by Alina Volkova -

Product I – Embodied Resonance – Hardware Selection and System Setup

RECAP:
Embodied Resonance investigates how the body of a person with lived experience of war responds to auditory triggers that recall traumatic events, and how these responses can be expressed through sound. The heart is central to this project, as cardiac activity – particularly heart rate variability (HRV)—provides detailed insight into stress regulation and autonomic nervous system dynamics associated with post-traumatic stress disorder (PTSD).

The conceptual direction of the work is shaped by my personal experience of the war in Ukraine and an interest in trauma physiology. I was drawn to the idea that trauma leaves measurable traces in the body—signals that often remain inaccessible through language but can be explored scientifically and translated into sonic form. This approach was influenced by The Body Keeps the Score by Bessel van der Kolk, which emphasizes embodied memory and non-verbal manifestations of trauma.

In the previous semester, the project focused on exploratory work with existing physiological datasets. A large open-access dataset on stress-induced myocardial ischemia was used to study cardiac behavior under rest, stress, and recovery conditions. Although not designed specifically for PTSD research, the dataset includes participants with PTSD, anxiety-related disorders, and cardiovascular conditions, offering a diverse basis for analysis.

During this phase, Python tools based on the NeuroKit2 library were developed to compute time- and frequency-domain HRV metrics from ECG recordings. Additional scripts transformed these parameters into MIDI patterns and continuous controller (CC) data for sound synthesis and composition. Initial experiments with real-time HRV streaming were also conducted, but they revealed significant limitations: many HRV metrics require long analysis windows and are computationally demanding, making them unsuitable for stable real-time sonification.

In the current semester, corresponding to the Product phase, the project transitions from simulation-based exploration to work with my own body. During earlier presentations, concerns were raised regarding the ethical implications of experiments that could potentially lead to re-traumatization, particularly when involving other participants with war-related trauma. In response, I decided not to extrapolate the experiment to other Ukrainians at this stage and to limit the investigation to my own physiological responses.

Furthermore, instead of exposing myself to recorded sirens at arbitrary times, I chose to record my ECG during the weekly civil defense siren tests that take place every Saturday in Graz. This context offers a meaningful contrast: for most residents of Austria, the siren test is a routine element of everyday life, largely stripped of emotional urgency. For someone with lived experience of war, however, the same sound carries associations of immediate danger. By situating the recordings within this real, socially normalized setting, the project examines how a familiar public signal can produce profoundly different embodied responses depending on personal history.

Before starting the experimental recordings, it was necessary to select and acquire appropriate sensors and a microcontroller. Prior to purchase, a short survey of available biosensing hardware was conducted, with particular attention paid to signal quality, availability of documentation, and the existence of example projects demonstrating practical use. An additional criterion was whether the sensors had been previously employed in projects related to heart rate variability (HRV) analysis.

For ECG acquisition, the DFRobot Gravity Heart Rate Monitor Sensor was selected. This sensor offered a favorable balance between cost and functionality, providing all required cables as well as disposable electrodes. Importantly, it had been used in a well-documented HRV-focused project, which served as a valuable technical reference during development and troubleshooting. In addition to ECG, a galvanic skin response (GSR) sensor from Seeed Studio was included to explore changes in skin conductance as an additional physiological marker of stress. While GSR was not part of the previous semester’s research, it was included experimentally to assess whether this modality could provide complementary information. At this stage, the structure and usefulness of GSR data were not yet fully predictable and were treated as exploratory. As a microcontroller, the Arduino MKR WiFi 1010 was chosen. 

The full list of acquired components is as follows:

  • Arduino MKR WiFi 1010
  • DFRobot Gravity Heart Rate Monitor Sensor (ECG)
  • DFRobot Disposable ECG Electrodes
  • Seeed Studio GSR Sensor
  • Seeed Studio 4-pin Male Jumper to Grove Conversion Cable
  • Breadboard (400 holes)
  • Male-to-male jumper wires
  • Male-to-female jumper wires
  • Potentiometers (100 kΩ)
  • LiPo Battery 1S 3.7 V 500 mAh (not used in final setup)

The total cost of the hardware provides amounted to approximately 80 EUR. The initial motivation for this choice was the possibility of wireless data transmission via WiFi or Bluetooth. In practice, however, wireless communication was not required. Due to the high motion sensitivity of both ECG and GSR sensors, recordings had to be performed in a largely static position, making a wired USB connection to the computer sufficient. For this reason, a battery intended for mobile operation was ultimately not used.

For software configuration, the Arduino IDE was installed. Although I had prior experience working with Arduino hardware several years ago, the interface had changed significantly. To support the Arduino MKR WiFi 1010, the SAMD Boards package was additionally installed via the Boards Manager. After software setup, all components were connected according to a simple wiring scheme that required no additional electronic elements. 

Figure 1. Wiring diagram of the experimental setup with Arduino MKR WiFi 1010, ECG sensor, and GSR sensor.

The Arduino ground (GND) was connected to the ground rail of the breadboard, and the 5 V output was connected to the power rail.

The ECG sensor was connected as follows:

GND (black wire) → ground rail on the breadboard

VCC (red wire) → 5 V power rail

Signal output (blue wire) → analog input A1 on the Arduino

The GSR sensor was connected as follows:

GND (black wire) → ground rail on the breadboard

VCC (red wire) → 3.3 V output on the Arduino

Signal output (yellow wire) → analog input A2 on the Arduino

Figure 2 illustrates the complete wiring configuration of the system, including the Arduino MKR WiFi 1010, ECG sensor, GSR sensor, and breadboard power distribution.

Figure 2. Physical hardware configuration used for ECG and GSR data recording.

by Alina Volkova -

Critical review of “Microbiophonic Emergences” by Adomas Palekas (Master’s Thesis, Institute of Sonology, 2024)

Adomas Palekas’s master’s thesis, entitled Microbiophonic Emergences, could be described as an interdisciplinary mixture of artistic reflection, philosophical speculation, and experimental sound practice. Combining ecological thought and artistic research, the text examines the relationship between sound, life, and perception. Drawing upon the Gaia hypothesis and Goethean science, the author advances a more sensitive and ethical mode of listening wherein the boundaries between the art and scientific observation are dissolved.

Overall, the presentation of the work is considerate and visually well-structured, though it significantly deviates from academic conventions. A clearly defined research question, hypothesis, or structured methodology is lacking. Instead, the text comes across as a long essay on listening, nature, and non-human agency. Its first half is dedicated to theoretical reflections on sound as a living force, while the second half introduces a series of artistic experiments and installations entitled Kwass Fermenter, Microbial Music I–III (On Bread, Compost and Haze, Aerials), Rehydration, Infection, Spectrum of Mutations: Myosin III and Kwassic Motion. According to the author, these works constitute a coherent artistic ecosystem in which microorganisms and sonic feedback interact.

