The convergence of wearable technology and sophisticated machine learning offers unprecedented opportunities for real-time health and performance optimization. The Polar H10 chest strap, traditionally recognized as the gold standard for accurate heart rate measurement, now serves as a powerful platform for holistic physiological analysis by streaming both high-fidelity Electrocardiogram (ECG) data and accelerometry data.
The application "Polar Emotions" leverages this dual capability, enabling the simultaneous and non-invasive monitoring of two critical indicators of the Autonomic Nervous System (ANS) state: Emotional State (derived from cardiac activity) and Respiratory Rate (RR) (derived from mechanical movement).
This integrated approach addresses a crucial need for long-session monitoring in fields ranging from sports science to clinical well-being. By utilizing the continuous data streams from the Polar H10, the application can perform:
Emotional State Classification (via Streaming ECG): A Convolutional Neural Network (CNN) processes the raw ECG signal to calculate Heart Rate Variability (HRV) metrics and complex morphological features. This allows for the real-time classification of an individual’s affective state (e.g., Stressed, Zen, or Normal).
Respiratory Rate Tracking (via Streaming Accelerometer): The 25 Hz accelerometry data, specifically the Z-component, is filtered and analyzed (via FFT or peak detection) to precisely determine the low-frequency Respiratory Rate (0.1 Hz to 0.75 Hz).
The capacity of "Polar Emotions" to monitor these signals simultaneously or separately over extended periods ensures the acquisition of rich, ecologically valid data, providing users and researchers with actionable insights into the dynamic interplay between the heart, breath, and emotional resilience.
Heart Rate Variability (HRV) is the physiological phenomenon of the variation in the time interval between consecutive heartbeats. It's not the same as heart rate, but rather a measure of the minute fluctuations in the rhythm. This variation is regulated by the Autonomic Nervous System (ANS), which has two main branches: the Sympathetic ("fight or flight") system, and the Parasympathetic ("rest and digest") system.
The balance between these two systems is a strong indicator of a person's emotional and physiological state.
🧘 Zen (Relaxed/Resilient): A high, regular HRV pattern is typically associated with a dominant parasympathetic system. This indicates a state of calmness, good recovery, and emotional resilience. The heart can respond quickly and flexibly to internal and external demands.
😠 Stressed (Fight or Flight): A low HRV, where the heartbeats are more rigid and less variable, suggests the sympathetic system is dominant. This state is characteristic of high stress, anxiety, fatigue, or being in a continuous "fight or flight" mode. The body is in a state of high alert, limiting its capacity for recovery.
👤 Normal (Balanced): An average or moderate HRV, showing a healthy mix of sympathetic and parasympathetic activity, reflects a general well-being and the body's ability to adapt effectively to daily life's stressors and relaxation periods.
In essence, a higher HRV signifies a healthier and more flexible heart-brain connection, often correlating with a more Zen and well-regulated emotional state, while a suppressed HRV is a common marker for stress.
The detection of human affective states, such as stress, relaxation, and baseline, is increasingly being achieved through the application of deep learning techniques to physiological signals. This approach fundamentally relies on exploiting the subtle modulations of the cardiovascular system, which is under the direct control of the Autonomic Nervous System (ANS). The use of a Convolutional Neural Network (CNN), trained on a comprehensive multimodal dataset like WESAD (WEarable Stress and Affect Detection), represents a state-of-the-art methodology for this purpose.
Emotional states are not merely cognitive; they are embodied physiological responses translated by the ANS, which comprises the Sympathetic ("fight or flight") and Parasympathetic ("rest and digest") branches.
A. Heart Rate Variability (HRV) as an ANS Proxy
Heart Rate Variability (HRV) is a key metric derived from the Electrocardiogram (ECG). It quantifies the beat-to-beat variations in the time interval between consecutive heartbeats (R-R or Inter-Beat Intervals, IBI). These variations are a non-invasive index of ANS balance.
Parasympathetic Dominance (Relaxation/Zen): This state is characterized by high HRV, reflecting strong vagal tone (via the vagus nerve). High variability signifies cardiovascular flexibility and resilience. Quantitative metrics, such as the Root Mean Square of Successive Differences (RMSSD) in the time domain or High Frequency (HF) power in the frequency domain, are typically elevated.
WESAD Protocol Examples: Amusement and Meditation conditions.
Sympathetic Dominance (Stress): This state is associated with low HRV. Sympathetic activation reduces the temporal variation in heartbeats, leading to a more rigid rhythm. An increased Low Frequency/High Frequency (LF/HF) ratio is often observed, indicating enhanced sympathetic activity and/or withdrawal of parasympathetic influence.
WESAD Protocol Example: Stress condition.
B. ECG Signal as a High-Resolution Input
While HRV is a derived metric, the Electrocardiogram (ECG) signal itself, often sampled at high frequencies (e.g., 700 Hz as in the RespiBAN device), contains morphological and temporal characteristics (P wave, QRS complex, T wave) that are subtly affected by autonomic discharge. Providing the raw ECG signal to the model allows the CNN to learn features beyond standard HRV metrics, potentially capturing finer nuances in cardiac activity linked to affect. Crucially, while the WESAD dataset primarily uses the RespiBAN, many real-world applications, including continuous monitoring, rely on devices like the Polar H10 chest strap, which typically provides the raw ECG signal at an acquisition rate of 130 Hz for subsequent analysis. This signal quality is essential as the CNN processes these raw streams.
