Mi Band 10 - Scientific Review (AMAZING!)

Scientific concepts, discoveries, and phenomena

Study aim & experimental approach

• The creator performs a systematic, scientific comparison of two Xiaomi Mi Band 10 devices to evaluate:
  • Heart-rate tracking accuracy
  • Sleep-stage tracking consistency/reliability
• Devices are worn on both the left and right wrists (two Mi Band 10 units), creating a replicate/within-subject comparison.
• A key practical research issue is highlighted:
  • It can be difficult to access raw data from Xiaomi ecosystems for independent analysis.

Heart-rate tracking: reference measurements & quantitative validation

Reference instruments used

• Polar H10 chest strap: primary reference for heart rate
• Garmin chest strap: mentioned, but the analysis uses Polar H10 as the reference

Metrics / analysis concept

• Agreement is quantified using correlation (R value) between Mi Band readings and the reference heart rate.
• Ideal agreement is described as Mi Band points lying along a blue “perfect match” line in plots.

Exercises tested

• Indoor cycling (easier tracking scenario)
• Weightlifting (hard scenario due to wrist motion/tension affecting sensor signal)
• Outdoor cycling (more realistic, harder-than-indoor scenario)
• Running (tested the next day)

Key findings by activity

• Indoor cycling

  • Very strong agreement: ~R = 0.99 on both wrists
  • Most measurements align closely with the reference

• Weightlifting

  • Reduced accuracy at higher heart-rate peaks
  • Mi Band often detects heart rate that is too low when the user’s arm tension is high
  • Reported correlations:
    • Left wrist: ~R = 0.89
    • Right wrist: ~R = 0.92
  • Conclusion: wrist-worn tracking is less reliable for peak HR under heavy arm tension; a chest strap is recommended

• Outdoor cycling

  • Strong performance overall
  • Reported correlations:
    • Right wrist: ~R = 0.91
    • Left wrist: ~R = 0.96
  • Hypothesized influence of ambient temperature:
    • Hot indoor conditions may improve peripheral blood flow / sensor signal quality (speculative; flagged for further investigation)

• Running

  • Extremely strong agreement
  • Both wrists described as near-perfect overlap with reference curves
  • Reported:
    • Left wrist correlation ~1.00
  • Conclusion: among the best-performing devices in the creator’s prior testing set (based on the initial run under sunny conditions)

---

Sleep-stage tracking: comparison to EEG ground truth concept

Reference for sleep staging

• Zmax EEG headband
  • Measures brain waves and serves as the sleep-stage reference

Conceptual validation method

• Mi Band sleep stages are compared to EEG-derived stages.
• Stage evaluation uses a confusion-matrix-like approach:
  • Each column sums to 100% (how Mi Band classifies what EEG defines as each stage)
  • Perfect agreement would place values along the diagonal (~100% diagonal)
• Additional validation uses time-series plots:
  • Sleep stage vs. clock time to check for expected sleep cycle structure

Key sleep findings

• Low agreement overall for sleep stages
  • Deep sleep: relatively better (~~90% agreement mentioned for one unit)
  • Light and REM: around ~50% agreement or worse
• Between-device inconsistency
  • The two Mi Band 10 units produce substantially different sleep-stage patterns, suggesting poor reliability
• Physiology/phenomenon inconsistency
  • EEG shows the typical progression through sleep cycles ending with REM
  • Mi Band output shows:
    • extra/unrealistic deep sleep spread across the night
    • REM distributed “randomly” rather than cycling
• Poorer unit noted:
  • Deep sleep often classified as light sleep
  • REM agreement also reported as poor

Quantitative summary and placement

• The creator estimates the Mi Band 10’s average agreement and worst-stage agreement versus previously tested devices.
• Overall conclusion: Mi Band 10 ranks among the lower-performing sleep-stage trackers, not in the “best” tier.

---

Broader scientific framing mentioned

• Sleep-stage tracking is treated as an AI/machine learning task:
  • It requires large amounts of training data linking sensor readings to EEG-labeled sleep stages.
• Heart-rate tracking is framed as a sensor-signal problem:
  • Signal quality may vary with movement, wrist tension, and possibly environmental temperature (hypothesis).

---

Methodology / test outline (as described)

• Device setup

  • Unbox two Mi Band 10 units (different colors)
  • Connect each to separate phones and separate Mi Fitness / Xiaomi accounts
  • Ensure syncing to Strava (sync issues investigated via reconnecting)

• Heart-rate tests

  • For each wrist (left and right), run the same exercise categories:
    • Indoor cycling (short session)
    • Weightlifting (short sets)
    • Outdoor cycling (two ~30–35 min rides)
    • Running (one run)
  • During exercise:
    • record heart rate on Mi Band 10
    • record reference heart rate using Polar H10 chest strap
  • After exercise:
    • compute correlation (R value) and compare time-series plots

• Sleep tests

  • Sleep one night with both wrist-worn Mi Band 10s
  • Compare Mi Band stage timeline to Zmax EEG headband staging
  • Analyze using:
    • stage agreement matrix (ignoring “awake” column due to reference issues)
    • time-series cycle alignment (allowing ±15 min offset for best fit)

---

Researchers / sources featured

• Rob: the video creator (described as a “post-doal scientist” specializing in biological data analysis)
• Rafael: colleague mentioned as someone who could replicate/test independently
• Zmax EEG headband: EEG reference device (not an individual researcher)
• Polar H10: reference heart-rate sensor
• Garmin: mentioned in context of reference/benchmark measurements (including V̇O₂max discussion)
• Runna: running coach app
• Levels: glucose tracking app
• Strava: sync destination/service
• Whoop: Whoop Strap mentioned
• Apple Watch / Apple: benchmark devices mentioned
• Aura Ring: benchmark device mentioned
• Oura Ring / Sleep 2 / NUA app: sleep tracking apps mentioned
• Fitbit / Google Fitbit devices / Pixel Watch: sleep tracking comparisons mentioned
• Huawei: comparison for heart-rate tracking devices mentioned