More conceptually, this framing of sonification as bi-directional means that sound should not just be generated from biological data but is also to be sent back into the system and used to affect it. Conceptually, this approach seeks to transform sonification into a dialogue rather than a representation. This claim of originality, however, feels somewhat overstated: bi-directional or feedback-based sonification has been explored conceptually and practically by many artists and researchers before him, mostly within the frames of bio-art and ecological sound practices. Palekas himself mentions only one precedent when, in fact, there exists a wide range of comparable works dealing with translating biological activity into sound and then re-introducing it into the same system. His treatment of the topic is therefore limited and without deep contextual awareness, giving the impression of a rediscovery of ideas that are well conceptualized in the discipline.

The artistic independence of the thesis and a strong, personal vision are explicit. Palekas’s voice is consistent; his writing also reflects genuine curiosity and sensitivity. But it is this very independence that alienates his research from the broader academic and artistic discourse. One misses the dialogue with other practitioners or with theoretical perspectives, except for the few philosophical sources mentioned above. The limited literature review weakens the credibility of his theoretical framework and makes it difficult to situate the work within contemporary sound studies or bio-art research.

The structuring of the thesis is much closer to a philosophical narrative than to a scientific report. The chapters are more intuitively than logically connected. Because explicit methodological framing is absent, the reader has to reconstruct the logic of the experiments from poetic descriptions. For example, the sonification tests with fermentation are told in narrative terms, sometimes mentioning sensors, mappings, and feedback without providing detailed diagrams, lists of parameters, or reproducible data.

From a communicational point of view, the thesis is well-written and easy to read. Palekas’s prose is expressive and reflective; his philosophical passages are a pleasure to read. At the same time, this lyricism too often supplants analytical clarity. The experimental results remain fuzzy; the measurements are given “by ear,” not through numerical analysis, and the reader cannot tell whether the effects observed are significant or only subjective impressions.

In scope and depth, the thesis is ambitious but uneven. It tries to combine philosophy, biology, and sound art, but the practical documentation remains superficial. The experiments are deficient in calibration and control conditions, as well as in quantitative evidence. The author himself recognizes that fermentation is hardly predictable and thus difficult to reproduce. But this admission only underlines the fragility of his conclusions. Without a presentation of clear data or even replicable protocols, the whole project remains conceptual rather than empirical.

Partial accuracy and attention to detail: the author provides some information about equipment and process – for example, relative calibration among the CO₂ sensors, use of Arduino, Pure Data, but no consistent system is provided for reporting values, frequencies, and time spans. References made to appendices and videos are incomplete, and none of the referred sound recordings and codes are available. The result is that the project cannot be scientifically evaluated or reproduced.

The section on literature review reflects selectivity: In situating his thought within broader ecological and philosophical frameworks, Palekas barely engages the rich corpus of research on bio-sonification, microbial sensing, and feedback sound systems. The lack of these sources increases the effect of isolation: the thesis feels self-contained rather than in conversation with a field.

This gap between theory and documentation is where the quality of the artifact is questioned. The installations and performances he describes conceptually are incompletely and poorly documented. It is not clear if the works were created for this thesis or collated from previous projects. Without recordings, schematics, or step-by-step documentation available, one cannot evaluate any artistic or technical outcomes. Put differently, Microbiophonic Emergences is a strong artistic and philosophical statement, but it is only a partially successful academic thesis. Its conceptual strength comes from the ethical rethinking of listening, the poetic vision of sound as life, and the attempt to dissolve the hierarchy between observer and observed.

The work unfortunately lacks in methodological rigor, detailed evidence, and sufficient contextual grounding. While Palekas seeks to establish a dialogue between humans and microbes, the outcome is just speculative and remains unverified. This invention of bi-directional sonification is not really new; moreover, the thesis overlooks the numerous past projects that have already elaborated on a similar feedback relationship between sound and living systems. Overall, the work is successful as a reflective, imaginative exploration of sound and ecology but fails as a systematically researched academic document. While the work evokes curiosity and wonder, it requires far stronger methodological and contextual grounding to meet the standards of a master’s thesis.

by Alina Volkova -

Experiment X: Embodied Resonance – Sound Design and Implementation in the DAW Environment

The final stage of the project focused on the integration of HRV-derived MIDI files into a digital audio workstation (Ableton Live) in order to create expressive and physiologically grounded sound textures. This process was divided into two main domains: rhythmic (drum) and melodic (spectral) layers, each mapped to specific HRV parameters.

The rhythmic component was implemented first. MIDI files generated from heart rate data were imported into dedicated drum tracks. A drum sampler loaded with the Boom Kit preset was used as the primary sound generator. This particular preset was selected based on aesthetic considerations and its extended control capabilities, which provided greater flexibility for parameter mapping.

Figure 1. Drum sampler setup with Boom Kit preset in Ableton Live.

To facilitate real-time modulation, Ableton’s stock Echo device was inserted after the drum sampler. Before the drum kit, a MIDI expression control device was placed to route MIDI CC data from the HRV-derived files to specific parameters within the effects chain. 

Figure 2. Full instrument rack configuration for the rhythm layer.

Among the tested mappings, a particularly compelling result was achieved by linking SDNN values to multiple targets simultaneously—namely, the feedback amount, reverb level, and transposition of the drum sounds. RMSSD was mapped to delay time, introducing irregularities and a sense of fragmentation that enriched the rhythmic texture and highlighted moments of local instability in cardiac variability.

The second major component of the sound design involved the melodic layer, constructed using frequency-domain metrics (VLF, LF, HF). MIDI files generated from these bands were distributed across three separate tracks in Ableton Live. 

The VLF data was used to control a bassline, realized through the free software synthesizer Vital. Two basic sine-wave oscillators were used, one of which was tuned an octave above the other to enhance tonal richness.

LF and HF bands were each assigned to their own melodic track, both using identical synthesis architecture. In each case, two Vital instances were grouped within an instrument rack. One synthesizer generated a sinusoidal waveform, while the other used a sawtooth. 

Figure 3. Oscillator configuration in Vital synthesizer using a sine waveform.

Figure 4. Oscillator configuration in Vital synthesizer using a sawtooth waveform.

This grouping enabled the use of Ableton’s chain selector feature to blend between the two timbres. The LF and HF values were routed to dynamically modulate the balance between the oscillators, producing smooth or aggressive harmonic profiles depending on the underlying autonomic activity.

Figure 5. Instrument rack in Ableton Live with activated chain selector.

To integrate the overall sympathovagal balance, the LF/HF ratio was mapped to the chain selector controlling the morphing between the two oscillators.

Figure 6. Full instrument rack setup with MIDI expression control mapped to chain selector.

