IoT-Based Skin Biometrics Device for Melanin Detection and Hydration Analysis Using ESP32-S3

An upscale skincare company approached us with an ambitious request: to create a home skincare professional diagnostic tool that could classify skin as "dry or oily" based on skin's levels of hydration and melanin in the skin, instead of just by using the terms "dry" and "oily."

Using these criteria, we would create a small portable device that measures both the Subsurface Melanin Index (MI) and Deep Layer Hydration of the skin, and then use this information to recommend ultra-personalized serums for individuals' skin care needs.

Technical challenges were major:

Ambient Light Interference:

Measuring light reflection from within the skin (dermis) is notoriously difficult when the user is in a brightly lit bathroom or under direct sunlight.

Diverse Skin Tones:

The device had to be equally accurate across all Fitzpatrick Skin Types (I-VI). Most consumer sensors struggle with "darker" skin tones due to high light absorption by surface melanin.

Cross-Platform Synchronization:

The device needed to sync data instantly with a mobile app that uses AI to analyze skin "aging" trends over months.

We have developed a compact consumer device that achieves near clinical-grade accuracy by utilizing a hybrid sensing technique composed of Diffuse Reflectance Spectroscopy and Bio-impedance Analysis.

1. Dual-Channel Optical Engine

An optical chamber has been designed using a custom-built shroud that completely encloses the optical chamber and prevents any external light from entering the optical chamber, providing a level of consistency in measurements performed in different environments where ambient lighting may differ.

The system will be utilized with three pre-calibrated wavelengths:

• Green – (~568 nm) is predominantly absorbed by hemoglobin and can therefore be used as a measure of degree of red and/or degree of luminescence and/or degree of inflammation in tissue.
• Red – (~660 nm) will penetrate deeper into the dermal tissue and can, therefore, be used to assess the structure of the dermal tissue beneath the skin.
• Near Infrared – (~880 nm) can be used to estimate the Melanin Index (MI) based on the spectral absorption of the pigment in the tissue.

To help eliminate any interference from ambient light, the light emitting diodes will be pulse modulated at a frequency of approximately 1 kHz and the modulated signals will be synchronously detected using a lock-in amplification circuit.

This phase-sensitive detection method will isolate the desired signal resulting in a rejection of better than 99% of any ambient light interference, therefore yielding highly stable, repeatable measurements.

2. Multi-Frequency Bio-impedance (BIA):

Surface moisture measurement can be straightforward but measuring 'Deep Hydration' is much more complex, requiring assessments of cellular activity. We incorporated gold-plated micro-electrodes designed to send a low-level / imperceptible electrical current (2 frequencies) through the skin:

• 5 kHz - Measures the amount of extracellular water
• 50 kHz - Measures the amount of intracellular water by penetrating the cell membrane

By having both measurements available, the device is then able to differentiate between wetness present on the skin (which can be temporary) vs. true hydration of cells.

3. Firmware and Connectivity

The hardware utilizes an ESP32-S3 processor that has both Wi-Fi and Bluetooth Low Energy (BLE) built into the chip.

The skin type calculation takes place on the device using the ESP-IDF framework, which we used to build a locally executed "Inference Engine"; the skin type calculation occurs on device in less than 200ms and provides the user with instant haptic stimulation when the measurement is successfully taken.

To establish authority, we didn't just rely on "marketing" numbers. We built a validation pipeline that mirrored clinical trials.

Fitzpatrick Scale Calibration:

We conducted a study with 200 participants across all six Fitzpatrick skin types. By adjusting the "Gain" of the optical sensors dynamically based on the initial "Scout Flash," we ensured that the Melanin Index remained accurate within a ±2% margin of error regardless of the user's ethnicity.

Correlation with Corneometry:

Our BIA sensor was calibrated against the Corneometer® CM 825 (the industry gold standard). We achieved a Pearson correlation coefficient of $r = 0.89$, proving that a consumer-priced IoT device could provide professional-level insights.

Dermatological Logic:

The back end "Recommendation Engine" of Dermatological Logic was created with the collaboration from a group of Board Certified Dermatologists.

The recommendation engine utilizes a Decision Tree Classifier that considers the Melanin Index, Hydration Levels, and Local UV Index (via API) to make recommendations for specific Active Ingredients such as Niacinamide or Hyaluronic Acid.

In the skincare industry, trust is tied to consistent results and data privacy.

Secure OTA (Over-the-Air) Updates:

As skin care continues to improve, so do the updates that would allow the device's software to continue to utilize these improvements. To prevent the installation of an unauthorized update (malicious firmware) via an OTA update, we have added a Digital signature to the new updates using the ECDSA cryptographic signature algorithm. ECDSA will prevent a man-in-the-middle from installing malicious firmware onto the device.

Privacy-First Cloud Architecture:

All skin biometric data is anonymized. The "User Profile" on our AWS backend is decoupled from the "Biometric Log." Even if a database breach occurred, the skin readings could not be traced back to a specific individual's identity (GDPR & CCPA Compliance).

Hardware Durability:

The durability of this product can be attributed to the fact that it will reside in a bathroom and has been designed according to the rigorous IP67 standards for waterproofness. The case is made using ultrasonic welding with hydrophobic coatings applied on its optical lens to ensure that there is no build up of moisture or fog inside the lens.

The launch of the analyzer transformed the client's business model:

• 40% Increase in customer retention for their subscription-based serums.
• Reduced Product Returns: By ensuring customers were using the "Right Product for the Right Day," return rates dropped by 15%.
• Empowered Users: For the first time, users could "see" their skin recovering from sun damage or dehydration before it became visible to the naked eye.