The adoption of a CNN is methodologically sound for physiological data classification. CNNs are highly effective at learning hierarchical features directly from raw data, bypassing the need for extensive, hand-crafted feature extraction.
Multimodal Time-Series Input: The input to the CNN is a temporal sequence, combining the raw ECG stream with other physiological modalities provided by the dataset (e.g., EDA, ACC, TEMP, BVP from both RespiBAN and Empatica E4).
Feature Learning: Convolutional filters (kernels) operate on the time series, learning to identify specific temporal patterns. Early layers may detect fundamental waveform features (peaks, slopes, specific beat intervals), while deeper layers learn complex, long-range rhythmic oscillations that are characteristic of the different affective states (e.g., a specific rhythm pattern indicative of stress).
Advantage over Traditional Methods: By processing the high-frequency raw data streams, the CNN is optimized to capture the subtle, non-linear transitions and interactions between physiological signals that define the shift between states like Zen, Stressed, and Normal (Baseline).
The WESAD dataset offers a well-defined framework for training and validating affective detection models.
Multimodality and Synchronization: The dataset's strength lies in its diverse, synchronized sensor data (RespiBAN chest-worn and Empatica E4 wrist-worn devices), allowing for a comprehensive view of the body's response.
Defined Ground Truth: WESAD provides two forms of ground truth for classification:
Protocol Labels (Objective): Labels derived from the controlled experimental conditions (Baseline, Stress, Amusement, Meditation).
Self-Report Questionnaires (Subjective): Data from validated scales like PANAS, STAI, and SAM, which capture the participant's perceived affective state.
A successful CNN model leverages the objective physiological changes in the ECG/HRV signals to accurately predict the discrete affective states defined by the WESAD labels, ultimately serving as a reliable tool for continuous, passive wearable stress and affect detection.
The ability to continuously and non-invasively monitor Respiratory Rate (RR), also known as breathing frequency, offers significant utility in both clinical and high-performance settings. Respiratory Rate Variability (RRV) and its absolute value are powerful indicators of overall physiological balance and metabolic status.
🩺 Medical Context (Clinical Utility): RR is a crucial component of vital signs. Subtle changes in baseline RR can be an early indicator of physiological deterioration (e.g., in cases of infection, sepsis, or respiratory distress). Continuous monitoring allows for the early detection of abnormal breathing patterns (e.g., Cheyne-Stokes respiration) and is vital in managing conditions like sleep apnea or monitoring recovery post-surgery.
🏃 Sport Context (Performance Utility): In sports science, the efficiency of the cardiorespiratory system is paramount. RR tracking helps assess the aerobic threshold and ventilatory efficiency. During high-intensity training, monitoring the ratio of oxygen consumed to CO2 produced (Respiratory Exchange Ratio) can be approximated, aiding in the determination of fatigue onset and optimizing training load. Furthermore, controlled breathing (low RR) is a core component of recovery and mental state training (e.g., mindfulness and meditation), which is linked to high VFC.
The Polar H10 chest strap, primarily known for its ECG acquisition, includes a built-in accelerometer which enables the non-invasive monitoring of respiration.
A. Physiological Basis of Signal Acquisition
Respiration involves the mechanical expansion and contraction of the chest cavity. This movement causes subtle shifts in the orientation and position of the device worn on the chest. The onboard 3-axis accelerometer captures these small displacement changes.
Target Component: The relevant motion is primarily captured by the accelerometer's Z-component—the axis perpendicular to the chest surface. This component is most sensitive to the forward/backward movement of the sternum and rib cage during inhalation and exhalation.
Sampling Rate: The accelerometric data is typically streamed at a moderate sampling rate, such as 25 Hz in the Polar H10, which is more than sufficient to capture the low-frequency respiratory signal.
B. The Respiratory Frequency Range
Normal human respiratory frequencies fall within a very specific and low-frequency band.
Target Bandwidth: The breathing signal is generally contained within the range of 0.1 Hz to 0.75 Hz.
0.1 Hz corresponds to 6 breaths per minute (slow/controlled breathing).
0.75 Hz corresponds to 45 breaths per minute (strenuous exercise/heavy breathing).
To isolate the RR signal from movement artifacts and the heart rate signal, specific digital signal processing techniques are applied in a live streaming context.
A. Initial Filtering and Isolation
The raw accelerometric data is subjected to a band-pass filter designed to retain only the energy within the 0.1 Hz to 0.75 Hz range.
Band-Pass Filtering: This is the first critical step to remove high-frequency noise (e.g., sudden body movements) and very low-frequency drift (e.g., posture changes).