This approach allowed real-time transformation of the harmonic character in response to physiological state—ranging from calm, sine-dominated tones to sharp, sawtooth-driven textures.

by Alina Volkova -

Experiment IX: Embodied Resonance – From Offline to Online genaration

Our original objective was simple on paper: write a Python utility that would stream real-time heart-rate-variability (HRV) data from a live ECG feed and let a digital-audio workstation turn those numbers into music. The script was expected to read 0.1-second ECG chunks, detect new R-peaks, derive the seven canonical HRV metrics (HR, SDNN, RMSSD, VLF, LF, HF, LF/HF) and hand fresh values to the sound engine every tick. In theory that would allow a performer’s physiology to shape the soundtrack moment-by-moment.

Phase 1 – “print-only” prototype
The first proof-of-concept program did nothing but compute the metrics and dump them to the terminal. Because the offline reference library (hrv_plot.py) already produced correct numbers, the live version mimicked its algorithm line-for-line, apart from using a sliding buffer instead of reading the whole file at once. The console printed something every 0.1 s, CPU usage looked negligible, and we assumed we were done.

Phase 2 – OSC transport and the DAW deterrent
To let a synthesiser listen, we wrapped the same loop with an Open Sound Control sender. The data reached Ableton Live but required an external bridge, custom track routing and an additional plug-in to map OSC to parameters. Latency was fine, usability was not: every DAW session began with ten minutes of cabling virtual ports. We dropped OSC and decided to embed the information in ordinary MIDI where any host can record or map it instantly.

Phase 3 – first real-time MIDI attempt, and the “melting clock”
The next script converted each 0.1-s update into a drum hit on one of sixteen cells; HR controlled velocity, SDNN and RMSSD drove two CC lanes. For the first five seconds everything grooved, then tempo collapsed. Profiling showed peak detection becoming slower and slower: our buffer was ten seconds long, so every tick the detector rescanned an ever-growing vector. With 500 Hz sampling that was half a million points per second.

Phase 4 – carving the code into independent daemons
We split responsibilities into two stand-alone scripts:
• hrv_live_print.py – generate just the metrics once every 0.1 s
• ecg_to_drum_online.py – listen to the stream and build MIDI

The HRV process still lagged behind wall-time, so the music engine received late packets and eventually starved.

Phase 5 – “peak-triggered” optimisation and the HR crisis
To cut CPU we tried a different paradigm: compute HRV only when the detector reported a new R-peak; between peaks resend the previous values. That immediately fixed the frame-rate problem but created another: HR oscillated between plausible and absurd numbers (e.g. 40 → 140 → 55 bpm within one second). The fault was a race condition—the timestamp of the newest peak was sometimes outside the analysis window because the buffer indices were updated before the metric window limits. After correcting the order of operations HR stabilised within ±2 bpm of the offline truth. Unfortunately SDNN, RMSSD and especially the spectral powers still diverged by factors of two to ten.

Why the divergence never vanished

  • Time-domain spread metrics need at least fifty RR intervals; with resting heart-rate that is a full minute, which our 10-s buffer could never deliver, hence wildly inflated variance.
  • Spectral power below 0.15 Hz requires windows ≥ 120 s; shortening the segment erases low-frequency bins, so LF and VLF shrink to almost zero or explode at random.
  • Re-detecting peaks on a moving buffer changes the set of RR intervals at every tick, adding jitter no post-hoc analysis has to face.
  • Any attempt to enlarge the window brings back the original slowdown; shrinking it further removes the physiological meaning altogether.

Take-away – offline wins, HRV is not a live control signal
After weeks of profiling, refactoring and cheating (full-track pre-detection disguised as streaming) the conclusion is clear: classic HRV statistics are intrinsically retrospective. They trade temporal resolution for statistical power. In a concert-length performance the musician would have to wait one–three minutes before a genuine LF/HF change manifests, which is musically useless. Short windows restore immediacy but turn the output into coloured noise and defeat the scientific basis of HRV.

A narrow exception
Plain heart rate—computed from the last RR pair—can be broadcast at 10 Hz with negligible latency, so HR alone might still serve as a slow modulator. It is, however, a single scalar that drifts over tens of seconds; by itself it is too static to drive anything more than subtle filter sweeps.

Final verdict
For faithful physiology-driven sound the only robust path is to run a full offline analysis first or to invent new, deliberately short-window descriptors instead of relying on canonical HRV. The experiment showed that real-time HRV is a conceptual mismatch rather than a coding glitch, and that insight will save us from chasing the same mirage in future projects.

by Alina Volkova -

Experiment VIII: Embodied Resonance – ecg_to_cc.py

After converting heart-rate data into drums and HRV energy into melodic tracks, I needed a tiny helper that outputs nothing but automation. The idea was to pull every physiologic stream—HR, SDNN, RMSSD, VLF, LF, HF plus the LF / HF ratio—and stamp each one into its own MIDI track as CC-1 so that in the DAW every signal can be mapped to a knob, filter or layer-blend. The script must first look at all CSV files in a session, find the global minimum and maximum for every column, and apply those bounds consistently. Timing stays on the same 75 BPM / 0.1 s grid we used for the drums and notes, making one CSV row equal one 1⁄32-note. With that in mind, the core design is just constants, a value-to-CC rescaler, a loop that writes control changes, and a batch driver that walks the folder.

If you want to see the full, visit: https://github.com/ninaeba/EmbodiedResonance

# ─── Default Configuration ───────────────────────────────────────────
DEF_BPM       = 75             # Default tempo
CSV_GRID      = 0.1            # Time between CSV samples (s) ≙ 1/32-note @75 BPM
SIGNATURE     = "8/8"          # Eight-cell bar; enough for pure CC data
CC_NUMBER     = 1              # We store everything on Mod Wheel
CC_TRACKS     = ["HR", "SDNN", "RMSSD", "VLF", "LF", "HF", "LF_HF"]
BEAT_PER_GRID = 1 / 32         # Grid resolution in beats

The first helper, map_to_cc, squeezes any value into the 0-127 range. It is the only place where scaling happens, so if you ever need logarithmic behaviour you change one line here and every track follows.

def map_to_cc(val: float, vmin: float, vmax: float) -> int:
norm = np.clip((val - vmin) / (vmax - vmin), 0, 1)
return int(round(norm * 127))

generate_cc_tracks loads a single CSV, back-fills gaps, creates one PrettyMIDI container and pins the bar length with a time-signature event. Then it iterates over CC_TRACKS; if the column exists it opens a fresh instrument named after the signal and sprinkles CC messages along the grid. Beat-to-second conversion is the same one-liner used elsewhere in the project.

for name in CC_TRACKS:
if name not in df.columns:
continue
vals = df[name].to_numpy()
vmin, vmax = minmax[name]
inst = pm.Instrument(program=0, is_drum=False, name=f"{name}_CC1")

for i in range(len(vals) - 1):
    beat = i * BEAT_PER_GRID * 4          # 1/32-note → quarter-note space
    t    = beat * 60.0 / bpm
    cc_val = map_to_cc(vals[i], vmin, vmax)
    inst.control_changes.append(
        pm.ControlChange(number=CC_NUMBER, value=cc_val, time=t)
    )

midi.instruments.append(inst)

