Synthetic PPG Test Videos

• Synthetic PPG test videos are artificially generated visual simulations of PPG signals that standardize benchmarking for smartphone-based heart rate measurement.
• They are created by mapping synthesized PPG waveforms to RGB video frames using statistical models, ensuring controlled and repeatable physiological signal patterns.
• This approach enables high-throughput, multi-device testing and objective performance comparisons, supporting regulatory compliance and accelerated mobile health innovation.

Synthetic photoplethysmographic (PPG) test videos are artificially generated visual stimuli designed to simulate the effects of real PPG signals as captured by consumer cameras, typically for the purpose of benchmarking, calibrating, or evaluating algorithms and devices in photoplethysmography-based heart rate (HR) measurement, especially on smartphones. These synthetic videos serve as standardized, repeatable inputs that can drive robust bench testing of smartphone health apps, particularly when manual or human-subject testing is logistically infeasible or when highly controlled physiological conditions are desired.

1. Principles and Motivation for Synthetic PPG Video Generation

Synthetic PPG test videos were introduced to address several core challenges in the evaluation of smartphone-based HR measurement:

• Device Fragmentation and Standardization: The vast diversity of camera hardware, image sensors, and firmware in smartphones introduces substantial measurement variability, complicating app evaluation across platforms.
• Scalability and Throughput: Human participant-based testing is limited by logistical, temporal, and ethical constraints. Synthetic videos allow many devices to be tested in parallel, accelerating algorithm and hardware validation.
• Controlled Physiological Simulation: Synthetic videos can precisely encode desired heart rate values, waveform morphologies, and a variety of signal perturbations (e.g., baseline drift, motion artifacts), facilitating robust performance analysis on edge cases seldom encountered in naturalistic testing.
• Benchmark Consistency: Standardized input signals enable objective comparison of different apps and devices over identical conditions, improving reproducibility in research and regulatory processes.

2. Methodology for Generating Synthetic PPG Test Videos

The generation of synthetic PPG test videos as described in the high-throughput bench-testing platform involves several key steps:

1. PPG Waveform Synthesis:
   • Synthetic PPG waveforms are generated using tools such as the NeuroKit PPG simulator. Parameters including heart rate, duration, respiratory sinus arrhythmia, baseline drift, and motion artifacts are defined to match physiological or experimental needs.
   • Real-world PPG waveforms from public datasets (e.g., MIMIC-III) may also be used.
2. Waveform-to-Pixel Mapping:
   • The synthesized PPG signal is downsampled to the intended video frame rate.
   • Each PPG value is inverted and linearly scaled to the 8-bit RGB pixel intensity range (0–255), with the mapping based on empirical pixel value distributions recorded from real finger-over-camera measurements under varying conditions (including skin tone and lighting).
3. Frame Construction:
   • For each frame $t$ and each RGB channel $c$, an $n \times m$ matrix of pixel values ($\mathbf{F}_{t,c}$) is drawn from a Gaussian distribution centered at the target value for that channel and frame:

   $$\mathbf{F}_{t,c} \sim \mathcal{N}(\mu = PPG_{t,c}, \sigma)$$

• The spatial mean of the frame aligns exactly with the downsampled PPG value for that time and channel, encoding the pulse waveform into spatio-temporal color fluctuations.

1. Compilation into Video:
   • The stacked RGB frames are written to a video file (e.g., MP4) with a frame timing corresponding to the sampling of the mapped signal.

This process produces synthetic videos in which the mean color intensity of each frame—when spatiallyly averaged—directly encodes the original PPG waveform, closely replicating the input signal morphology and HR dynamics.

3. Integration into High-Throughput Bench Testing Platforms

In the described platform, up to 12 smartphones are mounted facing a monitor displaying the synthetic PPG video. The smartphones, running their own HR measurement apps, record this stimulus as if a finger were applied to the camera. The measured HR and, if available, the extracted PPG waveform are collected for analysis. A host computer orchestrates playback and data logging for all devices.

