지난 2월 20일 e4ds 웨비나에서 인피니언 김성태 부장이 ‘MEMS 마이크로폰의 기본 원리 및 AI 트렌드에 맞는 고성능 마이크로폰 선정 방법’을 주제로 발표했다.
.jpg)
Microphone Market to Grow
at an Average Annual Growth Rate of About 4% by 2026 Voice recognition accuracy ↑·Noise ↓·Long-distance voice capture
MEMS, small size, high performance… mass production, cost reduction As the voice assistant market based on conversational AI expands, experts suggest that high-performance MEMS microphones can improve the user experience of voice-enabled applications.
On February 20, at the e4ds webinar, Infineon Vice President Kim Seong-tae presented on the topic of 'Basic principles of MEMS microphones and how to select high-performance microphones that fit AI trends.'
Artificial intelligence (AI) naturally understands and processes human language to perform functions. Recently, generative AI capable of natural language conversation, such as LLM (Large Language Model), has been rapidly increasing, and voice input is being installed as a key interface in products. For example, Samsung Electronics' S24 series and Galaxy Book Pro have installed generative AI by applying on-device AI technology. Apple also announced plans to strengthen HW and SW functions by integrating conversational AI models into Siri.
The principle of sensing is similar to the human auditory and visual functions. Just as humans perceive the environment through their senses, such as their eyes and ears, devices also perceive situations through sensors. Sensors can collect more accurate and extensive information beyond the limits of human senses. In addition, they comprehensively accept input information from multiple sensors through fusion algorithms for more accurate recognition.
'MEMS high-performance microphones' are attracting attention as a tool for accurately detecting voices. According to market research firms, the overall microphone market is expected to grow by an average of about 4.2% per year from 2022 to 2026.
Microphones are used in a variety of fields, from consumer applications such as smartphones, AirPods (True Wireless Stereo), headphones, video call equipment, AI speakers, to cameras, surveillance monitoring equipment, VR devices, medical devices, and industrial devices. Microphones are increasingly preferred for their high-spec, clear sound quality. When communicating through video calls, video conferences, speakers, etc., high-quality recording is required.
The two most widely used technologies for microphone configuration are 'MEMS' and 'ECM (Electronic Condenser Microphone)'. MEMS can achieve high performance in a small size by implementing the sensing part through a semiconductor process. It is quickly replacing ECM due to its low cost advantage in mass production.
'MEMS (Micro Electronic Mechanical system)' refers to a device that converts external physical signals into electrical signals within an ultra-small semiconductor structure. A MEMS microphone contains a MEMS part that senses sound, an ASIC and a PCB that extract the difference in capacitance caused by the generated sound into an electrical signal.
MEMS microphones are made up of two structures: a diaphragm that moves due to air vibration called a 'membrane' and a fixed diaphragm. When the membrane moves according to the vibration of the air at a certain distance, it moves up and down and the distance from the fixed diaphragm changes. At this time, the difference is recognized as an electrical characteristic and output.
Infineon is producing 'SBP MEMS (single)' and 'SDM MEMS (dual)'. The SBP structure is simple, so it can be made small and is relatively inexpensive, but SDM is a method to reduce noise and obtain a clean signal.
Infineon's XENSIV™ MEMS microphone is a SDM MEMS sensor that improves speech recognition accuracy, reduces noise, enables long-distance voice capture, enables contextual understanding and enables multimodal interaction.
It has a high signal-to-noise ratio (SNR) while consuming low power. A high SNR means that it can capture a lot of desired signals clearly. It has IP57-rated waterproof and dustproof characteristics, making it environmentally advantageous. Its low-noise characteristics also make it suitable for applications such as transparent mode and active noise cancellation.
Wide dynamic range allows distortion-free reception from low noise levels to loud sounds. Clock input allows changing various power modes to optimize power consumption. All products are calibrated and tested to key specifications and are easy to distribute with the highest quality standards.
In Infineon's MEMS microphone portfolio, the new product 'IM70D122' is an SDM MEMS product. In tests, the recognition rate of the existing product dropped from about 98% to about 24% when the sound became smaller and the distance increased from 2m to 16m, but under the same conditions, the IM70D122 showed about 78% recognition accuracy even at a distance of 39m. It also showed a significantly higher recognition rate than the existing product in noise tests.
Infineon claims that “with a 75 dB SNR, the microphone captures 40 percent more audio than standard microphones, such as those used in commercial voice assistants.”
Kim Sung-tae, Vice President, said, “Based on high performance and high SNR, Infineon has a market share of over 50% in the silicon microphone business,” and “It is ranked within the top 3 in the global MEMS sensor market overall.” He also said, “Infineon has been developing groundbreaking MEMS microphone products based on 15 years of know-how in all stages from wafer to finished product production.”