The wrapper main is pure housekeeping: parse CLI flags, list every *.csv, compute global min-max per signal, and then hand each file to generate_cc_tracks. Consistent scaling is guaranteed because min-max is frozen before any writing starts.

minmax: dict[str, tuple[float, float]] = {}
for name in CC_TRACKS:
vals = []
for f in csv_files:
df = pd.read_csv(f, usecols=lambda c: c.upper() == name)
if name in df.columns:
vals.append(df[name].to_numpy())
if vals:
all_vals = np.concatenate(vals)
minmax[name] = (np.nanmin(all_vals), np.nanmax(all_vals))

Each CSV turns into *_cc_tracks.mid containing seven tracks—HR, SDNN, RMSSD, VLF, LF, HF, LF_HF—each packed with CC-1 data at 1⁄32-note resolution. Drop the file into Ableton, assign a synth or effect per track, map the Mod Wheel to whatever parameter makes sense, and the physiology animates the mix in real time.

As a result, we get CC automation that we can easily map to something else or just copy automation somewhere else:

Here is an example of HR from patient 115 mapped from 0 to 127 in CC:

by Alina Volkova -

Experiment VII: Embodied Resonance – ecg_to_notes.py

The next step after the drum generator is a harmonic layer built from frequency-domain HRV metrics.
The script should do exactly what the drum code did for HR: scan all source files, find the global minima and maxima of every band, then map those values to pitches. Very-low-frequency (VLF) energy lives in the first octave, low-frequency (LF) in the second, high-frequency (HF) in the third. Each band is written to its own MIDI track so a different instrument can be assigned later, and every track also carries two automation curves: one CC lane for the band’s amplitude and one for the LF ⁄ HF ratio. In the synth that ratio will cross-fade between a sine and a saw: a large LF ⁄ HF—often a stress marker—makes the tone brighter and scratchier.

Instead of all twelve semitones, the script confines itself to the seven notes of C major or C minor. The mode flips in real time: if LF ⁄ HF drops below two the scale is major, above two it turns minor. Timing is flexible: each band can have its own step length and note duration so VLF moves more slowly than HF.

Below is a concise walk-through of ecg_to_notes.py. Key fragments of the code are shown inline for clarity. If you want to see the full, visit: https://github.com/ninaeba/EmbodiedResonance

# ----- core tempo & bar settings -----
DEF_BPM   = 75      # fixed song tempo
CSV_GRID  = 0.1     # raw data step in seconds
SIGNATURE = "8/8"   # eight cells per bar (fits VLF rhythms)

The script assigns a dedicated time grid, rhythmic value, and starting octave to each band.
Those values are grouped in three parallel constants so you can retune the behaviour by editing one block.

# ----- per-band timing & pitch roots -----
VLF_CELL_SEC = 0.8; VLF_STEP_BEAT = 1/4 ; VLF_NOTE_LEN = 1/4 ; VLF_ROOT = 36  # C2
LF_CELL_SEC  = 0.4; LF_STEP_BEAT  = 1/8 ; LF_NOTE_LEN  = 1/8 ; LF_ROOT  = 48  # C3
HF_CELL_SEC  = 0.2; HF_STEP_BEAT  = 1/16; HF_NOTE_LEN  = 1/16; HF_ROOT  = 60  # C4

Major and minor material is pre-baked as two simple interval lists. A helper translates any normalised value into a scale degree, then into an absolute MIDI pitch; another helper turns the same value into a velocity between 80 and 120 so you can hear amplitude without a compressor.

MAJOR_SCALE = [0, 2, 4, 5, 7, 9, 11]
MINOR_SCALE = [0, 2, 3, 5, 7, 10, 11]

def normalize_note(val, vmin, vmax, scale, root):
    norm  = np.clip((val - vmin) / (vmax - vmin), 0, 0.999)
    step  = scale[int(norm * len(scale))]
    return root + step

def map_velocity(val, vmin, vmax):
    norm = np.clip((val - vmin) / (vmax - vmin), 0, 1)
    return int(round(80 + norm * 40))

signals_to_midi opens one CSV, grabs the four relevant columns, and sets up three PrettyMIDI instruments—one per band. Inside a single loop the three arrays are processed identically, differing only in timing and octave. At each step the script decides whether we are in C major or C minor, converts the current value to a note and velocity, and appends the note to the respective instrument.

for name, vals, cell_sec, step_beat, note_len, root, vmin, vmax, inst in [
    ("VLF", vlf_vals, VLF_CELL_SEC, VLF_STEP_BEAT, VLF_NOTE_LEN, VLF_ROOT, vlf_min, vlf_max, inst_vlf),
    ("LF",  lf_vals,  LF_CELL_SEC,  LF_STEP_BEAT,  LF_NOTE_LEN,  LF_ROOT,  lf_min,  lf_max,  inst_lf),
    ("HF",  hf_vals,  HF_CELL_SEC,  HF_STEP_BEAT,  HF_NOTE_LEN,  HF_ROOT,  hf_min,  hf_max,  inst_hf)
]:
    for i, idx in enumerate(range(0, len(vals), int(cell_sec / CSV_GRID))):
        t        = idx * CSV_GRID
        scale    = MINOR_SCALE if lf_hf_vals[idx] > 2 else MAJOR_SCALE
        pitch    = normalize_note(vals[idx], vmin, vmax, scale, root)
        velocity = map_velocity   (vals[idx], vmin, vmax)
        duration = note_len * 60 / DEF_BPM
        inst.notes.append(pm.Note(velocity, pitch, t, t + duration))

Every note time-stamp is paired with two controller messages: the first transmits the band amplitude, the second the LF / HF stress proxy. Both are normalised to the full 0-127 range so they can drive filters, wavetable morphs, or anything else in the synth.

inst.control_changes.append(pm.ControlChange(CC_SIGNAL, int(round(norm_val  * 127)), t))
inst.control_changes.append(pm.ControlChange(CC_RATIO,  int(round(norm_lf_hf * 127)), t))

The main wrapper gathers all CSVs in the chosen folder, computes global min–max values so every file shares the same mapping, and calls signals_to_midi on each source. Output files are named *_signals_notes.mid, each holding three melodic tracks plus two CC lanes per track. Together with the drum generator this completes the biometric groove: pulse drives rhythm, spectral power drives harmony, and continuous controllers keep the sound evolving.

