PPG Sleep Staging: How Accurate Are Wearable Sleep Trackers?
How wearables use PPG to classify sleep stages, how accurate they are vs polysomnography, and which algorithms work best. Covers HRV, heart rate, and autonomic signals.
- Sleep & Sleep Apnea
Sleep staging from PPG signals leverages the autonomic nervous system's characteristic fingerprint on cardiac activity during different sleep stages, enabling wearable devices to approximate the information traditionally available only from polysomnography. While PPG-based sleep staging cannot match the precision of EEG-based polysomnography (PSG), modern algorithms achieve clinically useful accuracy for population-level sleep assessment, longitudinal tracking, and screening for sleep disorders.
Sleep architecture -- the cyclical progression through wake, light sleep (N1/N2), deep sleep (N3), and REM sleep -- profoundly affects cardiovascular health, metabolic function, cognitive performance, and mental health. The American Academy of Sleep Medicine (AASM) guidelines define sleep staging based on electroencephalography (EEG), electromyography (EMG), and electrooculography (EOG) recorded during polysomnography. PPG-based approaches bypass these expensive, laboratory-bound sensors by exploiting the systematic changes in autonomic cardiac control that accompany each sleep stage. For background on how PPG sensors capture cardiac signals, see our guide to PPG technology.
Autonomic Nervous System Changes During Sleep
The physiological basis for PPG-based sleep staging rests on the well-established coupling between sleep stages and autonomic nervous system (ANS) activity. Each sleep stage has a characteristic sympathovagal balance that modulates heart rate, heart rate variability, peripheral vascular tone, and respiratory patterns -- all of which are accessible from the PPG waveform.
NREM Sleep (N1, N2, N3)
During the transition from wakefulness to NREM sleep, parasympathetic (vagal) tone progressively increases while sympathetic activity decreases. This produces several measurable cardiac effects:
- Heart rate decreases as sleep progresses from wake through N1, N2, and N3. Boudreau et al. (2013) found that sleep onset and progression to deeper sleep stages were associated with a shift toward greater parasympathetic modulation, while REM sleep shifted toward greater sympathetic modulation (doi:10.5665/sleep.3230).
- HRV increases during NREM sleep, particularly high-frequency (HF) power (0.15-0.4 Hz), which reflects respiratory sinus arrhythmia and vagal tone.
- LF/HF ratio decreases during NREM sleep, reflecting the shift toward parasympathetic dominance.
- Blood pressure decreases by 10-20% during NREM sleep (nocturnal dipping), reducing PPG pulse amplitude variability.
Deep sleep (N3) shows the most pronounced autonomic signature: markedly slow, regular heart rate with high vagal tone and very low sympathetic activity. This makes N3 the most reliably detected NREM stage from PPG.
REM Sleep
REM sleep is characterized by autonomic instability with phasic surges of sympathetic activity superimposed on a background of parasympathetic tone. This produces:
- Increased heart rate variability with irregular, phasic increases in heart rate during phasic REM bursts.
- Increased LF power and LF/HF ratio compared to NREM sleep, though not as high as wakefulness.
- Respiratory irregularity, with variable breathing rate and amplitude that modulates the PPG signal.
- Peripheral vasoconstriction during phasic REM, which can reduce PPG signal amplitude.
The irregular, burst-like pattern of REM autonomic activity is distinct from both the stable vagal dominance of NREM and the sustained sympathetic activation of wakefulness, providing a basis for PPG-based REM detection.
Feature Extraction from PPG for Sleep Staging
Sleep staging algorithms extract features from PPG signals across multiple domains. The PPG signal is first processed to extract inter-beat intervals (IBIs), from which HRV features are computed, along with respiratory and morphological features.
