Page 16 - EE Times Europe Magazine

Page 16 - EE Times Europe Magazine | June2020

P. 16

16 EE|Times EUROPE

The Growing Importance of Innovative MEMS Microphones

The FluSense solution captures
crowd-level non-speech body sounds
such as coughs in an unobtrusive and
passive manner, combining this data with
patient counts estimated using thermal
images taken in hospital waiting rooms.
Together, these elements provide key
predictive information on epidemiolog-
ical trends for a given demographic. The
FluSense platform processes low-cost
microphone array and thermal imaging
data at the edge using a Raspberry Pi
and a neural computing engine (the
Intel Movidius). None of the information
stored is personally identifiable.
The solution can run deep-learning–
based acoustic models and algorithms for
estimating crowd sizes based on thermal
imaging in real time. The system can
detect coughs with an accuracy of up to
87%. The developers now aim to validate
the model in non-clinical settings such
as restaurants, public transportation, and
classrooms. High-performance micro-
Figure 2: High-performance microphones are vital in a wide range of voice-control features phones could increase detection rates
and applications. further under such conditions.

be distinguished from the real thing. Microphones are vital MICROPHONE PERFORMANCE
for providing the high-quality input that all of the applications Taking a closer look at microphone performance, there are several
mentioned here need in order to deliver an outstanding user factors to take into consideration: What are high-performance micro-
experience and excellent audio quality (Figure 2). MEMS micro- phones? Which microphone parameters are important and which ones
phones with best-in-class audio quality specifications can deliver are relevant for different use cases? Every microphone is capable of
the required performance. recording a range of sound pressure levels (SPLs); this is known as the
dynamic range of a microphone. The upper limit of the dynamic range
HEALTH TRACKING is defined as the acoustic overload point (AOP), while the lower limit
Monitoring vital signs with optical sensors is an established tech- is defined by the microphone’s self-noise. A microphone can pick up
nology. In some instances, however, space constraints limit the use only signals with an SPL above its self-noise. This lower threshold is
of existing sensors. One way to save space here is to combine several known as the “noise floor” of a microphone, and it defines the signal-
sensors — creating, for example, a microphone that can also monitor to-noise ratio (SNR). A microphone cannot record any sound below its
body temperature. Health tracking is a growing market for mobile noise floor. A microphone with a noise floor of 30 dB SPL, for example,
devices. Tracking applications will become more appealing as users cannot capture a human whisper at 25 dB SPL amplitude. Therefore,
become more health-conscious. High-performance microphones with microphones with a higher SNR (i.e., a lower noise floor) are well suited
ANC can be combined with body temperature sensors to provide a to picking up low-amplitude audio signals.
useful solution for tracking health and detecting a high temperature. SNR and AOP are important parameters for assessing individual
A TWS headset with the ability to track the wearer’s temperature and microphone performance. However, most devices today use several
issue a warning at the onset of a fever provides peace of mind; users microphones in an array. Smartphones, for example, have three or four
can rest assured that their health is being monitored. Detecting fevers microphones, while TWS incorporates up to six microphones (three per
at an early stage means that treatment can be started promptly. Having earbud). The numbers are even higher in conference systems. In short,
a record of a patient’s body temperature can also help with diagno- microphone arrays can contain anywhere from two to 32 microphones.
sis and treatment. Infineon has developed an ASIC that features an The performance of a microphone array depends on a combination of
I²C temperature sensor. Combining this with the MEMS produces a individual microphone characteristics and combined array characteristics.
high-performance microphone with temperature-sensing functionality The individual characteristics include the AOP and SNR, while the com-
— a solution that saves space by combining the two sensors. bined array characteristics include factors such as sensitivity matching
(whether all mics have almost the same sensitivity) and phase matching
INNOVATIVE APPLICATION EXAMPLE: FLUSENSE (whether all mics have a similar phase response). These features combine
An innovative device invented in the U.S. at the University of to improve overall audio capture and to ensure that the array produces
Massachusetts Amherst demonstrates the possibilities of using micro- higher-quality sound and has lower self-noise levels — comparable in
phones in medical tracking (https://www.umass.edu/gateway/feature/ many ways to watching a movie in normal resolution or full HD.
flusense). Designed to analyze coughing and detect crowd sizes, the Flu-
Sense device is made up of three components: a camera, a microphone, THE IMPORTANCE OF HIGH-QUALITY AUDIO RAW DATA
and a computer (Figure 3). The challenge for the developers was to Virtual assistants like Siri and Alexa are voice user interfaces (VUIs)
find an early way to predict and monitor the outbreak of influenza-like present in smart speakers. VUIs comprise an array of microphones
illnesses — characterized by key symptoms such as fever and coughing that are used to capture higher-quality raw audio data as input for the
— as feeding lab-confirmed cases into epidemic models takes time. application processor. The raw data input from high-SNR

11 12 13 14 15 16 17 18 19 20 21