Here is what we get if we make one mappinf for all patients

An here is example of MIDI files we get if we make individual mapping for each patient separetly





by Alina Volkova -

Experiment VI: Embodied Resonance – ecg_to_drum.py

To generate heart-rate–driven drums I set out to write a script that builds a drum pattern whose density depends on HR intensity. I fixed the tempo at 75 BPM because the math is convenient: 0.1 s of data ≈ one 1/32-note at 75 BPM. I pictured a bar with 16 cells and designed a map that decides which cell gets filled next; the higher the HR, the more cells are activated. Before rendering, a helper scans the global HR minimum and maximum and slices that span into 16 equal zones (an 8-zone fallback is available) so the percussion always scales correctly. For extra flexibility, the script also writes SDNN and RMSSD into the MIDI file as two separate CC automation lanes. In Ableton, I can map those controllers to any parameter, making it easy to experiment with the track’s sonic texture.

Let’s take a closer look at the code. If you want to see the full, visit: https://github.com/ninaeba/EmbodiedResonance

This script converts raw heart-rate data into a drum track.
Every 0.1-second CSV sample lines up with a 1⁄32-note at the fixed tempo of 75 BPM, so medical time falls neatly onto a musical grid. The pulse controls how many cells inside each 16-step bar are filled, while SDNN and RMSSD are embedded as two CC lanes for later sound-design tricks.

Default configuration (all constants live in one block)
 STEP_BEAT = 1 / 32  # grid resolution in beats
 NOTE_LENGTH = 1 / 32  # each hit lasts a 32nd-note
 DEF_BPM = 75    # master tempo
 CSV_GRID = 0.1   # CSV step in seconds
 SEC_PER_CELL = 0.1   # duration of one pattern cell
 SIGNATURE = "16/16" # one bar = 16 cells
 CC_SDNN = 11    # long-term HRV
 CC_RMSSD = 1    # short-term HRV

The spatial order of hits is defined by a map that spreads layers across the bar, so extra drums feel balanced instead of bunching at the start

 CELL_MAP = [0, 8, 4, 12, 2, 10, 6, 14, 1, 9, 5, 13, 3, 11, 7, 15]

generate_rules slices the global HR span into sixteen equal zones and binds each zone to one cell-and-note pair; the last rule catches any pulse above the top threshold

 def generate_rules(hr_min, hr_max, cells):
  thresholds = np.linspace(hr_min, hr_max, cells + 1)[1:]
  rules = [(thr, CELL_MAP[i], 36 + i) for i, thr in enumerate(thresholds)]
  rules[-1] = (np.inf, rules[-1][1], rules[-1][2])
  return rules

Inside csv_to_midi the script first writes two continuous-controller streams—one for SDNN, one for RMSSD—normalised to 0-127 at every grid tick

 sdnn_norm = clip((sdnn − min) / (range), 0, 1)
 rmssd_norm = clip((rmssd − min) / (range), 0, 1)
 add CC 11 with value round(sdnn_norm × 127) at time t
 add CC 1 with value round(rmssd_norm × 127) at time t

A cumulative-pattern list is built so rhythmic density rises without ever muting earlier layers

 CUM_PATTERNS = []
 acc = []
 for (thr, cell, note) in rules:
  acc.append((cell, note)) # keep previous hits
  CUM_PATTERNS.append(list(acc)) # zone n holds n+1 hits

During rendering each bar/cell slot is visited, the current HR is interpolated onto that exact moment, its zone is looked up, velocity is set to the rounded pulse value, and only the cells active in that zone fire

 hr = interp(t_sample, times, hr_vals)
 zone = first i where hr ≤ rules[i].threshold
 vel = int(round(hr))

 for (z_cell, note) in CUM_PATTERNS[zone]:
  if z_cell ≠ cell: continue
  beat = (bar × cells + cell) × STEP_BEAT × 4
  t0 = beat × 60 / bpm
  write Note(pitch=note, velocity=vel, start=t0, end=t0 + NOTE_LENGTH in seconds)

Before any writing starts the entry-point scans the whole folder to gather global minima and maxima for HR, SDNN and RMSSD, guaranteeing identical zone limits and CC scaling across every file in the batch

 hr_min, hr_max = min/max of all HR values
 sdnn_min, sdnn_max = min/max of all SDNN values
 rmssd_min, rmssd_max = min/max of all RMSSD values
 rules = generate_rules(hr_min, hr_max, cells)

Run the converter like this:

 python ecg_to_drum.py path/to/csv_folder --tempo 75 --time-signature 16/16

Each CSV becomes a _drums.mid file that contains a single drum-kit track whose note density follows heart-rate zones and whose CC 11 and CC 1 envelopes mirror long- and short-term variability—ready to animate filters, reverbs or whatever you map them to in the DAW.

Here are examples of generated MIDI files with drums. On the left side, there are drums made with individual mapping for every patient separately, and on the right side, mapping is one for all patients. We can see they resemble our HR graphs.

by Alina Volkova -

Experiment V: Embodied Resonance – understanding data for meaningfull sound mapping

Below are the four participants whose heart-rate-variability traces we have explored.
We deliberately chose them because they sit at clearly different points on the “health–illness” spectrum and therefore give us a compact but vivid physiological palette.

  • Subject 119 is our practical baseline. No cardiac findings, no anxiety-or-depression scores, no PTSD, no “Type-D” personality pattern. Anything we hear or see in this person’s HRV should approximate the dynamics of an uncomplicated, resilient autonomic system.
  • Subject 091 represents a “mind-only” disturbance. The heart itself is structurally sound, but the person carries the so-called Type-D trait (high negative affect, high social inhibition). This makes the autonomic system more reactive to worry or rumination even when the coronary vessels are normal.
  • Subject 044 adds psychological trauma on top of mild, non-obstructive angina. Clinically this participant scores high on anxiety and meets full PTSD criteria. We therefore expect brisk sympathetic surges, slower vagal recovery and more noise in the LF/HF ratio—an autonomic pattern often seen in hyper-arousal states.
  • Subject 005 is the most medically burdened case: non-obstructive angina, endothelial dysfunction, stress-induced ischaemia, plus anxiety, depression, Type-D personality and PTSD. In short, both the mechanical pump and the emotional “software” are under strain, so variability measures are likely compressed and heart-rate plateaus may appear where a healthy person would fluctuate.

Using these four contrasting bodies as our “voices” lets us investigate how the same 6-min rest → 12-min exercise → 6-min recovery protocol is translated into four distinct autonomic narratives—information we will later map into equally distinct sonic textures.