Heart Rate Variability Features
HRV analysis of PPG-derived IBIs is the primary feature source. Features are typically computed over 30-second or 5-minute epochs to match PSG scoring convention:
Time-domain features:
- Mean heart rate (HR) and its epoch-to-epoch variability
- SDNN (standard deviation of NN intervals): overall HRV reflecting both sympathetic and parasympathetic activity
- RMSSD (root mean square of successive differences): primarily reflects parasympathetic activity
- pNN50 (percentage of successive intervals differing by >50 ms): another parasympathetic marker
Frequency-domain features:
- Very low frequency (VLF) power (0.003-0.04 Hz): reflects thermoregulatory and hormonal fluctuations
- Low frequency (LF) power (0.04-0.15 Hz): reflects both sympathetic and parasympathetic modulation
- High frequency (HF) power (0.15-0.4 Hz): primarily reflects respiratory sinus arrhythmia and vagal tone
- LF/HF ratio: traditionally interpreted as sympathovagal balance index
- Total spectral power: overall autonomic modulation magnitude
Non-linear features:
- Sample entropy (SampEn): measures signal complexity; lower during regular NREM rhythms, higher during variable REM and wake states
- Detrended fluctuation analysis (DFA) alpha1: short-term fractal scaling exponent; differs between sleep stages due to different correlation structures
- Poincare plot descriptors (SD1, SD2): SD1 reflects short-term (beat-to-beat) variability, SD2 reflects longer-term variability
PPG-Derived Respiratory Features
Respiration modulates the PPG signal through three mechanisms: respiratory-induced intensity variation (RIIV), respiratory-induced amplitude variation (RIAV), and respiratory-induced frequency variation (RIFV, equivalent to respiratory sinus arrhythmia). These can be extracted to estimate respiratory rate and regularity:
- Respiratory rate: Differs between stages; typically 12-20 breaths/min during wakefulness, 10-16 during NREM, and variable (8-24) during REM.
- Respiratory regularity: The coefficient of variation of breath-to-breath intervals; low during stable NREM, high during REM.
For detailed methods on extracting respiratory signals from PPG, see our article on PPG respiratory rate estimation.
Movement and Signal Quality Features
Wrist-worn devices often include accelerometers that provide movement data complementary to PPG. Even without a dedicated accelerometer, PPG signal quality metrics serve as movement proxies:
- Signal quality index (SQI): Low SQI indicates motion, which correlates with wake or restless sleep.
- Motion artifact density: The fraction of epochs with detected motion artifacts.
- PPG amplitude stability: Stable amplitude suggests the subject is still (sleep), while large amplitude variations suggest movement (wake or stage transitions).
Classical Machine Learning Approaches
Early PPG-based sleep staging algorithms used hand-crafted features with classical classifiers, building on the HRV-based approaches originally developed for ECG-based sleep staging.
Random Forest and Gradient Boosting
Beattie et al. (2017) developed one of the most widely cited wrist-worn sleep staging algorithms, using movement, breathing, and heart rate variability features from an accelerometer and PPG (doi:10.1088/1361-6579/aa9047). In 60 self-reported normal sleepers, compared against a home sleep test with EEG channels, the algorithm achieved:
- Four-class accuracy (wake/light/deep/REM): 69%
- Cohen's kappa: 0.52
- Most common errors: light/REM and light/wake confusions
Wake detection is a persistent weakness of cardiac-based sleep staging: quiet wakefulness (lying still with eyes closed) produces HRV patterns similar to light sleep.
Walch et al. (2019) collected raw acceleration and PPG heart rate from the Apple Watch during polysomnography and compared motion, heart rate variability, and a circadian "clock proxy" feature across several classifiers (doi:10.1093/sleep/zsz180). Sleep-wake classification scored 90% of epochs correctly (93% of sleep epochs but only 59.6% of wake epochs), and three-class (wake/NREM/REM) accuracy was approximately 72% with all features.
Sleep stages follow a structured temporal sequence: wake typically transitions to N1, then N2, with deep sleep (N3) occurring primarily in the first half of the night and REM episodes lengthening toward morning. Hidden Markov models (HMMs) and their variants exploit this temporal structure.
Fonseca et al. (2017) validated wrist PPG heart rate variability plus accelerometry against PSG in 80 recordings from 51 healthy middle-aged adults, reporting a three-class (wake/NREM/REM) accuracy of 72.9% (kappa 0.46) and a four-class accuracy of 59.3% (kappa 0.42) (doi:10.1093/sleep/zsx097). In an HMM, the transition probability matrix naturally encodes the physiological constraints on sleep stage sequences -- for example, direct transitions from deep sleep to wake are rare, and the model learns to penalize such transitions.
Deep Learning Approaches
Deep learning has substantially advanced PPG-based sleep staging by learning feature representations directly from raw or minimally processed signals and by capturing complex temporal dependencies.
Convolutional Neural Networks
Kotzen et al. (2023) developed SleepPPG-Net, which combines a residual convolutional network for automatic feature extraction with a temporal convolutional network for long-range context. It was trained end-to-end on raw PPG from two public sleep databases totalling 2,374 patients and 23,055 hours (doi:10.1109/JBHI.2022.3225363). On a held-out test set it reached a median Cohen's kappa of 0.75 for four-class staging, versus 0.69 for the best prior approach, and 0.74 on an external database after transfer learning.
