5. What is Smart Speaker? (cont.)
"Alexa, what's the
weather?"
"Alexa, show my
calendar."
"Alexa, show me my
timers."
"Alexa, play some
music."
"Alexa, read my
notifications"
"Alexa, what's in the
news?"
"Alexa, find me a nearby
pizza restaurant"
"Alexa, ask Uber to
request a ride"
Voice Portal
12. ECM Microphone
A type of electrostatic capacitor-based microphone which eliminates the need for
a polarizing power supply by using a permanently charged material.
Built-in a FET as the amplifier.
Condenser diaphragm and plastic chassis are heat sensitive and should be careful
while soldering.
ECM Mic Circuit
ECM Mic Appearance ECM Mic Structure
13. MEMS Microphone
3 types of MEMS microphone technology
Piezoelectric (壓電式) (low sensitivity, high system noise)
Piezoresistive (壓組式) (low sensitivity, high system noise)
Capacitive (high sensitivity, low power consumption and low system noise)
Capacitive type of MEMS is the main stream in the market.
14. ECM Mic. vs MEMS Mic.
MEMS mics could offer the same SNR with much smaller size compare to traditional ECM mics.
MEMS mics also provide much better consistent response to sound across all operating temperatures.
Reference of Eetimes: http://www.eetimes.com/document.asp?doc_id=1280170
23. Beamforming Microphone Design (cont.)
• Beamformer is the Spatial Filtering.
• Base on these two idea
• Narrow band signal
• Far field plane wave
• Selectively amplify a sound source
at a particular location
• Take advantage of sound
propagation through space
• Use Delay Sum Beamforming
24. Beamforming Microphone Design (cont.)
• Endfire Microphone array
• Algorithm: sum of signal in front microphone
and inverted delay rear microphone.
• Distance: microphones shall be match to
sample rate for correct delay of sampling
Distance: Sound speed * sample time * n of sample
• Pattern: cardioid :180o : no Signal
( frequency < Aliasing frequency )
34300*1/(48000)*3=2.14
(cm)
A
B
SUM
28. Agenda
• What is Smart Speaker
• MEMS Microphone Array Technologies
• Cloud Voice Service
29. Cloud Voice Service
• Speech Recognition service(Speech to Text)
• NLP service (Natural Language Processing)
• TTS service (Text to Speech)
30. Cloud Voice Service (cont.)
Voice
Text
Intent
Feedback Activity
Device
Cloud
31. Cloud Voice Service (cont.)
• Amazon Alexa Voice Service
• Google Voice Service
• Baidu DuerOS
• Microsoft Voice Services
• IBM Watson Voice Service
• Nuance Communications
• Internal of Amazon AVS
32. Amazon Alexa Voice Service
• Alexa Voice Service(AVS)
• voice recognition service
• natural language understanding
service
• For voice-enable connected device
• Alexa Skills Kit
• API for voice application
• Include the ability to play music,
answer general questions, set an
alarm or timer, and more.
34. Amazon Alexa Voice Service (cont.)
• Application Case
Amazon Echo Vehicle Connectivity
( BMW / Ford / Hyundai )
35. Google Voice Service
• Actions on Google
• Design VUI
• work with the Google Assistant
• Support Google Home
• Support DiaglogFlow
• Support Firebase
• Google Speech Recognition
• Google TTS
• Google Natural Language API
40. Microsoft Voice Service
• Cognitive API with Microsoft Azure
• Speech API
• Speech to Text
• Text to Speech
• Speaker Recognition
• Speech Translation
• Language API
• Text Analytics
• Translator Text
• Bing Spell Check
• Language Understanding
• Content Moderator
42. IBM Watson Voice Service
• Voice Agent with Watson
• Improve telephone-based customer
service
• based on IBM Voice Gateway
• Support Service Orchestration Engine
• Speech to Text API
• Text to Speech API
• Natural Language Classifier
• Interpret and classify natural
language with confidence.
43. IBM Watson Voice Service (cont.)
• Application case
IBM Watson-powered
driverless electric bus
Softbank Pepper robot
44. Nuance Communications
• The company provide the software voice technology
used for Samsung’s S Voice and Apple’s Siri.
• Voice Search
• Intelligent Personal Assistance
• Knowledge Navigator
• Natural Language User Interface
• Dragon Software Developer Kit
• Dragon Mobile SDK
• PC Recognition Software