Development and Implementation of Telecom System for Hearing

1
Development and Implementation of Telecom System
for Hearing Speaking Impaired People
(Previously Dumb and Deaf Telecommunication
System)
Project Report
2
Presentation Agenda

• Overview
• System Software Design
• Implementation

3
Overview
4
TELKOMs BUSINESS OVERVIEW
GSM 800, 1800 Cellular
Network Interconenction
Data Multimedia
Fixed Wireless CDMA
Fixed Wireline PSTN, WLL,
5
TELKOM RD Center

• Well known as TELKOMRisTI since 1997 2003,
  after the restructurization change into pure RD
  CENTER
• Core business is product development in VAS ,
  Data Multimedia , High Speed network , Software
  Application and Product Management
• Some of projects are dedicated for social matter
  establishing Good Corporate Citizenship such as
  e-government, rural information community.

6
Different Ability (Difable) Community In Indonesia

• Almost 6-10 population is difable people
• According to Indonesias Constitution, all
  difable person are under government care, concern
  and protection . The implementation done by some
  foundation controlled by Social Department .
• Mostly of difable person are not wealthy enough
  to purchase Text Capable -Phone or CPE such as
  cellphone.
• Most of them live in home, suburban or even rural
  area.
• Some foundation related to difable people UPI
  (Universitas Pendidikan Indonesia), GERKATIN
  (Indonesian Movement for Hearing Impaired), YPAC
  ( Foundation for Difable Child), INVERSO BAGLIVO
  ( International Foundation for Hearing Impairment
  Community in Indonesia)

7
Type of Disabilities
81,90 Hearing Speaking Impaired are from
Congenital
8
Perspective to Difable Person

• 1945 Basic Constitution Article All citizens
  are entitle of job and decent livelihood
• There is no discrimination between normal people
  and difable people in their rights to have
  telecommunication services. All are customer ,
  having human rights in communication and
  Indonesian citizen.
• Creating some services that bring communication
  capability to all difable person with acceptable
  technology in economic and ergonomic aspect is
  prefered to becoming one of USO services.

9
Some telecommunication services already in place
for difable person

• Text Phone text is delivered through PSTN in
  real time communication.
• Mobile Phone not real time , difable person
  must bought CPE
• SMS Chatboard must be connected to CPE
• TELKOMSMS SMS through PSTN Wireline
• Interaksi2 - by Microsoft
• Some of them need purchasing of specific device,
  and communication in text mode only. TELKOMRD
  Center create a technology that can interchange
  text and voice communication

10
Research Objective

• Establishing an infrastructure to enhance POTS
  system capability to help difable people (
  hearing and speaking impaired ) in communicating
  with normal people through telco network.
• Serving most difable people in acceptable
  communication services.
• Become a triggering concept for other
  telecommunication cellular operator to also
  build similar service to help difable person.
• Implementing text to Speech and Automatic Speech
  Recognition concept and technology in
  telecommunication industry

11
Necessity Urgency

• Difable people have the same right as normal
  people in having access to telecommunication
  services.
• Facilities to help people with disabilities are
  rare in Indonesia.
• Many difable people have other potential
  capabilities that can help building Indonesia,
  therefore access to telco services will help them
  to maximize their capabilities

12
Participants

• TELKOMRisTI Paulus Hutagaol (Chief Researcher),
  Eka Kelana, Rudy Harinugroho, and Romzi Imron
• Responsible for DDTS development
• Advanced Telecommunications Research (ATR) Japan
  Satoshi Nakamura, PHd and Sakriani Sakti
• Support Automatic Speech Recognition (ASR)
  Technology
• Indonesia University of Education (UPI) Juang
  Sunanto and Permanarian Somad
• System Testing
• Providing advice in social aspect issue

13
System Software Design
14
Conceptual Design
Normal people
Text-To-Speech (TTS)
Automatic Speech Recognition (ASR)
Hearing Speaking Impaired People
15
Communication flow from hearing speaking
impaired people to normal people
Typed text are HELLO , IM TRYING TO
COMMUNICATE
HELLO , IM TRYING TO COMMUNICATE
TTS
TTS
16
Communication flow from normal people to hearing
speaking impaired people
HELLO , IM TRYING TO COMMUNICATE
Typed text are HELLO , IM TRYING TO
COMMUNICATE
ASR
17
Critical Design

• Automatic Speech Recognition Concept, Engine,
  Corpus Database for Indonesian language
• Call Flow (Signalling) Traffic Engineering and
  ASR and TTS integration
• Telco line data comm link condition (frequency,
  bandwidth)
• Operating System Configuration (I/O Latency,
  memory management, supporting library)

18
System Boundary

• The system should be implemented as a network
  element.
• The system must support available CPE and
  computing device.

19
System Components
20
Development Process (ASR)

• Collecting speeches/utterances from 110 speakers
  covering all dialects of major Indonesian tribes,
  resulting in more than 48.000 speech data (digit
  and simple words/sentences).
• Processing speech data to meet ATRs standard
• Validating, Training and Testing speech engine
  using the speech data

21
Development Process (System)

• Develop the VUI using TAPI library of windows OS
• Develop the TUI using available supporting
  library for real time text communication (similar
  to chat engine)
• Develop the bridge using TAPI library of windows
  OS
• Develop the TTS Controller using the TTS
  supporting library
• Develop the ASR Controller using the ASR
  supporting library, linux glibc and networking
  library (socket API)

22
Implementation
23
Current Implementation

• Lab Scale, fully functional
• Text-to-speech translation only provides male
  voice
• ASR can recognize digit very well (99) and
  limited set of words/sentences arranged in cases
  (90)

24
Trial System Activity

• Conducted in 20 April 2004
• System was tried by 2 hearing and speaking
  impaired person from Sekolah Luar Biasa Cicendo.
• We invited researchers from ATR japan and BPPT
  Jakarta, and academic from ITB Bandung and
  STTTelkom Bandung
• Some notes To enhance the server quality and to
  add more recognizable speech database

25
Next Development

• Indonesian LVCSR (Large Vocabulary Continuous
  Speech Recognition)
• TTS Female voice
• ASR Controller enhancement Speech End Point
  Detection

26
Commercialization

• Language Translation
• Digit Collecting (PIN, Call Center, etc)
• Location Based Information System
• Game via phone
• Voice to SMS/SMS to voice
• Distance Learning
• Web/HTML reading
• Etc,

27
(No Transcript)
28
The End

• Thank You