This setup enables:

• Concurrent testing: Multiple devices receive identical, temporally aligned synthetic physiological signals, allowing direct device-to-device performance comparisons.
• Control and repeatability: Any desired waveform (including those representing rare or pathologic conditions) can be synthesized and replayed in repeatable fashion.

4. Validation and Performance Metrics

Validation of the synthetic PPG test video approach is performed by comparing the HR input encoded in the synthetic video against HR outputs measured by a clinically-validated smartphone app (e.g., Google Fit HR). Key metrics used include:

• Mean Absolute Percentage Error (MAPE):

$$\mathrm{MAPE} = \frac{1}{N} \sum_{i=1}^{N} \left| \frac{y_i - \hat{y}_i}{y_i} \right| \times 100\%$$

where $y_i$ is the reference HR and $\hat{y}_i$ is the measured HR.

• Correlation Coefficient (Pearson’s $r$):

$$r = \frac{\sum_{i} (x_i - \bar{x})(y_i - \bar{y})}{\sqrt{\sum_{i}(x_i - \bar{x})^2} \sqrt{\sum_{i}(y_i - \bar{y})^2}}$$

where $x_i$ and $y_i$ are paired input and measured PPG values.

The platform achieves:

• MAPE between input and measured HR: $0.11\% \pm 0.001\%$
• Correlation coefficient (input vs. measured PPG waveforms): $0.92 \pm 0.008$
• For HR, correlation is 1.0
• Bench-testing using synthetic videos classified all tested smartphones as passing industry accuracy standards (ANSI/CTA: MAPE < 10%), with these results highly predictive of clinical (in vivo) performance.

5. Significance for Device Compatibility and Mobile Health Evaluation

The use of synthetic PPG test videos in bench platforms delivers several advantages for the field:

• Compatibility Assurance: Uniform, repeatable signals allow detection of device-specific hardware, software, or firmware anomalies that may not appear in clinical testing due to uncontrolled variability.
• Development Acceleration: Synthetic videos enable automated, high-throughput regression testing and QA over new device releases or app updates, with the same rigor as human-subject protocols.
• Standardization: The approach establishes a quantitative, repeatable framework for regulatory and industrial compliance assessments, supporting the drive toward widely adopted accuracy standards in mobile health.
• Coverage of Edge Cases: Developers can proactively test scenarios that are difficult or unethical to reproduce with human participants, such as extremely low or high heart rates, severe signal noise, or artifacts.

6. Limitations and Future Prospects

Synthetic test videos inherently do not simulate certain real-world complexities, including:

• The full optical and physical interaction of a finger with the camera and flash (e.g., skin deformation, pressure variability, non-planar contact)
• Variations introduced by human movement and behavioral anomalies
• LED/flash illumination spectrum effects

A plausible implication is that while the synthetic video approach is extremely effective for “signal pathway” validation and algorithm benchmarking, it should be complemented by selected clinical/ergonomic studies to fully qualify real-world app performance.

Future extensions may include the injection of user-behavior artifacts, further customization of spectral noise profiles, or adaptation to new sensing modalities.

---

Summary Table

Step/Aspect	Description
PPG waveform simulation	NeuroKit PPG or real data; HR, artifacts controllable
Mapping to RGB video	Framewise mean pixel value encodes PPG for each channel
Spatial noise addition	Gaussian-distributed 8-bit pixel values per frame/channel
Video generation	Frames compiled at target framerate as stimulus MP4
Bench platform use	Monitor displays video; devices record; outputs logged
Validation metrics	MAPE, correlation coefficient (input vs. output HR/PPG)
Role in health/device QA	Scalability, standardization, predictive accuracy

---

Synthetic PPG test videos form a foundation for scalable, standardized, and high-fidelity evaluation of smartphone-based HR measurement, supporting both algorithmic innovation and robust device compatibility assessment in mobile health.