Attia et al. (2024) extended this line with SleepPPG-Net2, adding multi-source domain training and a domain-shift uncertainty layer across six datasets (2,574 recordings). Versus SleepPPG-Net, external-set median per-patient kappa improved by up to about 19%, with MESA-test median κp 0.75 and external κp of 0.74 (CFS), 0.69 (ABC), 0.69 (HomePAP), 0.57 (CAP), and 0.66 (SLEEPAI) (arXiv:2404.06869). For the full staging-plus-AHI SOTA map, see PPG sleep tracking SOTA.
Recurrent and Temporal Models
Sleep staging benefits from temporal context -- knowing the sleep stage of adjacent epochs constrains the current epoch's classification. Several architectures exploit this:
Radha et al. (2021) first trained a deep recurrent neural network on a large ECG sleep dataset (292 participants, 584 recordings), then adapted part of its weights to a smaller wrist PPG dataset (60 healthy participants, 101 recordings) through transfer learning (doi:10.1038/s41746-021-00510-8). The best strategy achieved four-class accuracy of 76.36% and Cohen's kappa of 0.65, significantly outperforming both the PPG-only and ECG-only baselines.
Transformer-Based Models
Transformer architectures apply self-attention across sequences of consecutive 30-second epochs, which lets the model weight neighboring epochs differently depending on the transition patterns and capture longer-range context such as the ultradian sleep cycle.
Performance Comparison and Benchmarks
Performance comparison across studies is complicated by differences in datasets, population demographics, class definitions (three-class vs. four-class vs. five-class), and evaluation protocols. The following table summarizes representative results:
Scroll horizontally if needed
| Study | Method | N Subjects | Classes | Accuracy | Kappa |
|---|---|---|---|---|---|
| Beattie et al., 2017 | Accelerometer + PPG features | 60 | 4 | 69% | 0.52 |
| Walch et al., 2019 | Neural net (motion, HR, clock proxy) | not stated in abstract | 3 | ~72% | not reported |
| Fonseca et al., 2017 | PPG HRV + accelerometer | 51 (validation) | 3 | 72.9% | 0.46 |
| Fonseca et al., 2017 | PPG HRV + accelerometer | 51 (validation) | 4 | 59.3% | 0.42 |
| Radha et al., 2021 | RNN, ECG-to-PPG transfer learning | 60 (PPG) | 4 | 76.4% | 0.65 |
| Kotzen et al., 2023 | SleepPPG-Net (ResNet + TCN) | 2,374 (total) | 4 | not reported | 0.75 |
| Attia et al., 2024 | SleepPPG-Net2 (multi-source) | 2,574 (total) | 4 | see paper | 0.75 (MESA-test median κp) |
For context, in the AASM inter-scorer reliability program (more than 2,500 scorers), agreement with the majority score averaged 82.6%, falling to 67.4% for N3 and 63.0% for N1 (Rosenberg and Van Hout, 2013; doi:10.5664/jcsm.2350). PPG-based methods are approaching but have not reached this benchmark, particularly for the more challenging stage distinctions (N1 vs. N2, REM vs. wake).
Clinical Applications and Limitations
Population-Level Sleep Assessment
PPG-based sleep staging is most valuable when applied at scale over extended periods. While individual-night staging has meaningful uncertainty (a substantial fraction of epochs may be misclassified), trends averaged over weeks to months are more reliable. Large epidemiological studies using wearable PPG data have revealed population-level associations between sleep architecture and health outcomes that were previously inaccessible due to the cost and inconvenience of PSG.
Sleep Disorder Screening
PPG-based sleep analysis shows promise for screening several sleep disorders:
- Obstructive sleep apnea (OSA): PPG can detect apnea-related oxygen desaturations (via SpO2), cyclical heart rate patterns, and sleep fragmentation. Using 12,923 polysomnography recordings from six databases, Levy et al. (2023) developed OxiNet, a deep learning model that estimates the apnea-hypopnea index from the oximetry signal alone; it missed 0.2% of moderate-to-severe OSA patients in the test set, versus 21% for the best benchmark (doi:10.1038/s41467-023-40604-3).
- Insomnia: Total sleep time, sleep onset latency, and wake after sleep onset (WASO) can be estimated from PPG-based sleep/wake classification, though accuracy for detecting brief wake episodes remains limited.
- Circadian rhythm disorders: Long-term PPG-derived sleep timing data can identify delayed or advanced sleep phase patterns without requiring sleep diaries or actigraphy.
