When integrating biometric sensors in a hearable or wearable, the project team may come across a number of challenges that have the potential to slow down the product development cycle. Whether it’s mechanical, electrical, software, or testing procedures, let’s take a look at common questions from an engineering point of view that may arise during the product development process and how solving potential issues early can minimize setbacks and get your product to market successfully.
Valencell also conducted a webinar on this topic that you can see here:
Product Management Team
Before we get into the technical aspects, it’s important to understand the bigger picture concerns that a product or project manager might have when integrating a biometric sensor into a wearable and how it has the potential to slow the product development cycle. Understanding this process and how potential delays can be avoided is a key success factor.
In most cases, sensor integration with the mechanics, optics, and electronics should be planned as early as possible in the production cycle, so most issues can be solved before it’s too late in the manufacturing process. This can be done by validating form, fit, and function of the sensor in the design mockup stage before the ID freeze takes place.
It’s also essential to conduct early reviews with the biometric sensor provider to ensure power supply considerations are validated before engineering samples are given to avoid delays with the electronics. Electronic integration and assembly can impact the performance of a unit once it’s mass produced, so it’s important to think through the manufacturing test plan ahead of time as well. This will also ensure that assembly of the wearable doesn’t compromise the ability to test the biometric sensors at the end of development, which is essential to keeping the production cycle on track.
Planning early for these aspects of production can help avoid the majority of potential pitfalls with biometric sensor integration.
The Mechanical Engineering Team
The mechanical engineering team is likely to have the most questions about potential issues when it comes to integrating a biometric sensor. One of the most common areas of concern is where a sensor can be positioned.
From a system perspective, location and accuracy are determined primarily by assessing the signal-to-noise ratios. While the short answer is that a sensor can be placed almost anywhere with some modifications to the design and algorithms, there are areas of the body that are superior to others. The fingers, ears, and head are ideal locations partly because of the higher blood perfusion, providing more physiologically relevant PPG signals than the limbs. Additionally, the ear has the benefit of substantially lower motion artifacts throughout exercise and everyday life activities. That said, with the right optomechanics, high-performance biometrics can be achieved on the wrist and arm as well.
The two other major concerns for a mechanical engineering team are: 1) the factors that influence the size of the sensor on the front end and 2) how a sensor can be attached to a wearable. The size of the sensor module is usually determined by the intended use and form factor of the device, as a hearable will need to be smaller than a wearable. One key consideration in the size of the sensor module is the spacing between the emitter and detector, since wider spacing captures less total light but has a higher ratio of good light to bad light. When there isn’t a lot of space, engineering techniques can be used to lessen the spacing impact.
Regarding sensor attachment to the wearable, the best attachment option is normally selected depending on a particular manufacturer’s process preferences and costing. Ultrasonic welding is often selected based on preferential experience, with factors such as the wearable’s material selection and surface finish, the manufacturer’s fixture design and process control, and the consideration of a power-on-self-test all contributing to the decision process. Alternately, many manufacturers prefer to attach the sensor with dispensed adhesive, citing considerable experience with gluing plastics or a desire for wearable materials that may not be compatible with ultrasonic welding. Several adhesives have been successfully used in production, but it should be noted that cyanoacrylate is not the preferred option.
Form-Factor Specific Mechanical Questions
In addition to the mechanical engineering concerns outlined above, there are a few other form factor specific questions that often come up during design for both hearables and wearables.
For hearables, which ear is best for the sensor and if the skin needs to be in contact with the sensor are the most common questions that come up. Though a hearable can be used in either ear, the choice is usually driven by design characteristics. For example, if the control mechanism (buttons or touch) is in the right earbud, placing the sensor in the left earbud to avoid noise interference is usually the best option.
Sensors also don’t necessarily need to be in direct contact with the skin, but it helps. The skin surface can be used to stabilize the device and sensor module, which is important for accuracy because motion between the skin and the sensor causes problems. External light also causes problems, and space between the skin and the sensor potentially exposes the sensor to more external light.
