Exclusive interview with Yole analyst: What other developments are there for MEMS microphones today?

This article is translated from Ferrelectronics

When it comes to sensor technology for sound and voice control, innovation is happening on all fronts, from the audio devices themselves to the software and algorithms and MEMS microphones that are making leaps and bounds, as well as AI-driven data analytics.

Ferrelectronics recently spoke with Dimitrios Damianos, Technology & Market Analyst and Business Developer for Custom Projects at Yole Développement, about this trend and how it will ultimately impact future applications such as acoustic event detection, speech recognition and context awareness, and even emotion/empathy perception using voice (something Amazon and Apple already have patents on).

As voice control evolves, design engineers will need to consider the unique requirements and issues surrounding this technology.

FE: You said the next innovation in MEMS and sensors will be in audio, sound and voice control. But isn’t it already here? What will be different?

Dimitrios Damianos (DD): Yes. MEMS microphones have been in use since they were included in the first Motorola Razr phone in 2003. Since then, they have come a long way: they have replaced traditional electret condenser microphones (ECMs), offering better performance, sensitivity and lower cost, with billions of them shipped each year.

Voice control as a human machine interface (HMI) has been making waves since a few years ago. There are now many devices with voice/virtual personal assistant (VPA) capabilities, such as smartphones, smart watches and the latest smart speakers and cars. Innovation in the audio field is actually happening on a larger and more comprehensive scale. MEMS microphones need to have best-in-class performance (sensitivity) as well as low power consumption, because these devices are always on. In addition, the sound must be captured at a high quality for efficient processing and high-quality rendering. You know the concept in computer science: garbage in, garbage out, meaning that if you want to get some context from the data, it must be of at least a certain quality. This is why MEMS microphones continue to improve.

At the system level, you also need to consider the entire audio chain from the device to the audio codec, audio software and algorithms (noise cancellation, beamforming, etc.) and digital signal processors (DSPs), including audio amplifiers and speakers. So innovation is happening on all fronts, in optimizing all of these variables, especially in using artificial intelligence to analyze data, which ultimately affects acoustic event detection, speech recognition, and contextual awareness.

FE: Which technological advances will accelerate adoption and open up new applications? What role will the edge play?

DD: In addition to some new technologies in MEMS microphones (piezoelectric, optical) and MEMS microspeakers, the use of voice as a human-computer interaction is also accelerating, mainly because of advances in AI computing. Now, most computing is done in the cloud, where models are trained and inference is also done in the cloud. This allows the data to be analyzed, which has huge value.

However, the data in this case is often in the hands of the global GAFAMs (Google, Apple, Facebook, Amazon, and Microsoft), which sometimes raises privacy concerns. We are seeing a shift towards training in the cloud and inference at the edge to reduce latency issues. Eventually, both training and inference will be done at the edge to address privacy concerns. In this case, everything is done locally on the device and no data is sent to the cloud. All training is done in a small form factor, close to the device (at the edge), and at low enough power that machine algorithms are being re-architected and new computing architectures are being investigated, such as neural networks.

FE: What about costs? Will they need to come down significantly to achieve the market size you predict?

DD: Cost is not an issue. MEMS microphones are produced in billions per year at very low prices, typically between $0.1-0.3, depending on the manufacturer and order size. The specific market size we predict for MEMS microphones will be achieved in two ways: the increasing attachment rate of MEMS microphones in various consumer devices, and the growth of terminal system capacity. The adoption of voice as a human-machine interface will depend on the cost, performance and functionality of the entire system, including MEMS microphones, speakers, audio processors or computing chips, etc.

FE: What are the future applications?

DD: We are heading towards the Internet of Voice (IoV) era as voice is increasingly adopted as an interface for all kinds of everyday devices. So really, the future is here, and it will continue to get better as hardware and software continue to improve, providing users with a more inclusive and personalized experience. In this way, more and more people will use voice assistants in their daily lives as various latency, power consumption, computation and privacy issues begin to become clear.

One future application will be emotion/empathy awareness using voice (and sometimes other sensor) data, where your tone of voice can infer your emotions. Amazon and Apple already have patents for this. Amazon also has a new wearable device, the Amazon Halo wristband, that analyzes your tone of voice.

FE: When will we see hearing aids based on MEMS microphones, especially given the growing elderly population?

DD: Each hearing aid manufacturer has different requirements and wants to develop a specific microphone, which makes this a high-demand market (needing high-quality microphones), which in turn leads to high selling prices for high-quality microphones. Given these limitations, it does not seem to be a very profitable market for various microphone manufacturers.

However, MEMS microphones are increasingly being used in hearing aids, although traditional ECMs are still the most common microphones used in this application. The small size of MEMS microphones has always been their main advantage, but now they have similar or better performance than ECMs in terms of noise performance, power consumption, stability, and hearing aid reproducibility. MEMS microphones enable new features such as directional hearing, speech recognition, and amplification to become more precise, ultimately resulting in better hearing aids.