Fourier Transform (FFT): The Fast Fourier Transform (FFT) is a highly efficient method used to transform the time-domain signal into the frequency domain. By analyzing the frequency spectrum of the filtered signal, the dominant frequency peak within the target 0.1 Hz to 0.75 Hz band corresponds directly to the respiratory frequency.
B. Peak Detection Method
Alternatively or in parallel, peak detection algorithms can be applied directly to the filtered time-domain signal.
Time-Domain Analysis: The filtered accelerometry signal will show clear peaks and troughs corresponding to the maximum expansion and maximum contraction of the chest.
RR Calculation: The time interval between successive peaks represents the period of one full breath cycle. The inverse of this period yields the respiratory frequency in Hz, which is then typically converted to breaths per minute (BPM).
This real-time, continuous calculation of RR from the chest-worn accelerometer provides a valuable, simultaneous physiological context to cardiac metrics like ECG and HRV, enabling a more holistic and accurate assessment of an individual's stress, recovery, and fitness status.
This application allows you to monitor and analyze your emotional states through ECG (electrocardiogram) and accelerometer data from your Polar device. Polar Emotions is an advanced biometric monitoring application designed to work with Polar H10 devices. It provides real-time analysis of your electrocardiogram (ECG) and respiratory rate to help you understand your stress levels and physiological state.
Connect to Device
The application should connect to the Polar H10 automatically.
Wait for the connection confirmation
When it is ready Tap the button (with Bluetooth icon) to start ECG streaming
Start Recording
If you want to stream both ECG and ACC (acceleartion), tap "Start ECG & ACC" to begin recording
The app will simultaneously capture ECG and accelerometer data
Recording duration and heart rate will be displayed in real-time
View Settings
Tap the ⚙️ button to access application settings
Configure your preferences for analysis and display
ECG Display Controls:
Zoom In/Out: Use the magnifying glass buttons to adjust the ECG signal zoom level
Pan Left/Right: Use the arrow buttons to navigate through the recorded signal
Time Display: Current recording time is shown at the top of the ECG view
Emotion Detection:
Tap "Search for emotions" to analyze your ECG data
The app identifies three emotional states:
Stress (red): High arousal, tension
Normal (blue): Neutral baseline state
Zen (green): Relaxed, calm state
Respiratory Analysis:
Tap "Respiratory analysis" to extract breathing patterns from your ECG
Results include breathing rate and variability metrics
Pie Chart:
Visual representation of time spent in each emotional state
Color-coded segments: Red (Stress), Blue (Normal), Green (Zen)
Statistics Panel:
Shows percentage breakdown of emotional states
Duration spent in each state
Overall recording duration
View Saved ECGs: Access previously recorded sessions
Choose ECG File
Browse and select previously saved ECG recordings
Files are automatically saved with timestamp
Select ACC File
Load corresponding accelerometer data
Required for respiratory analysis
Calendar View
View recordings organized by date
Quick access to historical data
Back: Return to current recording
New Detection: Start a fresh recording session
History: View all saved ECG recordings in table format
Toggle Statistics:
Tap the ▼ button to expand/collapse detailed statistics
View comprehensive respiratory metrics:
Breathing rate (breaths per minute)
Heart rate variability (HRV) parameters
Coherence scores
Statistical analysis results
Signal Visualization:
Same zoom and pan controls as main screen
Switch between ECG and processed signal views using the return button
View both ECG and accelerometer traces
Information Display:
Date and time of recording
Total duration
Real-time signal quality indicators
Start/Stop Recording:
Begin continuous respiratory rate monitoring
Stop the session at any time
Statistics Panel (Collapsible):
Current: Instantaneous breathing rate
Average: Mean breathing rate over session
Min/Max: Range of detected breathing rates
Confidence: Measurement reliability indicator
Method: Detection algorithm used
Valid Measurements: Count of successful readings
Line Chart:
Real-time graph of breathing rate over time
X-axis: Time elapsed
Y-axis: Respiratory rate (breaths/min)
Auto-scrolling display for continuous monitoring
Controls:
Start/Stop: Toggle recording
Back: Return to previous screen
Settings: Adjust monitoring parameters
Ensure proper sensor contact with skin
Minimize movement during recording (if not exercizing)
Record for at least 5-10 minutes for accurate emotional analysis
Stay still during respiratory analysis for best results
Stress Detection: Based on heart rate variability patterns
Zen State: Indicates high coherence and relaxation
Normal State: Baseline physiological state
Longer recordings provide more reliable statistics
Recordings are automatically saved with timestamps
Access history to track emotional patterns over time
Use calendar view for long-term trend analysis
Connection Issues:
Ensure Bluetooth is enabled
Check that Polar device is charged
Try restarting both app and device
Analysis Problems:
Verify recording duration is sufficient (minimum 5 minutes)
Check signal quality indicators
Re-record if signal contains too many artifacts
Display Issues:
Use zoom controls to adjust view
Toggle statistics panel if screen feels cluttered
Check that device screen rotation is enabled
For additional help or to report issues:
Access app settings for more information
Check for app updates regularly
Consult Polar device documentation for sensor-specific guidance
Version 18.0 - User Guide