HR

In general, each cardiovascular system answers physical load differently, depending on the baseline autonomic tone, fitness, psychological state, and comorbidities.

sub 115 – “healthy”
The heart “ramps up” slowly. Pulse climbs in stair-like steps with brief dips between peaks—the body is constantly trying to regain homeostasis. After the exercise HR quickly falls almost to baseline. This is typical of a well-trained, adaptive cardiovascular system with a large functional reserve.

sub 091 – “mental / Type D”
Resting HR is slightly above normal, yet overt anxiety is absent. The response is inertial: about a minute after load starts HR jumps from 75 → 91 bpm, then plummets to 76, followed by alternating short rises and falls. When the exercise stops, HR drops to 68 (below the initial value) and only then drifts back to ~75. Such a “swing” may reflect conflicting sympathetic vs. parasympathetic signals: outward calm while inner tension builds and discharges in bursts.

sub 044 – “PTSD”
Even before load the heart is already in “fight-or-flight” mode (~83 bpm). The onset of exercise changes little, but at the third minute HR spikes to 98. It then declines stepwise yet remains high (89–97) even during recovery. The absence of a deep post-exercise dip shows that the parasympathetic “brake” hardly engages— the body struggles to shift into a resting state.

sub 005 – “sick / multiple pathologies”
Baseline HR is 68. At the start of exercise it leaps to 80, after which it very slowly, step-by-step, drifts back toward normal and stays there until the end of the trial. The pattern resembles sub 044 but with lower background stress and slightly better recovery.

In healthy subjects the heart rate rises smoothly in step-like waves during effort and quickly settles back to baseline once the task stops, reflecting a well-balanced push-and-pull between sympathetic “accelerator” activity and parasympathetic (vagal) “brakes.” In patients with cardiac disease, PTSD or Type-D traits the resting rate is already elevated, the response to stress is sharper or more erratic, and the return to baseline is sluggish, showing a system locked in chronic “fight-or-flight” with weaker vagal damping.

Sound-design mapping:

Based on HR, I want to create drum patterns, where the lowest HR generates only kick, and then if HR rises, we can add more percusive elements.

SDNN

SDNN tracks how much the RR-intervals expand and shrink over time.

  • When the value climbs, the spacing between beats becomes more irregular—your heart is “dancing” around the mean to satisfy moment-to-moment demands.
  • When the value sinks, the intervals line up almost like a metronome—either the organism is in very deep rest or the rhythm is held in a tight sympathetic “clamp” and cannot flex.
  • Usually this parameter is calculated for 24 hours, but since we dont have such luxury, we stick to 2min time window

Looking at all four traces, the healthy control shows the widest SDNN swing and more frequent surges during the load phase. This wide dynamic range tells us the cardiac pacemaker is quick to loosen and tighten the rhythm, i.e. it adapts smoothly to the body’s changing needs. By contrast, the clinical subjects operate in a narrower corridor: their SDNN rarely strays far from baseline, signalling that the heart either remains compressed by chronic sympathetic tone or cannot recruit enough parasympathetic “slack” to respond fully.

For the sound-mapping layer, I’d like to add a short delay on the Drums track and control athe mount of feedback and reverb of echoes.

RMSSD

RMSSD is conceptually close to SDNN, but it is calculated as the square root of the mean of the squared differences between successive RR intervals. In essence, RMSSD captures beat-to-beat (“breath-by-breath”) variability, whereas SDNN reflects overall dispersion of intervals within the chosen window. Because of this local focus, we processed RMSSD in a shorter analysis window—30 seconds with a 10-second step—to obtain a curve that is smooth yet sensitive to rapid changes.

The comparative analysis reveals patterns similar to SDNN, with several noteworthy differences. First, in the clinical subjects, the RMSSD range is almost twice as narrow, indicating reduced high-frequency variability—the heart is working in a more “rigid” mode. Second, in both pathological cases, RMSSD rises noticeably toward the end of the protocol: once exercise stops, sharp spikes appear that are virtually absent in the healthy subject. This delayed surge suggests a late engagement of parasympathetic control—the body remains under sympathetic drive for a prolonged period and only during recovery tries to compensate, generating erratic, uneven intervals.

To highlight the heart’s “rigidity,” I plan to map RMSSD to the pitch shift of drums.

Spectral HRV analysis

For spectral HRV analysis the gold-standard window is about five minutes, but in this project we need a compromise between scientific reliability and near-real-time responsiveness. We therefore use a 120-second window with a 60-second step.

VLF

Within such a short segment, the VLF band (0.003–0.04 Hz)—which is thought to reflect very slow regulatory processes such as hormonal release and thermoregulation—cannot be interpreted with the same statistical confidence as in traditional five-minute blocks. Even so, the plots still reveal that surges in VLF power tend to appear just before rises in heart-rate amplitude. In our context that timing may hint at micro-shifts in core temperature or other slow-acting homeostatic mechanisms that prime the cardiovascular system for the upcoming workload.

VLF → Sub-bass “body boil” layer
We will map the slow-acting VLF band to a very low sub-bass drone in the first octave (≈ 30-60 Hz). As VLF power rises the pitch of this bass note is gently shifted upward by a few semitones and a subtle vibrato (slow LFO) is added. The result feels like liquid starting to simmer: higher VLF = hotter “water,” faster wobble, and a slightly higher fundamental. A pre-recorded low-frequency sample (e.g., pitched-down kettle rumble) can sit under the main mix; its playback pitch follows the same VLF curve to reinforce the sensation. When VLF falls, both pitch and vibrato relax, letting the bass settle back to its calm, foundational tone. This approach sonically frames VLF as the deep thermal undercurrent of the body.

LF & HF

Low-Frequency (LF, 0.04–0.15 Hz) and High-Frequency (HF, 0.15–0.40 Hz) power curves mirror the behavior we already saw with SDNN and RMSSD. That is perfectly logical: SDNN and LF both track activity of the sympathetic branch, which dominates under stress—heart rate accelerates, blood pressure rises, LF climbs. Conversely, HF and RMSSD follow the parasympathetic (“vagal”) branch: as the body relaxes, the heart slows, breathing deepens, and HF increases.

Patient 091 delivers a textbook illustration of that theory. At rest his HF overtops LF, but the moment exercise begins the autonomic balance flips—LF jumps above HF and keeps climbing, whereas HF rises more modestly and then drops back near the end, underscoring how hard it is for his system to settle.

By contrast, our other symptomatic patients (044 and 005) show LF dominating HF throughout, signalling a chronically tense autonomic state.

The healthy subject 115 starts with LF and HF almost equal; as effort mounts the gap widens in a smooth, orderly fashion.

Sonically, LF and HF make a natural complement to the VLF layer. We can map their absolute values to pitch in the second octave while using the LF/HF ratio to shape timbre—blending from a pure sine wave toward an edgier saw-like texture. When LF/HF is below 2, the sound stays near sine (calm, vagal tone); as the ratio exceeds 2, it becomes increasingly “saw-toothed,” evoking sympathetic arousal. Also, with this parameter, I would like to control the music scale from major to minor.

In practice, we see 115 and 091 resting in the neutral range and pushing upward only during exercise—the difference being that 115 ascends gradually, whereas 091 leaps. Both glide back to baseline once the task ends.

Patients 044 and 005, however, begin with ratios already high and jittering, telegraphing their persistent sympathetic load.

