Company News
GMEMS-A high school team on the other side of the ocean
- Time：2021/5/9 Posted：GMEMS (Shenzhen) Technology Co., Ltd.
- A team of scholars on the other side of the ocean
General Micro (GMEMS) is such a company. It adopts a combination of software and hardware, and proposes a leading end-to-end voice interaction solution, which groundbreakingly solves the problem of complete ingestion of voice signals in a high-noise environment, that is, The cocktail party issue mentioned earlier.
General Micro CEO Dr. Wang Yunlong told Yiou: "The reason why this problem can be solved is due to the strong R&D strength of General Micro." Yiou understands that Universal Micro has strong strengths in three major aspects of hardware, software and algorithms. . "
In terms of hardware, General Micro Technology can integrate IC and MEMS design and packaging. It has independent research and development of MEMS microphone chips, and its products have been updated several times. The main product currently being supplied is the GMEMS2.0 chip with a signal-to-noise ratio of 66dB. The world's leading GMEMS2.5 chip with a signal-to-noise ratio of 70dB and a dynamic range of 135dB or more will soon enter mass production in the second half of 2019.
GMEMS also plans to launch GMEMS3.0 with a new subversive four-diaphragm architecture. The signal-to-noise ratio can reach 76dB to 80dB on the same platform as measured. Wang Yunlong said: “GMEMS3.0 will surpass similar products from Lou's Electronics, Goertek, STMicroelectronics and other large companies to reach the industry-leading level.”
In terms of software, Universal Micro has three core technologies: speech processing, adaptive beamforming, and voiceprint recognition. Through adaptive machine learning sound spectrum analysis, it can effectively separate the user's sound source, interference, background noise, and echo, and there is no need to debug due to dynamics. Problems caused by noise, echo, and unstable signal sources. The software's acoustic design error tolerance and signal unsteady state tolerance are both high. Through the training of the voiceprint library, the software can effectively recognize the user's voice, and can effectively solve the cocktail party problem and robustness problem.
In addition, in terms of algorithms, Universal Micro uses the world's leading deep learning algorithm to easily solve the voice interaction function of the end test, and will realize functions including micro sensor array, wake-up function, noise reduction, echo cancellation, and voice enhancement. At the same time, the algorithm has the advantages of high wake-up rate, small amount of required wake-up word data, and high wake-up accuracy, which is better than similar algorithms on the market.
General Micro's strong technical strength stems from the strong technical team behind it. CEO Dr. Yunlong Wang introduced that the company currently has more than 150 people and the core R&D team is located in the United States. The company has nearly 40 people in research and development. The company's chief scientist Li Xilin is from Tsinghua, CTO Wu Guanghua is from the University of California, Berkeley, and VP Lu Yanzhen is from the University of Sheffield, UK. In addition to myself, the four core founders of the company are all PhDs, thus ensuring the company's continuous leadership in technology.
