3D facial recognition has come into the spotlight after Apple’s introduction of Face ID in the iPhone X. Companies in the supply chain have also become the new focuses of investors. In fact, 3D sensing technology has been around for years, but did not have commercial applications of facial recognition in smartphones until 2017. The reasons why it happened in 2017 can be explained in terms of components, modules and subsystems, computing and systems, end-devices and applications.

 

First, components are crucial to the commercial uses of facial recognition. A series of articles by LEDinside have illustrated how these key components have met smartphone users’ needs, such as unlocking devices conveniently and securing mobile payment, from built-in authentication methods like passwords and unlock patterns, to more advanced biometric recognition, such like fingerprint, iris scanning and 3D facial recognition. Fingerprint scanning uses capacitive sensors, and iris scanning relies on LED IR and IR cameras. As for 3D facial recognition, the key components of structured light solutions include IR lasers, diffractive optical elements, and IR cameras. New generation solutions integrate more and more components to provide superior user experience.

 

Moving on to modules and subsystems. Compared with products of traditional IC companies, a fingerprint scanner requires an additional manufacturing process of coating. IC companies either complete the design and manufacturing all by themselves, or outsource this part to professional modules makers. When it comes to iris scanning and structured light solutions, the individual modules of TX (light source) and RX (IR camera) are combined to a subsystem. Traditionally, light source components and IR camera are made by different suppliers.

 

In TX modules, in addition to chip packaging, the integration of wafer-level optics and DOEs can also be found. In terms of camera, the manufacturing of lenses and optical sensors is separated for a long time. As 3D facial recognition uses structured light solutions, current development of modules and subsystem grows increasingly complicated and involves wider range of professional sectors, which is beyond the current capability of traditional IC suppliers. That is to say, if a single IC supplier wants to complete the whole fabrication process of subsystems, its investment in R&D and manufacturing will be tremendous. In addition, it will face considerable technical barriers and market risks, which traditional component makers cannot afford.

 

Thirdly, the breakthrough of algorithm is the main drive for the advent of 3D facial recognition in iPhone X. The algorithm of fingerprint and iris sensing is different from that of 3D facial recognition. In order to achieve self-adaptability, 3D facial recognition requires Neuro Network Engine to complete the process of facial recognition at reasonable computing speed and power consumption. This is also the main reason why Apple can release this product in 2017. However, there are still a few key considerations regarding the corresponding system. First, from the perspective of computer vision, the 3D modules and subsystems are designed to produce data that is more suitable for computers to process, which will in turn influence the selection and design of entire DOEs components and camera modules.

 

Second, the precision of subsystem assembly, determined by mechanic and material technology, has to meet the algorithm and the demand in end-applications, or otherwise, need to combine another reference of distance, such as TOF technology. These two considerations make it more complicated to initiate the production of new modules, because it is difficult to design a single module that can fit in a wide range of applications. In terms of the system, its use in unlocking smartphones should also consider power consumption. In order to detect the faces to be recognized in the state of low power consumption, the choice of IR camera or RGB camera will also make differences in the module design of 3D facial recognition subsystem.

 

The last point is end-devices and applications of this technology. Existing commercial uses of 3D sensing modules can be found in game consoles, personal computers (Windows Hello or RealSense technology), industrial sectors, and rear-camera of mobile phones. But these applications attract far less attention than Apple because iPhone records a sales of about 200 million units each year. Even the only one model currently available can contribute to the module demand of more than 70 million units throughout the year. Apple can afford large scale of research and development of this technology because of its high revenue, large gross profit margin and support from suppliers.

 

On the other hand, Android 3D facial recognition ecosystem is faced with two challenges. First, it is not sure whether iPhone X can show promising market prospect which is good enough for Android smartphone vendors to follow the trend. Second, in the independent operation of the Android smartphone supply chain, 3D sensing value chain consists of various parties, including the smartphone vendors, major wafer makers, component suppliers, and module makers. It is still unclear who will take the risk in the future and lead the products towards making profits. Only when joint understanding is achieved within Android eco-systems can the industries cross the technology chasm, and accelerate the large-scale adoption of 3D facial recognition technology.