The next stage will be to develop an algorithm that converts these metrics into MIDI messages that can be mapped to parameters inside a DAW, or, alternatively, to build a SuperCollider or Pure Data patch implementing the same control scheme.

by Alina Volkova -

Experiment IV: Embodied Resonance – plot HRV metrics with python

Before we can compare a “healthy” and a “clinical” heart, we first need a small tool-chain that does three things automatically:

  1. detects each normal-to-normal (NN) beat in a raw ECG trace,
  2. converts those beats into the core HRV metrics (HR, SDNN, RMSSD, VLF, LF, HF, LF/HF) and
  3. plots every curve on an interactive dashboard so that trends can be inspected side-by-side.

Because the long-term goal is a live installation (eventually driving MIDI or other real-time mappings), the script is written from the start in a sliding-window style: at every step it re-computes each metric over a moving chunk of data.
Fast-changing variables such as heart-rate itself can use short windows and small hops; spectral indices need at least a five-minute span to remain physiologically trustworthy. Shortening that span may make the curves look “lively,” but it also distorts the underlying autonomic picture and breaks any attempt to compare one participant with another. The code therefore lets the user set an independent window length and step size for the time-domain group and for the frequency-domain group.
Let’s take a closer look at the code. If you want to see the full, visit: https://github.com/ninaeba/EmbodiedResonance

1. Imports and global parameters

import argparse
import sys
from pathlib import Path

import numpy as np
import pandas as pd
import plotly.graph_objs as go
import scipy.signal as sg
import neurokit2 as nk
  • argparse – give the script a tiny command-line interface so we can point it at any raw ECG CSV.
  • NumPy / pandas – basic numeric work and table handling.
  • scipy.signal – classic DSP tools (Butterworth filter, Lomb–Scargle).
  • neurokit2 – robust, well-tested R-peak detector.
  • plotly – interactive plotting inside a browser/Notebook; easy zooming for visual QA.

2. Tunable experiment-wide constants

FS_ECG   = 500              # ECG sample-rate (Hz)
BP_ECG   = (0.5, 40)        # band-pass corner frequencies

TIME_WIN, TIME_STEP = 60.0, 1.0   # sliding window for HR / SDNN / RMSSD
FREQ_WIN, FREQ_STEP = 300.0, 30.0   # sliding window for VLF / LF / HF
FGRID = np.arange(0.003, 0.401, 0.001)

BANDS = dict(VLF=(.003, .04), LF=(.04, .15), HF=(.15, .40))
  • Butterworth 0.5–40 Hz is a widely used cardiology band-pass that suppresses baseline wander and high-frequency EMG, yet leaves the QRS complex untouched.
  • 60s time-domain window strikes a balance: long enough to tame noise, short enough for semi-real-time trend tracking.
  • 300s spectral window is deliberately longer; the literature shows that the lower bands (especially VLF) are unreliable below ~5 min.
  • FGRID – dense frequency grid (1 mHz spacing) for a smoother Lomb curve.

3. ECG helper class – load, (optionally) filter, detect R-peaks

class ECG:
    def __init__(self, fs=FS_ECG, bp=BP_ECG, use_filter=True):
        ...
    def load(self, fname: Path) -> np.ndarray:
        ...
    def filt(self, sig):
        ...
    def r_peaks(self, sig_f):
        ...
  1. load – reads the CSV into a flat float vector and sanity-checks that we have >10 s of data.
  2. filt – if the --nofilt flag is absent, applies a 4-th-order zero-phase Butterworth band-pass (via filtfilt) so that the baseline drift of slow breathing (or cable motion) does not trick the peak detector.
  3. r_peaks – delegates the hard work to neurokit2.ecg_process, which combines Pan-Tompkins-style amplitude heuristics with adaptive thresholds; returns index positions and their timing in seconds.

4. HRV class – sliding-window metric engine

class HRV:
    ...
    def time_metrics(rr):
        ...
    def lomb_bandpowers(self, rr, t_rr):
        ...
    def time_series(self, r_t):
        ...
    def freq_series(self, r_t):
        ...
    def compute(self, r_t):
        ...
  • time_metrics converts every RR sub-series into three classic metrics
    – HR (beats/min), SDNN (overall beat-to-beat spread, ms), RMSSD (short-term jitter, ms).
  • Why Lomb–Scargle instead of Welch?
    The RR intervals are unevenly spaced by definition.
    • Welch needs evenly sampled tachograms or heavy interpolation → can distort the spectrum.
    • Lomb operates directly on irregular timestamps, preserving low-frequency content even if breathing or motion momentarily speeds up/slows down the heart.
  • lomb_bandpowers:
    1. Runs scipy.signal.lombscargle on de-trended RR values.
    2. Integrates power inside canonical VLF / LF / HF bands.
    3. Computes LF/HF ratio, but guards against division by tiny HF values.
  • time_series / freq_series slide a window (120 s or 300 s) across the experiment, jump every 30 s, calculate metrics, and store the mid-window timestamp for plotting.
  • compute finally stitches time-domain and frequency-domain rows onto a 1-second master grid so that all curves overlay cleanly.

5. Tiny colour dictionary

COLORS = dict(HR='#d62728', SDNN='#2ca02c', RMSSD='#ff7f0e',
              VLF='#1f77b4', LF='#17becf', HF='#bcbd22', LF_HF='#7f7f7f')

Just cosmetic – keeps HR red, SDNN green, etc., across all subjects so eyeballing becomes effortless.

6. plot() – interactive dashboard

def plot(ecg_f, hrv_df, fs=FS_ECG, title="HRV (Lomb)"):
    ...
  • Left y-axis = filtered ECG trace for QC (do peaks line up?).
  • Right y-axis = every HRV curve.
  • Built-in range-slider lets you scrub the 24-minute protocol quickly.
  • Hover shows exact numeric values (handy when you are screening anomalies).
  • different backgrounds for phases

7. CLI wrapper

if __name__ == '__main__':
    main()

Inside main() we parse the file name and the --nofilt flag, run the whole pipeline, save the HRV table as a CSV sibling (same stem, suffix .hrv_lomb.csv) and open the Plotly window.

The four summary plots included below are therefore not an end-point but a launch-pad: they give us a quick visual fingerprint of each participant’s autonomic response, and will serve as the reference material for deeper statistical comparison, pattern-searching, and—ultimately—the data-to-sound (or other real-time) mappings we plan to build next.

by Alina Volkova -

Experiment III: Embodied Resonance – HRV Metrics and meaning

Heart-rate variability, or HRV, is the tiny, natural wobble in the time gap from one heartbeat to the next. It exists because two automatic “pedals” are always tugging at the heart. One pedal is the sympathetic system, the same chemistry that makes your pulse race when you are startled. The other pedal is the vagus-driven parasympathetic system, the brake that slows the heart each time you breathe out or settle into a chair. The more freely these pedals can trade places, the more variable those beat-to-beat spacings become. HRV is therefore a quick, non-invasive way to listen to how relaxed, alert, or exhausted the body is.

