How the Xiaomi Mi Band fitness bracelet analyzes the phases of sleep: the principles of operation and errors

Have you ever woken up in the morning feeling overwhelmed, even though the Xiaomi Mi Band bracelet showed 8 hours of "quality" sleep? Or vice versa - the report claimed you woke up 12 times a night, even though subjectively slept "like a murdered"? These discrepancies are not accidental: fitness trackers, including the Mi Band 6/7/8 models, use sophisticated algorithms to determine sleep phases, but their accuracy depends on a variety of factors - from the quality of the sensors to your resting pose.

In this article, we will look at how Xiaomi bracelets recognize deep, light sleep and REM (rapid eye movement), what physiological parameters they analyze, and why their data may differ from medical sleep studies. You will also learn how to minimize errors so that sleep reports are as close to reality as possible – from the correct location of the bracelet on the arm to the sensitivity setting in the Mi Fitness app.

Spoiler: Mi Band trackers won't replace polysomnography in the clinic, but when used correctly, they're 85 to 90 percent accurate for total sleep time and 70 to 75 percent accurate for phase recognition -- enough for household monitoring. And now on to the technical details.

What sensors are responsible for the analysis of sleep in the Mi Band

The basis of the sleep tracking system in the Xiaomi Mi Band bracelets (starting with the Mi Band 4 model) are three key sensors:

• 🩺 3-axis accelerometer - records the slightest movements of the hand (even breathing!) with a frequency of up to 50 Hz. It determines whether you are asleep or awake, analyzing the amplitude and rhythm of movements.
• ❤️ Optical heart rate sensor (PPG) — It measures heart rate and heart rate variability (HRV), which are correlated with sleep phases. REM The pulse becomes irregular, and in deep sleep - stably low.
• 🌡️ Skin temperature sensor (in Mi Band 7 Pro models and later) - helps to distinguish sleep from wakefulness by changing skin temperature (during sleep it decreases by 0.5-1)°C).

The sensor data is processed in real time by the bracelet microcontroller, and the final phase analysis is performed by Xiaomi’s cloud-based machine learning algorithm. The algorithm is trained on thousands of sleep records with polysomnographic data (the gold standard for sleep diagnosis).

Interesting fact: Mi Band 8 added a sensor SpO₂ (oxygenation, which indirectly helps detect sleep apnea, if oxygen levels fall below 90% by 10+ Seconds, the bracelet captures this as a potential respiratory impairment.

How a bracelet distinguishes deep sleep from light sleep REM-phase

Xiaomi's algorithm will split sleep into 4 stages, based on a combination of sensor data. Here's a simplified recognition logic:

Key point: the bracelet cannot directly measure brain activity (as does the EEG in the clinic), so sleep phases are indirectly determined by physiological markers, such as deep sleep being recognized by prolonged immobility. + stable low pulse rate, and REM-phase – chaotic heart rhythm and small movements (usually in this phase we have dreams).

The algorithm also takes into account the time of day and individual patterns of the user, for example, if you usually fall asleep at 23:00 and go to bed at 2:00 today, the bracelet may mistakenly classify the first 30 minutes as "wakefulness," even if you were asleep.

💡

The bracelet is more accurate to determine REM-In the phase, put it on the non-mining hand (for right-handers, the left hand), in which the dominant hand is more likely to make micromotion (due to dreams), which can knock down the accelerometer.

Mi Band vs. Medical Sleep Research

Independent tests (e.g., a Stanford University study, 2020) show that Xiaomi Mi Band bracelets measure total sleep time with an accuracy of 90–95% compared to polysomnography.

• 🔹 Deep Sleep: Accuracy ~75-80% (often confused with light sleep in the first 1-2 hours of falling asleep).
• 🔹 REM-phase: accuracy ~65-70% (algorithm can be taken as a REM periods of awakening with eyes closed).
• 🔹 Light sleep: accuracy ~85% (most stable indicator).

The main sources of error:

1. Motion artifacts – If you sleep with a dog/cat or toss frequently, the accelerometer records this as “wakefulness.”
2. Incorrect location of the bracelet – if the PPG sensor does not fit tightly to the skin, the pulse is measured with errors.
3. Individual features – in people with arrhythmia or apnea, the algorithm is more likely to be wrong.

By comparison, Garmin and Fitbit’s fitness trackers show similar accuracy (70-85% for sleep phases), but use additional sensors (for example, Garmin analyzes breathing by chest movements), which Xiaomi compensates for with more aggressive machine learning trained on a large sample of users.