Known Limitations
Several limitations constrain PPG-based sleep staging accuracy:
- N1 detection: N1 is poorly detected by all PPG methods due to minimal autonomic differentiation from wakefulness and N2.
- Wake detection during the night: Brief awakenings (1-5 minutes) are frequently missed because HRV changes lag the EEG state transition.
- Sleep disorders: Patients with atrial fibrillation, heart failure, or autonomic neuropathy have altered HRV patterns that violate the assumptions of sleep staging algorithms trained on healthy populations.
- Medications: Beta-blockers, anticholinergics, and other medications affecting autonomic function alter the HRV-sleep stage relationship.
- Age effects: HRV decreases with age, reducing the dynamic range of features used for staging. Algorithms trained primarily on younger adults may perform poorly in elderly populations.
For researchers working on sleep staging algorithms, our algorithms reference provides foundational signal processing methods, and understanding the conditions affecting sleep and cardiovascular health is essential for developing robust classification systems.
References
- Boudreau, P., Yeh, W. H., Dumont, G. A., & Boivin, D. B. (2013). Circadian variation of heart rate variability across sleep stages. Sleep, 36(12), 1919-1928. https://doi.org/10.5665/sleep.3230
- Beattie, Z., et al. (2017). Estimation of sleep stages in a healthy adult population from optical plethysmography and accelerometer signals. Physiological Measurement, 38(11), 1968-1979. https://doi.org/10.1088/1361-6579/aa9047
- Walch, O., Huang, Y., Forger, D., & Goldstein, C. (2019). Sleep stage prediction with raw acceleration and photoplethysmography heart rate data derived from a consumer wearable device. Sleep, 42(12), zsz180. https://doi.org/10.1093/sleep/zsz180
- Fonseca, P., et al. (2017). Validation of photoplethysmography-based sleep staging compared with polysomnography in healthy middle-aged adults. Sleep, 40(7), zsx097. https://doi.org/10.1093/sleep/zsx097
- Kotzen, K., et al. (2023). SleepPPG-Net: A deep learning algorithm for robust sleep staging from continuous photoplethysmography. IEEE Journal of Biomedical and Health Informatics, 27(2), 924-932. https://doi.org/10.1109/JBHI.2022.3225363
- Attia, S., et al. (2024). SleepPPG-Net2: Deep learning generalization for sleep staging from photoplethysmography. arXiv:2404.06869. https://arxiv.org/abs/2404.06869
- Radha, M., et al. (2021). A deep transfer learning approach for wearable sleep stage classification with photoplethysmography. npj Digital Medicine, 4, 135. https://doi.org/10.1038/s41746-021-00510-8
- Rosenberg, R. S., & Van Hout, S. (2013). The American Academy of Sleep Medicine inter-scorer reliability program: sleep stage scoring. Journal of Clinical Sleep Medicine, 9(1), 81-87. https://doi.org/10.5664/jcsm.2350
- Levy, J., et al. (2023). Deep learning for obstructive sleep apnea diagnosis based on single channel oximetry. Nature Communications, 14, 4881. https://doi.org/10.1038/s41467-023-40604-3
Related articles
How Accurate Are Fitness Trackers for Heart Rate During Exercise?
Fitness tracker heart rate accuracy during exercise drops significantly. Learn what the research says and how to get reliable data during workouts.
PPG Sleep Tracking SOTA: Staging Kappa, AHI/REI Cutoffs, and What Wearables Actually Measure
State of the art for PPG and wearable sleep tracking in 2026: four-class staging Cohen's kappa (SleepPPG-Net2, wav2sleep), AHI/REI screening cutoffs (≥5 / ≥15 / ≥30), ODI and hypoxic burden, plus public datasets and leaderboards.
Home Sleep Apnea Test vs Polysomnography: Where Wearables Fit and What They Cannot Replace
HSAT and polysomnography answer different clinical questions. This guide explains when each test is appropriate, what wearables and PPG can support, and what they still cannot replace in sleep apnea care.
Which Wearable Has the Most Accurate Heart Rate Monitor in 2026?
Comparing heart rate accuracy across Apple Watch, Garmin, Fitbit, Samsung, Polar, and Oura Ring. Based on independent clinical validation studies.
Sleep Regularity Index: Formula, Range, and Wearable Use
The sleep regularity index scores whether the same minute is asleep or awake 24 hours later. Phillips et al. 2017 rescale that probability onto 0 to 100.
Devices in this article
Research records with the datasets, papers and published validation results for each device.
