Bachelor-Arbeit-Sophia-Haberland/README.md

Overview

• Data source: SingleRecordings/<id>/{hr.csv, rr.csv, timestamps.csv}
• Scripts:
  • plot_meditation_data.py: per-recording plots (raw + moving averages + segment boxplots)
  • aggregate_segments_analysis.py: aggregate metrics/plots across recordings with exactly 4 marks (conditions)

Results

Condition	HR (bpm)	RR (ms)	RMSSD (ms)
Breathing Scene 1	72.61	843.33	52.70
Spring Scene	72.33	860.78	56.95
Summer Scene	72.50	862.00	41.15
Autumn Scene	72.89	850.44	41.89
Breathing Scene 2	73.78	839.94	42.59

Timestamp Timing Analysis

This table presents the timing analysis for each recording, focusing on durations in seconds. It includes the time from the first heart rate (HR) recording to the first timestamp marker, the intervals between consecutive timestamp markers, and the time from the last timestamp marker to the end of the HR recording.

Recording ID	Breathing Scene 1 (s)	Spring Scene (s)	Summer Scne (s)	Autumn Scene (s)	Breathing Scene 2 (s)
01	214.037	160.618	168.298	156.118	79.05
02	189.707	196.964	162.457	145.152	101.075
03	177.139	184.446	174.979	160.041	104.475
04	253.685	167.817	169.903	150.954	103.016
05	220.387	157.653	184.168	149.795	138.08
06	253.251	167.564	177.214	153.651	96.502
07	218.86	163.38	177.967	154.913	122.008
08	184.833	160.343	180.283	154.76	152.834
10	182.163	164.495	184.733	144.337	113.889

Summary Statistics for Timestamp Timing Analysis

Metric	Average (s)	Min (s)	Max (s)
Time to first marker	210.451	177.139	253.685
Interval 1→2	169.253	157.653	196.964
Interval 2→3	175.556	162.457	184.733
Interval 3→4	152.191	144.337	160.041
Time from last marker to end	112.325	79.050	152.834

Plots

Aggregate Boxplots

Aligned Average Curves

Outlier Detection

Scripts filter any RR values outside of the range [300, 2000] ms. Values outside of this range are replaced by interpolated values. For current samples no outliers were detected.

How calculations are done

• HR and RR loading

  • Read timestamp (ms) → convert to datetime; set as index.
  • HR column: hr (beats per minute). RR column: rr_ms (milliseconds).

• Moving averages (per recording)

  • Time-based rolling means over irregular timestamps using windows: 5 s, 10 s, 30 s, 60 s.
  • X-axis is seconds from each recording’s start.

• Mark handling (per recording)

  • Vertical lines at marks from timestamps.csv.
  • Segment boxplots use values in: each interval between consecutive marks and the final interval (last mark → end).

• RMSSD (HRV)

  • Definition: RMSSD = sqrt(mean(diff(RR_ms)^2)) using successive RR intervals (ms).
  • Aggregate per-condition RMSSD: computed over all RR samples within each condition’s time window (recording-level); requires ≥2 RR samples.
  • Aligned time-series RMSSD: first compute a 30 s time-based rolling RMSSD per recording, then align and average across recordings (see below).

• Aggregate metrics across recordings (exactly 4 marks)

  • Conditions (English labels):
    • Breathing Scene 1 (pre-first mark)
    • Spring Scene (1st–2nd mark)
    • Summer Scene (2nd–3rd mark)
    • Autumn Scene (3rd–4th mark)
    • Breathing Scene 2 (post-last mark)
  • Per recording: compute medians for HR and RR within each condition; compute RMSSD within each condition.
  • Summary CSV aggregates these per-recording values; boxplots show distributions. The mean of each condition is marked by a black dot on the boxplots.

• Aligned average curves (HR, RR, RMSSD)

  • Only recordings with exactly 4 marks are used.
  • For each recording, durations of the five segments are measured; median segment proportions across recordings define a normalized 0–1 time axis with aligned boundaries.
  • Each recording’s series is piecewise-linearly time-normalized to this axis, interpolated to a common grid, and then averaged (mean ± 1 SD). For RMSSD, a 30 s rolling RMSSD is used before alignment.