When we measure HRV we usually pull out a few headline numbers.

SDNN is the overall statistical spread of beat intervals during a slice of time, for example one minute. A wide spread means the heart is flexible and ready to react. A very narrow spread means the system is locked in one gear, as happens in chronic stress or heart failure.

RMSSD zooms in on the jump from one beat to the very next, averages those jumps, and reflects how strongly the vagus brake is speaking. During slow, deep breathing RMSSD grows larger; during mental tension or sleep deprivation it falls.

Frequency-domain measures treat the heartbeat trace like a piece of music and ask how loud each note is. Very-low-frequency power, or VLF, comes from extremely slow body rhythms such as hormone cycles and temperature regulation. Low-frequency power, or LF, sits in the middle and rises when the sympathetic pedal is pressed, for example in the first minute of exercise or during mental arithmetic. High-frequency power, or HF, sits exactly at breathing speed and is almost pure vagus activity: it swells during calm, diaphragmatic breathing and shrinks when breathing is shallow or hurried. A simple way to summarise the tug-of-war is the LF-to-HF ratio. When the sympathetic pedal dominates the ratio climbs; when the vagus brake dominates the ratio slides downward.

In a healthy, rested adult who is quietly seated the heart rate is steady but not rigid. SDNN and RMSSD show a modest but clear jitter, HF power pulses in step with the breath, LF is similar in size to HF, and the LF/HF ratio hovers around one or two. If the same person begins brisk walking heart rate rises, HF power fades, LF power grows, and the LF/HF ratio can shoot above five. During slow breathing meditation RMSSD and HF surge while LF/HF drops below one. In someone with chronic anxiety or PTSD the resting pattern is different: SDNN and RMSSD are low, HF is thin, LF/HF is already high before any task, and it climbs even higher during mild stress. The pattern can be even flatter in advanced heart disease, where both pedals are weak and total HRV is minimal.

Put simply, HRV lets us watch the nervous system’s soundtrack: fast notes reflect breathing and relaxation, mid-notes reflect alertness, and the overall volume tells us how much capacity the system still has in reserve.

The raw material for every HRV metric is the NN-interval sequence:
NNi is the time in seconds between two consecutive normal (sinus) beats.

SDNN is the standard deviation of that sequence.

SDNN = √[ Σ (NNi – NN̄)² / (N – 1) ]
Units are milliseconds because the intervals are expressed in ms. A resting, healthy adult who sits quietly will usually show an SDNN between roughly 30 ms and 50 ms. Endurance athletes can sit in the 60–90 ms range, while chronically stressed or cardiac patients may drift below 20 ms.

RMSSD focuses on the beat-to-beat jump and is dominated by parasympathetic (vagal) tone.

RMSSD = √[ Σ ( NNi – NNi-1 )² / (N – 1) ]
Again the unit is milliseconds. Typical resting values in a calm, healthy adult are about 25–40 ms. Slow breathing, a nap, or meditation can push it up toward 60 ms, whereas sustained mental effort, anxiety, sleep deprivation, or PTSD often pull it down below 15 ms.

Frequency-domain indices start from the same NN series but first convert it into a power-spectrum, most accurately with a Lomb–Scargle periodogram when the points are unevenly spaced:

P(f) = (1/2σ²) { [ Σ NNi cos ωi ]² / Σ cos² ωi + [ Σ NNi sin ωi ]² / Σ sin² ωi }
where ω = 2πf and f is scanned from 0.003 Hz upward.

Power is then integrated over preset bands and reported in ms² because it represents variance of the interval series per hertz.

Very-low-frequency power VLF integrates P(f) from 0.003 Hz to 0.04 Hz. In a healthy resting adult VLF is often 500–1500 ms². Because the mechanisms behind VLF (thermoregulation, hormones, renin-angiotensin cycle) change only slowly, values can drift greatly between individuals and between days.

Low-frequency power LF integrates P(f) from 0.04 Hz to 0.15 Hz. A quiet, healthy adult usually sits near 300–1200 ms². LF rises when the sympathetic accelerator is pressed, for example during the first few minutes of exercise or a stressful mental task.

High-frequency power HF integrates P(f) from 0.15 Hz to 0.40 Hz, exactly the normal breathing range. Calm diaphragmatic breathing drives HF toward 400–1200 ms², whereas rapid or shallow breathing in anxiety or hard exercise cuts HF sharply, sometimes below 100 ms².

LF/HF is the simple ratio LF ÷ HF. At rest a ratio near 1–2 suggests a balanced tug-of-war. If the ratio soars above 5 the sympathetic branch is clearly on top; if it falls below 0.5 the vagus brake is dominating (seen in deep meditation or in some fainting-prone individuals).

All of these numbers rise and fall in real time as the two branches of the autonomic nervous system jostle for control, so plotting them across the exercise-rest protocol lets us see how quickly and how strongly each person’s physiology reacts and recovers.

MetricUnitsWhat It Reflects (plain-language)Typical Resting Range in Healthy Adults*When It Runs High (what that often means)When It Runs Low (what that can signal)
Mean Heart Rate (HR)beats per minute (bpm)How fast the heart is beating on average50 – 80 bpmPhysical effort, fever, anxiety, dehydrationExcellent cardiovascular fitness, medications that slow the heart
SDNNmilliseconds (ms)Overall “spread” of beat-to-beat intervals—long-term autonomic flexibility40 – 60 msGood recovery, calm alertness, athletic conditioningChronic stress, heart disease, PTSD, over-fatigue
RMSSDmsVery short-term vagal (rest-and-digest) shifts from one beat to the next25 – 45 msDeep relaxed breathing, meditation, lying downSympathetic overdrive, poor sleep, depression
VLF Powerms²Very-slow oscillations (< 0.04 Hz) tied to long hormonal / thermoregulatory rhythms600 – 2000 ms²Possible inflammation, overtraining, sustained stress loadOften low in severe autonomic dysregulation
LF Powerms²“Middle-speed” swings (0.04–0.15 Hz) from blood-pressure reflex & controlled breathing (~6 breaths/min)600 – 2000 ms²Active mental effort, controlled slow breathing, standing upBlunted baroreflex, autonomic failure
HF Powerms²Fast vagal modulation (0.15–0.40 Hz) synchronized with normal breathing (3–9 breaths/min)500 – 1500 ms² (age-dependent)Relaxation, slow diaphragmatic breathing, lying downAnxiety, rapid shallow breathing, fatigue
LF/HF Ratio— (dimension-less)Balance of sympathetic “drive” (LF) vs vagal “brake” (HF)0.5 – 2.0Acute psychological stress, caffeine, upright postureExcess vagal tone, certain medications, autonomic failure