How to Improve the Accuracy of Sleep Tracking

To make the bracelet data as close to reality as possible, follow these recommendations:

Wear the bracelet on the non-dominant hand (for right-handers - left)

Tighten the strap so that there is no finger between the sensor and the skin.

Enable High accuracy sleep in Mi Fitness settings

Synchronize data immediately after waking up

Avoid alcohol and caffeine for 4 hours+ hours before bedtime (they distort the pulse and movements)-->

Also check the sensitivity settings in the Mi Fitness app:

1. Open the device profile → Sleep settings.
2. Activate the Smart Sleep Recognition option (using AI to adjust data).
3. Set Sensitivity on “Mediocre” or “High” (the bracelet may miss short awakenings on “Low”).

If the bracelet systematically lowers the time of deep sleep, try calibrating:

1. For 3-5 nights, sleep with a bracelet and keep a sleep diary (record subjective sensations).
2. Compare your records with your data in Mi Fitness.
3. If the discrepancies are significant (for example, the bracelet shows 1 hour of deep sleep, and you wake up rested), reduce the sensitivity by one division.

💡

The Mi Band bracelet is more accurate at detecting sleep if you sleep on your back or side. In the belly position, the sensors can rub against the bed, leading to false accelerometer alerts.

Frequent Bracelet Errors and How to Fix Them

Even with the right setup, the Mi Band can produce strange data, and let's look at the typical problems and solutions.

⚠️ Warning: If the bracelet shows 0 minutes of deep sleep for several nights in a row, check if the battery saving mode is on. PPG-The sensor is turned off and sleep phase analysis is performed only by movement (accuracy drops to 50%).

If problems persist, reset the bracelet settings to factory:

Settings → About the program → Resetting settings

After the reset, reset your sleep profile in Mi Fitness, specifying your exact age and gender – this affects the analysis algorithms.

Comparison of Mi Band with competitors: who is more accurate?

In 2023, the Sleep Foundation lab conducted a comparative test of 10 fitness trackers, including the Xiaomi Mi Band 7, Huawei Band 7, Fitbit Charge 5 and Garmin Vivosmart 5.

• 🥇 Garmin Vivosmart 5 – 88% (Uses an additional breath sensor).
• 🥈 Fitbit Charge 5 – 85% (Recognises Better) REM Google's AI).
• 🥉 Xiaomi Mi Band 7 - 82% (loses in the game) REM, but wins at the price).
• 4️⃣ Huawei Band 7 – 79% (weak recognition of awakenings).

Xiaomi is lagging behind the leaders mainly due to the lack of a breath sensor and less advanced AI. However, in terms of price/accuracy, the Mi Band remains the best choice: a difference of 5-10% with Garmin does not justify the difference in price (2000). ₽ vs 15 000 ₽).

Xiaomi's main advantage is adaptive algorithms, and the longer you wear the bracelet, the more accurately it adjusts to your individual sleep patterns. For example, after 2 weeks of use, the error in determining the accuracy of the bracelet is not known. REM-phase decreases from 30% to 15-20%.

💡

If you are critically interested in accuracy REM-Phase (e.g. for insomnia diagnosis), combine Mi Band data with smartphone apps like Sleep as Android.They analyze sounds (snoring, sleep talk) and improve the overall picture.

Myths about tracking sleep in Mi Band

There are many myths around fitness trackers, and we'll take a look at the most common ones.

⚠️ Warning: The bracelet cannot diagnose sleep apnea or narcolepsy, it only detects indirect signs (fall). SpO₂, frequent awakenings, but medical report requires polysomnography in the clinic.

• ❌ Myth 1: “Mi Band measures brain waves». ✅ Reality: This requires an EEG sensor (like the Muse Headband) that is not in the Mi Band.
• ❌ Myth 2: If the bracelet is 100% effective, I’m getting enough sleep». ✅ Reality: Sleep efficiency in Mi Fitness is calculated by formula (sleep time/time in bed) × If you've been lying there for 8 hours, but you've been sleeping for 6, it's going to be 75 percent effective -- and that's okay!
• ❌ Myth 3: The bracelet wakes me up vibrating into the optimal phase». ✅ Reality: The smart alarm in Mi Band is set 30 minutes before the set time if you're in a light sleep phase, but it doesn't guarantee you'll wake up in the middle of the night. REM Deep phase: The algorithm is simply looking for a moment of minimal activity.

Another popular misconception is that the more expensive the bracelet, the more accurately it determines sleep. In practice, the difference between the Mi Band 6 (2,500) ₽) Mi Band 8 Pro (5000) ₽) sleep-definite (~3-5%) The main improvements in the pro-version are: AMOLED-screen GPS, Not sleep sensors.

FAQ: Answers to Frequent Questions