Slide 1

Report
Forschungszentrum Telekommunikation Wien
[Telecommunications Research Center Vienna]
Interfaces between Speech and Non-Speech
Audio Technology
Michael Pucher (FTW Vienna, ICSI Berkeley)
Contents
 Text-to-Speech Synthesis (TTS)
 Automatic Speech Recognition (ASR, STT)
 Dialog Systems
 Multimodal Mobile Applications
 Resources
© ftw. 2005
Auditory representations
Non-linguistic
Sound signals
Music
Perspectival, spatial cues
Paralinguistic
Speaker characteristics
Affective states and attitudes
Linguistic
Pragmatics and discourse
Structural prosodic elements
ASR
© ftw. 2005
Lexical semantics and syntax
Dialog Systems
TTS
TTS Examples

16kHz natural voice

16kHz unit selection synthesis (server-based)

8kHz diphone-based synthesis with lexicon
(embedded or distributed)

8kHz diphone-based synthesis without lexicon
(embedded)

Application specific lexicon
- Gerald R. Ford  tSE-r6ld a:R fo:rd
© ftw. 2005
TTS Evaluation
Pronuniciation
Articulation
Overall
Voice
Listening
Comprehension
SmartPPCNoLex
SmartPPCLex
100
95% CI Comprehension of words
Natural
Source
MobileSPNoLex
MobileSPLex
CTTS
CorpPPC PCM
80
60
40
20
CorpPPC GSM
0
PC
PC
tP
ar
Sm
tP
ar
x
Le
No
x
Le
SourceLabel
Sm
l
ra
tu
x
Na
Le
No
SP
ile
ob
M
x
Le
SP
ile
ob
M
M
PC
C
SM
G
© ftw. 2005
5
PP
rp
95% CI
4
C
3
R
2
AM
1
Co
PP
rp
Co
C
PP
rp
Co
CorpPPC AMR
TTS and Non-Speech Audio
TTS
FEATURE STATUS
Comprehensible TTS
Low worderror-rate
Solved
Diphone
based TTS
Natural TTS
Single style
prosody
Solved
Unit selection
Expressive TTS
© ftw. 2005
Various
prosodic
styles
Not solved
?
COMBINATION
WITH NONSPEECH AUDIO
TTS provides lexical
information
Add structural prosodic
elements
TTS provides structural
prosodic elements
Add affective states and
attitudes
Add pragmatic
information, dialog
turns
Limited Expressiveness of Speech 1

Limited expressiveness of Expressive TTS = Limited
expressiveness of speech

Limited expressiveness of speech because of unlimited
expressiveness1 of speech
- Because everything is expressible in language, the
messages are less useful for certain purposes (too
complex)
- Simpler, less expressive codes (sounds, icons) may be
used in context and lead to shorter messages

Disadvantages of speech
- Seriality
- Non-universality
© ftw. 2005
Types of ASR and Applications
 Isolated word recogniton
Command & control
 Large vocabulary Speech
recognition
Broadcast news transcription
 Conversational Speech
recognition
Meeting transcription
 Speech Recognition in
noisy environments
Car navigation
Speaker dependent or speaker independent
© ftw. 2005
Other Related Technologies
 Speech
- Speaker verification
 NLP
- Dialog act detection
- Topic detection
© ftw. 2005
Music Information Retrieval (MIR)
 Query By Humming (Fraunhofer)
- Non-speech sound as an input pattern to search for
other non-speech sounds
- http://www.musicline.de/de/melodiesuche/input
 Performer Style Identification
 Melody and Rhythm Extraction
 Music Similarity
 Genre Classification
© ftw. 2005
Dialog Systems - ASR

3 Types of Recognition in stateof-the-art Dialog Systems
- Isolated word
<rule id="exit">
<one-of>
<item>exit</item>
<item>quit</item>
</one-of>
</rule>
- Recognition grammar
- Statistical Language Model
(SLM) + grammar for more
robustness
„um ah to san francisco from new york“
1. Apply SLM
2. Apply grammar on results of SLM
© ftw. 2005
<rule id=„commands">
<item repeat="0-1">
move
</item>
<one-of>
<item>forward</item>
<item>backward</item>
</one-of>
</rule>
Dialog Systems – TTS and Audio
 Loquendo TTS Mixer
- Play and mix TTS and
audio files
- Fadein, fadeout
- Pause and resume
- Record
© ftw. 2005
Paolo Massimino : Loquendo
S.p.A.
From Marked Text to Mixed
Speech and Sound
Dialog Management 1
 Usages of non-speech audio
-
Replace prompts
Indicate dialog turns and dialog states
Indicate menu structure (3Daudio)
Create listen & feel of the application
System response time
 Questions
- Bargein, Streaming and Standardization
© ftw. 2005
Dialog Management 2
 A good bad example
- Uses only speech
- Audio enhancement for
transitions
- Audio enhancement for
states
© ftw. 2005
Bob Cooper : Avaya Corporation
A Case Study on the planned and
actual Use of Auditory Feedback
and Audio Cues in the Realization
of a Personal Virtual Assistant
Dialog Managment 3
 VoiceXML Version 2.0
- W3C (Word Wide Web Consortium) standard for
voice dialog design
- Form filling paradigm similar to web forms
 Synthesis Markup Language (SSML) Version
1.0
<prosody contour="(0%,+20Hz) (10%,+30%) (40%,+10Hz)">
good morning
</prosody>
<voice gender="female">
Any female voice here.
<voice age="6"> A female child voice here. </voice>
</voice>
© ftw. 2005
Limited Expressiveness of Speech 2
 Limited expressiveness of human-machine
voice dialog compared to a natural dialog
- Natural dialog is probable multimodal
- Role of non-speech sound in human communication
© ftw. 2005
The Importance of Multimodality
for Mobile Applications
 Multimodal communication is perceived as
natural
 Disadvantages of unimodal interfaces for
mobile devices
- Small displays
- No comfortable alphanumeric keyboards
- Visual access to the display is not always possible
 Disadvantages cannot be overcome by
increasing processor and memory capabilities
© ftw. 2005
Multimodal Dialog Managment

Speech Application
Language Tags
(http://www.saltforum.org)


Possible combination
with non-speech
audio at all states
and transitions
Similar to (unimodal)
dialog systems
Minhua Ma : University of Ulster
Paul Mc Kevitt : University of
Ulster
Lexical Semantics and Auditory
Display in Virtual Storytelling
© ftw. 2005
Asymmetric Multimodality

For Multiparty applications
- Users select preferred modalities (e.g. speech, visual, music?)
- System is able to translate content from one modality to another

MONA – Mobile Multimodal Next Generation Applications
- Multiuser quiz application
Input
Output Preference=Speech
Output Preference=Visual
Speech
Speech
Speech-To-Text
Text
Text-To-Speech
Text
© ftw. 2005
Resources
 TTS
-
Festival 2.0, to build unit selection voices
Festival Lite, for embedded TTS
FreeTTS, a Java speech synthesizer
The Mbrola project, many synthetic voices available
 ASR
- Sphinx
- Htk
 Multimodal Systems
- SALT implementations
© ftw. 2005
Thank you for your attention
Contact:
[email protected]
http://userver.ftw.at/~pucher
© ftw. 2005

similar documents