Standing on the back of the Chinese Alpha dog, Keda News

In addition to Chinese football, the recent eye-catching event, Google’s Alphago and Li Shishi’s Go War, how artificial artificial intelligence is, see how the University’s core technology development was held on March 30 at Shenzhen’s Archimedes Internet Corporation. Days - Mai into the cloud, leading the smart interaction neo-ism" Salon activities will know.

Photo: The venue is popular

At the event site, HKUST shared its understanding of artificial intelligence and efforts to achieve this, and brought a complete intelligent hardware voice interaction solution and open platform core interaction technology.

Photo: Ma Hanjun, deputy general manager of Keda Xunfei Open Platform

Ma Hanjun, deputy general manager of Keda Xunfei Open Platform, said that artificial intelligence is becoming the trend of the times, and now it has entered the eve of the real outbreak. At present, AI has caught up with humans in computing intelligence and perceptual intelligence. Alphago is a good example of victory over Li Shishi, and cognitive intelligence is the key challenge of current artificial intelligence. The rapid development of artificial intelligence will also help the boom of the IOT era, and the vision of the University of Science and Technology is to accelerate this process. At present, the application of Xunfei's natural open interactive platform has exceeded 90,000, and the number of connected terminals exceeds 880 million. Next, Xunfei will realize the artificial intelligence from "can hear and talk" to "through the combination of deep neural network and big data." A breakthrough that can understand and think. In 2015, Xunfei released AIUI, an intelligent interactive solution tailored for intelligent hardware, which redefines the human-machine voice interaction standard in the era of Internet of Everything, with far-field noise reduction, dialect recognition, speech correction, and more. Functions such as round dialogue, through the unified interface, open service expansion, flexible ability to match the human-computer interaction and product experience. HKUST News hopes to work with partners to change the world with artificial intelligence.

Photo: Zhang Liangchun, Product Manager, Feiyun Platform Business Unit, Keda News

In this event, Zhang Liangchun, product manager of the Flying Cloud Platform Business Unit of Keda News, brought the latest weapon to break the bottleneck of far-field voice interaction - the five-microphone array with a far-field pickup distance of up to 5 meters and 360 support. ° Full-plane pickup angle, support continuous wake-up, echo cancellation, and voice interruption. Zhang Liangchun said that using the microphone array module of Keda Xunfei can help developers to develop products quickly and save product development cycle. Xunfei also offers soft core solutions including single wheat, double wheat, four wheat linear, ring five wheat and ring seven wheat, especially the ring seven wheat program supports 7 meters pickup distance, 360° sound source positioning, ±10° sound Source positioning accuracy, the audio signal-to-noise ratio of the array input is higher. The hardware solution with the voice wake-up closed-loop optimization service, speech recognition, semantic understanding deep customization, identity authentication, natural speech synthesis and other capabilities of the Keda Xunfei open platform can make the developer's products "as bright and outstanding as the fireflies in the dark."

Photo: Dr. Wang Haikun, Institute of Science and Technology, University of Science and Technology

Dr. Wang Haikun from the Institute of Science and Technology of Keda University said that noise, reverberation, interference and echo are the problems that need to be solved in acoustic signal processing. The main functions of the microphone array are sound source point, suppression of background noise, signal extraction and separation, which are the solutions to the above problems. The best means. The sound source localization technology uses the microphone array to calculate the angle and distance of the sound source distance array to realize the tracking of the target sound source; and then through the beam forming technology, that is, a beam is effectively formed in the desired direction, and only the beam signal is picked up, thereby Achieve simultaneous extraction of sound sources and suppression of noise. For reverberation, efficient de-reverberation processing is achieved using microphone array + depth learning. As for the processing of the echo, the adaptive filtering method is mainly used to estimate the echo path and use the reference signal to obtain the horn sound picked up by the microphone and subtract it. Xunfei's echo cancellation technology has reached the international top level, and all the acoustic processing technical indicators are international leaders. Through this series of technologies, the Xunfei microphone array achieves excellent acoustic signal processing, ensuring a good user experience.

Photo: Tang Xi, Director of Intelligent Hardware Business, Feiyun Platform Business Unit, Keda News