Rilispedia.com – KaiKuTek and other notable homegrown startups led by Taiwan Tech Arena (TTA) had showcased their innovative solutions and sought for global cooperation at CES 2020.
Mike Wang, the technical director of KaiKuTek, has lived in the United States for more than 30 years. He has more than 25 years of experience in analog/RFID design. He saw the potential of the application of the 60GHz mmWave WiGig (Wireless Gigabit, 802.11ad) in gesture recognition and went forward to establish KaiKuTek in January 2017. After raising US$10 million in seed round and Series-A funding, KaiKuTek has successfully developed the world’s first 60GHz gesture recognition SoC that combines mmWave, a deep learning algorithm and AI accelerator. It is expected that mass production will begin by the end of the year.
Taiwan ranks highly in mathematical proficiency, and with the heavy promotion of AI by the government and the industry over the years, the country has cultivated many AI talents. It is because of this that Mike Wang decided to return to Taiwan and set up a company. However, he believes that most of the AI applications in Taiwan are currently limited to a few areas like finance and shopping. Realistically, maximum synergy can only be achieved when AI is combined with hardware. At present, there is an enormous startup manufacturing industry chain in Taiwan. KaiKuTek’s AI accelerator can meet the needs of specific applications of AI and core technology development in this industry and other markets.
“We have fused three major technologies, each having a high threshold, into one to create the 60GHz gesture recognition total solution.” Mike Wang said: “The first is mmWave technology that includes radar signal processing, IC and antenna design. The second is deep learning and an AI algorithm. The third is AI.”
This powerful integration skill spawned the world’s first fully integrated embedded system that combines the 60GHz mmWave radar and dedicated AI accelerator. KaiKuTek has demonstrated its patent planning and mapping ability, including the special mmWave IC circuit design know-how that improves efficiency and reduces power consumption, as well as the special AI/deep learning algorithm that improves accuracy and reduces IC implementation, power consumption and complexity. In addition to providing a complete 60GHz hand/finger gesture recognition/tracking SoC and relevant algorithms, KaiKuTek will also develop other mmWave/deep learning technologies for different markets and applications. Moreover, KaiKuTek will also dive into sensor fusion in the future.
KaiKuTek has selected 60 GHz because it is a free and unlicensed band. Its maximum bandwidth can reach 10Gb/s (current maximum available bandwidth) in the future. Many other standards, including 802.11ad/WiGig and 802.11ay, are all on this band. More importantly, its radial distance resolution is 1.5 cm. This distance is especially important for gesture recognition applications because finger movement is very subtle. In addition, mmWave signal attenuation in the air occurs rapidly, so 60GHz is ideal for line-of-sight and short-range applications (fewer than 10 m). With a directional antenna, it can provide better privacy protection and reduce interference.
Mike Wang emphasized that the product adopts sensor-side computing, which has low-latency, provides immediate benefits including very economical. Specifically, applying a deep learning algorithm to sensor-side SoC eliminates bandwidth issues commonly found in cloud computing. Also, compared to being run on software processors, a deep learning algorithm can provide lower latency and power consumption when combined with the dedicated AI processor on the SoC.
At present, gesture recognition/tracking has a wide range of applications. These include mobile phones (provides more diverse control experience), IoT wearable devices (smart watch, headset, smart bracelet), games (motion detection, game control), smart home/appliances (TV, lighting controls), in-vehicle controls (audiovisual controls, call answering), and security (keyless entry, alarm system). KaiKuTek has three main targets: smartphones/wearable devices, games and AIoT. However, startups are limited by the economy, costs and resources, KaiKuTek will prioritize the first two targets.
“We founded the company and developed the mmWave radar 3D gesture recognition/control project because we did not want to be a ‘me too’ trend-follower. We increased our value and advantage through innovative integration methods. Also, by combining hardware and services, we can provide customers with customized value-added services,” said Mike Wang. At present, KaiKuTek provides different business models such as built-in module supply and personalized gesture and gesture app development platform tailored to mobile devices/wearable devices, consumers and developers. Personalized gestures allows the user to add gestures and upload them to the cloud via the paid app. After training is completed on the cloud server, the gestures will be updated to the device. Lastly, KaiKuTek will establish a profit-sharing system with service providers to build a sustainable, profitable business model.