How to Make Smart Homes Even Smarter?
The impact of voice-controlled digital assistants such as
Amazon Alexa, Google Assistant, Apple Siri, Microsoft Cortana and Samsung Bixby
cannot be overstated. Consumers are using voice commands for turning on the
lights, streaming music, asking questions and accessing third-party services in
record numbers. Interacting directly with smart home assistants has created a
new sales channel: shop by voice.
CTA projected consumers would buy 4.4 million Amazon Alexa
and Google Home devices during the 2017 holiday season, up 22 percent, in its
U.S. Consumer Technology Sales and Forecasts. The report pegged 2017 sales in
the smart home category at 27 million units (a 50 percent increase over 2016),
earning $3.3 billion (a 48 percent increase).
These smart assistants work pretty much the same way. There
is a trigger phrase that captures the system’s attention, such as Apple’s “Hey
Siri,” which lets users invoke Siri hands-free. This isn’t as simple as it
seems. A very small speech recognition engine — the Apple iPhone calls it the
“Always-on Processor” — has access to the microphone signal and runs all the
time, listening for those two words.
The “Hey Siri” detector uses a Deep Neural Network (DNN) to
convert the acoustic pattern of your voice into a probability distribution over
speech sounds. It then computes a confidence score to confirm you said “Hey
Siri.” If the score is high enough, Siri wakes up.
The audio input that follows in each digital assistant then
is transmitted to the cloud for processing via Wi-Fi. A smart device equipped
with Bluetooth first sends its audio file to a smartphone and then transmits
the file to the cloud using a cellular or Wi-Fi connection. Placing the main
voice recognition engine in the cloud efficiently eliminates the need for dedicated
resources in the device.
If they work the same way, how can suppliers make these smart
assistant devices smarter?
Involve More Developers
With the competition to build digital assistant products
intensifying, success will depend on how well they perform, but also the
ecosystem to which they connect. The ease that third-party developers can
integrate voice-control technology into their products is critical.
Intel released the Intel Speech Enabling Developer Kit for
Amazon Alexa that provides an audio frontend solution for voice control. It
includes algorithms for acoustic echo cancelation, noise reduction, beamforming
(a signal processing technique used for directional signal transmission or
reception) and a custom wake-word engine tuned to “Alexa.” It also comes with
dual Intel DSPs and an eight-microphone circular array.
Similarly, Cirrus Logic’s Voice Capture Development Kit for
Alexa Voice Service (AVS) is optimized for the development of Alexa-enabled
smart speakers, portable speakers and compact audio devices. It includes a
reference board that features Cirrus Logic’s CS47L24 smart codec, CS7250B
digital MEMS (microelectromechanical systems) microphones and “SoundClear”
algorithms for voice control, noise suppression and echo cancelation. The
turnkey voice capture reference design enhances “Alexa” wake word detection and
audio capture in real-world conditions, even from moderate distances and in
noisy environments.
Get a Clean Audio Signal
This increases the chances that the voice recognition engine
will understand what’s being asked of it. The application of MEMS technology
led to the development of small, high-performance microphones supporting user
interactions with “smart” devices. Polysilicon is the material often used as a
membrane for microphone devices. But this material is brittle and defects can
occur during manufacturing.
At the AVS International Symposium & Exhibition in
Tampa, FL, researchers from WPI-Advanced Institute for Materials Research/Micro
System Integration Center at Japan’s Tohoku University presented a potential
replacement material — amorphous metallic glass (cobalt tantalum boron) and it
shows promise for MEMS mics from a manufacturing standpoint and also from the
perspective of achieving high signal to noise ratios.
Provide Short Range, Very High-Speed Transfer of 4K/HDR
Video
The IEEE standard 802.11ad could play a role using the 60
GHz spectrum instead of the 5 GHz and 2.4 GHz used by most Wi-Fi connections.
It provides data throughput speeds of up to 7 Gbps (vs. 3.2 Gbps for the
current speed champion, 802.11ac Wave 2). Since it uses frequencies in the
millimeter range 802.11ad, its range is limited to about 30 feet. For high
volume data exchange such as 4K video transfers or streaming from your smart
device to your smart TV it should work well. But when longer ranges are needed,
the Wi-Fi 802.11ac version is a better choice to provide the best consumer
experience.
Comments
Post a Comment