Wrist-wearable fitness bands and smart watches are moving from basic accelerometer-based “smart pedometers” to include biometric sensing such as heart rate monitoring. This trend is being driven by manufacturers who are looking for differentiation in the rapidly growing wearables market and by educated consumers who want to maximize their performance and fitness with more effective training.
-
Measurement accuracy of heart rate
-
Tracking accuracy while exercising
-
Validating performance across many unique individuals
-
Reducing size and thickness of wearable designs
-
Extending battery life
Wearable system designers have the choice of implementing non-invasive optical or electrical techniques. Electrical techniques typically use chest straps with two contacts on the skin to measure heart rate. This method is usually accurate but requires end users to wear a bulky band around their chest with a battery and wireless communication, which can be uncomfortable and inconvenient. Optical techniques are preferred as they eliminate the uncomfortable chest strap by using an optical emitter and sensor integrated with the wristband.
Principle of Wrist Based Optical Heart Rate Monitoring
As shown in Figure 1, infrared (IR) light from an LED shines on the skin of the user’s wrist. Light entering the skin gets absorbed, scattered and reflected from tissue, bones, veins and arteries. An optical sensor is used to detect these weak reflections. Light reflected from tissue and bones is time-invariant and results in a direct current (DC)-only level. As the heart pumps, the light reflected from arterial blood changes and results in an alternating current (AC) signal. Advanced digital signal processing removes the DC signal and calculates the heart rate from the weak AC signal. This processing technique generates what is known as a photoplethysmography (PPG) signal.
Many factors contribute to accurate PPG signals in an optical design. These factors include the wavelength of the optical emitter(s), the sensitivity of the system at those wavelengths, spacing between the emitter and sensor, the amount of light generated by the LED, the number of emitters and noise in the system. The ideal system for heart rate monitoring for one user may not work well for another due to differences in skin pigmentation and other physiological characteristics of the unique wrist. Designers must choose carefully and use adaptive techniques where possible in order to work well with a broad consumer population. Failure to do so can result in dissatisfied customers and a high return rate of wearable products. Let’s now consider a few of the critical aspects of heart rate monitor system design.
Wavelength of Emitter
Typical fingertip and ear-based heart rate monitoring systems use infrared (IR) emitters in the 850 nm to 940 nm band. IR emitters are not ideal for wrist-based solutions as the wrist does not have the same blood-rich capillaries close to the skin found in fingertip or ear locations. Green (525 nm) wavelengths have been found to provide superior performance in wrist locations for people with light complexions. Unfortunately, dark-pigmented skin can absorb green wavelengths. Yellow (590 nm) wavelengths have been found to work best with dark-pigmented skins. For best performance across the widest number of users, both green and yellow LEDs can be used at a small tradeoff to increased cost and power consumption. An adaptive technique is used to select the best signal to calculate heart rate for each unique individual.
“Opto-Coupling” to Skin
It is important to have good coupling of the optical signals between the wrist and the wearable device as air gaps reduce accuracy. Having a flexible wristband enables a close but comfortable fit. If the band is too tight, coupling is improved but blood flow can be restricted, resulting in worse accuracy. If the band is too loose, it will move freely. A common issue with single-LED systems is that
the wristband may need to be adjusted farther up on the arm or rotated slightly for optimal performance. Using two LEDs spaced on opposite sides of the optical sensor minimizes issues due to placement of the band as well as with tilting. Tilting issues can occur during exercise when one side of the wristband has good skin coupling while the other side has an air gap. Three LEDs are recommended for wearable designs to ensure the highest accuracy across a wide cross-section of end users. For example, the high-performance Scosche Rhythm Plus fitness band uses two green LEDs and one yellow LED in a triangular setup.
Emitter Light Energy
The emitter light energy is primarily determined by the LED drive current, voltage, pulse on-time, half-angle and luminous intensity. Having an optical sensor system that enables control over several of these parameters allows the software to configure itself optimally for each unique individual. For example, green LEDs have a high forward voltage that may require a tradeoff with LED voltage and power output. A higher voltage for the LED is not always feasible, so a longer pulse-on time can be used to increase emitter light energy while staying within the LED’s normal operating parameters. An auto-sensing function can adjust the LED driver current and/or the LED on-time to optimize the reflected signal for a unique individual. This automatic DC sensing helps minimize the dynamic range requirements of the system’s analog-to-digital converter (ADC) and places the signal in the optimum range for detecting the weak AC heart rate signal.
Tracking Accuracy When Exercising
The biggest weakness of most commercially available wrist wearables is that they cannot accurately track heart rate while the user is exercising. It is very challenging to adequately compensate for motion and physiological artifacts when exercising. Accelerometers are typically used in wearables and can be effective when combined with advanced signal processing to reduce motion artifacts. These algorithms can use accelerometer data to reject heart rate samples corrupted by noise or to actively cancel the noise. Despite these algorithms, the heart rate signal can be temporarily lost. An adaptive algorithm that recognizes when the sensor data is not valid by using quality grading enables estimation techniques to provide consistent tracking accuracy when the user is exercising. It is also important to validate the algorithms with a large sample of users representing different skin tones, ethnicities, ages and weights.
Reducing End Product Size and Thickness
Adding heart rate sensors requires more space in a wearable design. Many existing heart monitoring wearable designs use large discrete photodiodes combined with an analog front end (AFE) and MCU. The AFE includes the LED drivers, ADC, analog filtering and control. A smaller high-sensitivity photodiode integrated with the ADC along with analog filtering and LED drivers can have much lower noise floor, use fewer ADC bits, and smaller footprint. For example, Silicon Labs’ Si114x optical sensor integrates high- sensitivity photodiodes, a 17-bit ADC, low-noise analog filtering, up to three dynamically configurable LED drivers and an I2C interface in a compact 2 mm x 2 mm clear QFN package. A typical AFE in 3 mm x 3 mm package and discrete photodiode in a 2 mm x 2 mm package has a 3x larger footprint.
Increasing Battery Life
The largest consumer of power for a heart rate monitor is typically the LED sampling power and the signal processing for motion artifact reduction. A key factor in power consumption is the sample rate used. Accurate heart rate monitoring can require faster sample rates when exercising with high beats per minute (BPMs). Using a dynamic algorithm that changes the sample rate based on BPM can maintain accuracy while minimizing power. Interpolation of samples can also enable lower power than raising the sample rate. Sensors that can dynamically change their LED drive currents can auto-sense DC level to reduce power and improve performance. Dynamically using one, two or three LEDs in the system design can also maintain high performance while minimizing power consumption.
Conclusion
Designing an optical heart rate monitoring solution in a wearable product presents many technical challenges. A high-performance integrated optical sensing solution such as Silicon Labs’ Si114x sensor family, combined with heart rate algorithms validated under clinical studies, enables developers to design robust wearable systems that maximize battery life and minimize the physical size of wrist-based heart rate monitoring devices. Silicon Labs has taken a major step toward making optical heart rate monitoring an integral feature for the fast-growing consumer wearable-device market.