Electrical Engineering Team
While there are a lot more exceptions to the rules from a mechanical perspective, questions or concerns from an electrical engineer are more straight forward by comparison. The most common is which interface method should be chosen—a UART or an I2C?
The answer to that depends on the availability of the interface at the host. If the host controller can support both types of interfaces, other considerations come into play. For instance, I2C will be the best option if there are multiple peripherals on the same bus, or if there is a preference for a clocked interface. On the other hand, UART can be a popular choice due to its simplicity. If power consumption is the primary concern, a more detailed analysis is necessary to determine the best choice. The tradeoffs involve the differences between static and dynamic power consumption of UART vs I2C as well as chip-specific details of each peripheral’s implementation.
Another electrical engineering concern is the power supply. The system should be designed to support both the average and peak currents of the sensor system and ensure that electrical noise doesn’t degrade biometric sensor performance. Best-practices are to use low noise power supplies and isolate sensor circuitry from the potential noise sources.
As far as the sensing system, VDD for the MCU is typically going to be different than the voltage that drives the LED. Furthermore, VLED is likely to be different for a wearable and hearable. This is due to dependence of the turn-on voltage of the different LEDs that one would use for a hearable versus a wearable.
The next step in the process of working through potential biometric sensor integration issues from a software engineer’s point of view. Some of the common questions that might occur during the integration process include:
- What information must be supplied to the sensor?
- What information is directly measured versus what is a derived metric/assessment?
- How is performance optimized depending on the use case?
To get the most accurate results, activity characterization and resting heart rate of the individual are the most important pieces of information that must be supplied to the MCU. This will help to make assessment for things like VO2 max and energy expenditure as accurate as possible. When the information is not supplied, the default mode will process results—though not with the same precision. Weight, age, and gender are other key pieces of information that can improve accuracy when coordinated with data from the biometric sensor.
Once you have direct measurements of continuous heart rate during activity, resting heart rate, and physical activity characterization (such as step count, running pace) from the optical sensing system or accelerometer, further assessments can be made. For instance, V02 max metrics can be derived once we know heart rate and distance.
When it comes to various use cases and optimization, the goal is to use as little power as possible while still being able to deliver the desired features to the user.
With Valencell’s PerformTek technology, this can be accomplished by through the feature activation and algorithm configuration, where power consumption is optimized based on the activity. The host controls the power consumption by deactivating algorithms that are not needed for that specific use case in order to save power.
Testing for Accuracy
Once a biometric sensor has been integrated into a hearable or wearable, ensuring accuracy of the device is a priority. When trying to determine how accurate a particular device will be, testing protocols are developed based on the use case of the wearable. If it’s a fitness device, a protocol for each of the exercises the user may undertake (such as running, cycling, lifting weights etc.) will need to be developed for testing. If it’s a lifestyle device, you’ll need a different protocol because the optical noise will vary depending on the environment the device is being used.
It’s also important to remember that accuracy is a statistical parameter. Testing one person or even a few cannot guarantee accuracy across a broad and diverse user base. At Valencell, a minimum of 30 individuals are tested with a focus on maintaining a diverse representation of physical habitus, gender, ethnicity, skin tone, and even hair coverage. This means hundreds or thousands of data sets per product launch. Once you have sufficient diversity in your testers to accurately reflect the population of your users and you’ve done a lot of testing protocols, basic statistical analysis can be done.
Once devices move to production, testing at the production line may come up as another potential concern. During the development process, the design’s accuracy is verified. This same accuracy needs to be obtained at the manufacturing line to ensure the device-under-test is within specification.
For Valencell’s benchmark modules, each of the following failure mechanisms are tested with a custom test system to ensure the quality of our product:
- Electrical failures
- Optical emitter issues
- Optical detector issues
- Mechanical issues in the optical path
Hopefully this has been valuable in understanding the different considerations that different parts of your product development team will encounter in building a biometric wearable or hearable. If you have questions or would like further detail on any of this, please let us know at email@example.com.