Smart speakers promote MEMS microphone market growth

Voice assistants are no longer new, but the smart speakers equipped with AI voice assistants have become the new smart home products that domestic and foreign technology giants are vying to launch, which will undoubtedly drive the growth of the microphone market, the indispensable sensor in speech recognition. But growth does not include all types of microphones. MEMS microphones will benefit most from the voice interaction requirements of smart speakers, smart headphones, robots and other applications with miniaturization, good consistency and low power consumption, but the application of microphone arrays will also bring technical challenges.

The MEMS microphone market continues to grow.

The data shows that the MEMS microphone market in 2016 was $993 million, close to the $1 billion mark, and the $700 million electret condenser microphone (ECM) market, the entire microphone market is about $1.8 billion. Market research firm Yole Développement predicts that MEMS microphone shipments will maintain a high growth rate in the next five years with the support of smart voice assistants and in-vehicle applications, with a compound annual growth rate of 11.3% and annual shipments by 2022. It is expected to exceed 8 billion. ECM microphone shipments are slowly shrinking, with shipments of only about 3 billion by 2022.

Figure 1: ECM/MEMS microphone market size (Source: Yole Développement)

After Amazon launched the smart speaker Echo with voice assistant Alexa in 2014, smart speakers equipped with virtual voice assistants such as Google Assitant, Cortana and Siri appeared. Major IT giants are launching smart speaker products, indicating that the consumer MEMS microphone array market is poised for growth. Obviously, smart home products represented by smart speakers will boost the growth of the MEMS microphone market. Dong Wei, senior market engineer of South China Division of STMicroelectronics MEMS Products Division, also said: "With the increasing penetration of smart products in smart homes, the traditional way of interaction can no longer meet the needs of users, and voice as an important means of interaction More and more popular. As an important interactive window, MEMS microphones have achieved great success in the mobile phone and notebook markets, and smart speakers will become the third wave of MEMS microphone applications."

Figure 2: Market trends in wearable devices (including Hearables)

Far-field speech recognition faces challenges such as high SNR and AOP

The future of smart speakers will play a role in promoting the MEMS microphone market. In turn, smart speakers will also bring multiple challenges to MEMS microphones. As a smart home product, smart speakers are different from the near-field speech recognition of mobile phone voice assistants and headphone voice assistants. Smart speakers are often equipped with MEMS microphone arrays, and techniques such as beamforming (Beam-forming) are used for noise reduction speech processing to achieve far field. Speech recognition, no need to operate with both hands. For far-field speech pickup applications, combined with the requirements of algorithm applications, MEMS microphones with high signal-to-noise ratios are required. The sensitivity of the microphone array application to the microphone is consistent, the signal-to-noise ratio consistency and phase consistency requirements are very high.

In addition to the well-known signal-to-noise ratio (SNR), the Acoustic Overload Point (AOP) is also an important quality indicator for MEMS microphones. High SNR enables devices to achieve long-distance radio reception. AOP is a performance metric for evaluating MEMS microphones at high sound pressure levels. SNR and AOP are important for an increasingly diverse range of intelligent voice devices. Reports show that the market share of high signal-to-noise ratio MEMS microphones above 64dB has reached more than 50%. Many manufacturers have also increased their previous 120dB to more than 130dB.

In the far-field speech recognition scene, the receiving environment of the microphone is inevitably interfered by various noises, mixing and even reflections, resulting in a significant decrease in the accuracy of the MEMS microphone for the acquisition and extraction of sound signals, thus affecting the accuracy of speech recognition. . At this point the device needs to identify which is a useful sound through the MEMS microphone array. In addition, if the MEMS microphone array is always in the radio state, it will consume a lot of power, which is very unsuitable for smart mobile devices. Therefore, how to make MEMS microphones more intelligent, and to achieve low power consumption while ensuring user experience is also a challenge. In addition, with the development of smart devices in the direction of miniaturization and diversification, EMI interference and miniaturization of MEMS microphones still need to be improved. Finally, the cost of MEMS microphones still has no advantage over traditional ECMs.

In addition to technical indicators, it is still more important to consider the design structure and production method algorithms.

Dong Wei said that MEMS microphone arrays have high performance requirements for microphones. However, in addition to the characteristics of the device itself, the microphone algorithm is the key to whether the product can be accepted. For example, for the filtering of environmental noise, MEMS microphones need to be integrated with technologies such as deep neural networks (DNN) and machine learning. It is difficult to judge how many microphones will be more suitable for the microphone array. Various chip companies and algorithm companies in the industry have launched their own array solutions. Different hardware and algorithm combinations have also launched many excellent smart speaker products. Therefore, the choice of the number of microphones depends on the design, positioning, and specific application scenarios of the solution provider. Implementations involving functions such as beamforming and keyword wake-up are more dependent on speech algorithm companies.

MEMS microphones will merge DSP into intelligent direction in the future

The application trend of MEMS microphones is always on, just like the ears, to keep them open. The significance of the fusion DSP is that the simple algorithm microphone handles itself and meets the requirements of low power consumption. ST and DSP Group introduce MEMS microphones with integrated keyword wake-up to help customers solve the low-power microphone low-power challenge of integrating ST's low-power MEMS microphones, DSP Group's in a tiny system package (SiP) The ultra-low-power speech processing chip and Sensory's speech recognition firmware use ST's advanced packaging technology to achieve a very good lightweight package, extremely long battery life and advanced features.

With the decline in the cost of MEMS microphones, the selection of MEMS microphones for smart speaker microphone arrays is already a trend, and the MEMS microphone market is currently developing very well, and is being developed in many fields.

For more applications of MEMS microphones in the future, Dong Wei believes that the automotive market will be another application of MEMS microphone bursts. The car operating system is networked, providing a rich application, and also puts forward new requirements for the human-computer interaction interface of the car. As the most natural and safe human-car interaction mode, voice will dominate the interaction mode inside the car. He also said that the MEMS microphone market will continue to grow in the next few years, and new competitors will lead to more intense market competition.