simon Open-Source Speech Recognition Developed by the non profit - - PowerPoint PPT Presentation

simon
SMART_READER_LITE
LIVE PREVIEW

simon Open-Source Speech Recognition Developed by the non profit - - PowerPoint PPT Presentation

simon Open-Source Speech Recognition Developed by the non profit organization Simon Listens in cooperation with Cyber-Byte IT services Introducing: David 17 years old Hobbies: Music TV Friends Girls Page 2 of 13


slide-1
SLIDE 1

simon

Developed by the non profit organization Simon Listens in cooperation with Cyber-Byte IT services

Open-Source Speech Recognition

slide-2
SLIDE 2

Page 2 of 13

Introducing: David

  • 17 years old
  • Hobbies:
  • Music
  • TV
  • Friends
  • Girls
slide-3
SLIDE 3

Page 3 of 13

Introducing: David

  • 17 years old
  • Hobbies:
  • Music
  • TV
  • Friends
  • Girls
slide-4
SLIDE 4

Page 4 of 13

The Challenge

  • According to EU: 6.4 % suffer from disabilities affecting

their arms and hands

  • 500 Mio * 6.4 % = ~ 32m
  • Reference: Finland + Norway + Sweden = ~ 20m
  • Same study: 0.4 % suffer from speech impairment
  • ~ 2m

Source: eurostat, [hlth_db_emtyag]

slide-5
SLIDE 5

Page 5 of 13

simon: The Idea

  • Speech recognition system
  • Extremely flexible
  • Adapts to the user
  • Even speech impairments are no issue
  • Community: Potential
  • Free software
slide-6
SLIDE 6

Page 6 of 13

simon: The Implementation

  • Based on
  • Julius
  • HTK
  • KDE4
  • C++
  • Supports
  • GNU/Linux
  • Windows
slide-7
SLIDE 7

Page 7 of 13

Speech recognition

Speech model Language model Acoustic model Recognizer

  • Vocabulary
  • Grammar
  • Sounds
slide-8
SLIDE 8

Page 8 of 13

Speech recognition

1.Recording utterance

slide-9
SLIDE 9

Page 9 of 13

Speech recognition

1.Recording utterance 2.Parameterization

slide-10
SLIDE 10

Page 10 of 13

Speech recognition

1.Recording utterance 2.Parameterization 3.Statistic comparison with acoustic model 4.Comparison with vocabulary and grammar 5.Output most probable solution

slide-11
SLIDE 11

Page 11 of 13

simon: Flexibility

Speech model Language model Acoustic model Recognizer

  • Vocabulary
  • Grammar
  • Sounds
slide-12
SLIDE 12

Page 12 of 13

simon: Flexibility

  • Language model: Scenarios
  • „Use-case-packages“

– Vocabulary – Grammar – Commands – Trainings texts

  • Acoustic model: Base models
  • Static
  • Adapted
slide-13
SLIDE 13

Page 13 of 13

simon 0.3 alpha 3

  • Demonstration of simon 0.3:
  • Firefox
  • Amarok
  • XBMC
slide-14
SLIDE 14

Page 14 of 13

Get involved

  • Donate speech: http://www.voxforge.org
  • Create and share scenarios
  • Software development
  • Translation and documentation
  • Contact: support@simon-listens.org
  • Me personally: grasch@simon-listens.org
slide-15
SLIDE 15

Questions?

slide-16
SLIDE 16

Thanks!