How it works – Voice Recognition How it works – Voice Synthesis

movedearAI and Robotics

Nov 17, 2013 (3 years and 4 months ago)

78 views

How it works – Voice Recognition
“Hello! Will you be
at home for dinner?”
hello will you be at home for dinner
“hello”
How it works – Voice Synthesis
➊ The mobile device receives a text
message

➋ The system retrieves the message.
Statistical probability algorithms
determine the correct usage and
pronunciation for each word in
context with those around it

Hello!
Will you
be at
home for
dinner?
home
dinner
dinner
həˈləʊ wɪl juː biː æt həʊm fɔː(ɹ) ˈdɪnə

ˈləʊ
wɪl
juː
biː
æt
həʊm
fɔː(ɹ)
ˈdɪ


➌ The individual sounds needed to synthesize
each word are selected from a database

➍ The system pieces together the individual
sounds selected into complete words

➎ It then says the word or sentence using the
built-in voice engine

ˈdɪ
“Play artist Coldplay!”
“artist”
tɪst
➊ User speaks command into microphone

➋ System converts sound input
into digital signal

➌ The signal is processed into a series of data
fragments called feature vectors, each the same
length

➍ The system searches for the most probable
match for the sequence of feature vectors using
statistical sound representations called
acoustic models, enabling the system to
recognise words

➎ Each word is examined in context with those
around it and statistical probability algorithms used
to determine the intended command

➏ The appropriate response for the command is
triggered
r
tist
ur
ar
artist
Coldplay
coldplay
play
artist