The team of National Yang Ming Chiao Tung University (NYCU) led by Assistant Professor Yao Wei Huang, the Yushan Young Fellow Scholar at the Tin Ka Ping Photonics Center and Department of Photonics and the Dr. Hao-Chung Kuo at Semiconductor Research Center of Hon Hai Research Institute (HHRI) and the team led by Dr. Yu-Heng Hong have collectively and successfully developed the “Metasurface- and PCSEL-Based Structured Light for Monocular Depth Perception and Facial Recognition.” This new system is not only able to accurately recognize the face of Michelangelo’s David Statue but also significantly reduces the power consumption and the area of optical components so that the wish for future consumers to buy a “no notches” mobile phone is just around the corner!
The collaboration between NYCU and Hon Hai Technology Group, developing a face recognition system with a size that is much smaller and with lower power consumption than the system now developed by the 3D Surface Imaging technology [1]. In addition to the mobile phone application, it can be used in the face recognition system for game systems and the banks’ online login accounts. This significant research result was published as “Metasurface- and PCSEL-Based Structured Light for Monocular Depth Perception and Facial Recognition” in the leading global journal – Nano Letters. The article was featured on the cover of the February issue, highlighted as the cover story, and reported by the American Chemical Society. Corresponding patent applications have been successfully granted in Taiwan and China, and are in the process of being filed in the United States.
In the prototype test on a replica of Michelangelo’s David Statue, the depth information created by the system in comparison to that obtained from an iPhone dot matrix projector can be used to accurately recognize facial features and achieve close recognition confidence scores, representing its commercial feasibility. Notably, it uses approximately 0.15 times less power than the existing dot projector systems and requires 233 times less optical component area. In addition, the number of structured light points increases by 1.34 times, and the field of view increases by two times. This technology is highly miniaturized, lightweight, and energy-efficient, making it easy to integrate into spatial computing, mobile, and wearable devices.
Assistant Professor Yao Wei Huang stated that the size of the proposed system is approximately as wide as three human hairs. In the future, it is expected that the notches on the mobile phone screen will disappear, and the face unlocking function will be more power efficient, practically solving the technological challenges. This research demonstrates the potential applications of effective, miniature, low-power imaging solutions.
This research was supported by Hon Hai Precision Industry Co., Ltd., the Forward-looking Cooperation Project of the National Science and Technology Commission (Principal Investigator: Prof. Edward Yi Chang, Chair Professor of the International Semiconductor Industry College at NYCU; Co-Principal Investigators: Prof. Tien-Chang Lu, Chair Professor of the Department of Photonics at NYCU, and Prof. Chun-Hsiung Lin, International Semiconductor Industry College at NYCU), and the Youth Project Grant from the Ministry of Education.
[1] 3D surface imaging is a common application in smartphone facial recognition, computer vision, and autonomous driving. These systems usually consist of a dot matrix projector that contains several components: a laser, a lens, an optical waveguide, and a diffractive optical element. The diffractive optical element is a special grating that divides the laser beam into about 32,000 infrared dots. Therefore, when people look at a locked screen, the facial recognition system projects a series of dots onto their faces, and the device’s lens captures the infrared dots along with the established depth information to confirm their identity. However, the current dot matrix projector systems are relatively large for small devices such as smartphones.
