshold) - Peak falling (`boolean`. True when the detection signal crosses downward through the trigger threshold) - Peak-to-peak (`float`. Held raw peak-to-peak amplitude) - Max above threshold (`float`. Held raw amplitude above the rising trigger baseline) - Min below threshold (`float`. Held raw amplitude below the falling trigger baseline) Modes explained PeakLocked: - Uses raw values paired with confirmed detect peaks - Suitable when amplitude should follow the peaks confirmed by the detection signal CycleExtrema: - Uses raw extrema over two half-cycles: rising to falling for the upper excursion, and falling to next rising for the lower excursion - Suitable when amplitude should follow the full raw waveform shape between trigger crossings ### Implementation notes Uses a custom `PeakDetector` class stored in `3rdparty/extras` (to be reused in Respiration and EDA processes). Raw baselines are interpolated at trigger crossings so amplitude is measured relative to the estimated raw value at the threshold crossing, not only at the nearest sample.
Removed CycleExtrema mode from PeakDetection and simplified the object to a single detect-based amplitude path. peak_to_peak, max_above_threshold, and min_below_threshold now all follow the detect-confirmed logic, while a new raw_peak_to_peak output is computed independently from raw slope-based turning points.
TestingTesting for main peak detection logic was already done in previous pull request #44. Code for this part has not changed. Here, I only tested the new amplitude analysis functions. Was tested on Linux (booted on a T2 chip, 2019 MacBook Pro) Example file : https://drive.google.com/file/d/19X2KxdvfEoWtXr1ylX8LNEi4I0KpySqP/view?usp=sharing Tested with live data from Sensors2OSC (acceleration). The live data was then recorded and saved with CSV recorder, and output from the Peak Detection object was compared to output from python offline processing. 
comparison of ossia object output and python output Python file |
|
Code looks pretty good! You should rebase on origin/master to make sure that everything is correct before merging as right now there's a few commits in master and not in here |
2 participants
Summary
Adds a peak-detection and amplitude-analysis utility for real-time signals. The object uses a normalized
Detection signalto track four event types — rising trigger crossing (PEAK_RISING), falling trigger crossing (PEAK_FALLING), local maxima (PEAK_MAX), and local minima (PEAK_MIN) — based on configurable trigger, reload, and fallback parameters. Amplitude is measured from a separateRaw signal.The object outputs threshold-relative amplitudes (
Peak-to-peak,Max above threshold,Min below threshold) using raw values paired with confirmed detect peaks, and also provides an independentRaw peak-to-peakoutput computed directly from raw turning points.The peak-detection logic is adapted from Sofian Audry’s Plaquette implementation, while the raw-signal amplitude processing was added in this object.
Functionality
Inputs are:
float, ideally normalized and scaled 0 to 1)float, original signal used for amplitude measurement)Parameters are:
floatbetween 0 and 1. Level used for rising and falling trigger crossings)floatbetween 0 and 1. Level used to reset / re-arm the detector)floatbetween 0 and 1. Drop percentage between trigger and apex required to confirm maxima and minima)enum) — updates detect-based peak-to-peak on max, min, or bothOutputs are:
boolean. True when a local maximum is confirmed on the detection signal)boolean. True when a local minimum is confirmed on the detection signal)boolean. True when the detection signal crosses upward through the trigger threshold)boolean. True when the detection signal crosses downward through the trigger threshold)float. Held detect-based raw peak-to-peak amplitude)float. Held raw peak-to-peak amplitude computed from raw turning points)float. Held raw amplitude above the rising trigger baseline)float. Held raw amplitude below the falling trigger baseline)Amplitude behavior
Detect-based amplitudes:
Peak-to-peak,Max above threshold, andMin below thresholduse raw values paired with confirmed detect peaksRaw peak-to-peak:
Implementation notes
Uses a custom
PeakDetectorclass stored in3rdparty/extras(to be reused in Respiration and EDA processes).Raw baselines are interpolated at trigger crossings so threshold-relative amplitudes are measured relative to the estimated raw value at the threshold crossing, not only at the nearest sample.