|
Network
of Excellence |
RecoPhone |
The
Internet Portal for Speach Recognition Technology |
The
IP consists of 10 partners from
seven different European countries. The wide geographic distribution of the
partners over Europe provides a wide variety of experience. A good balance
between pre-accession countries and members of the European Union will
contribute to the integration of the pre-accession countries and the exchange of
scientific and technical expertise. The Project consists of 5
enterprises and 5 universities and research
institutes. The high level of participation from enterprises will give strong
impetus to the implementation of the results. The universities and research
institutes are all very highly rated institutions with a very strong scientific
research base and therefore well able to provide the scientific leadership
required to fulfill the aims of the IP. The partners are widely experienced in
different fields of speech recognition and information technologies and their
applications.
ICS
(Intelligent Computer Systems) is a SME in Germany develop,
manufacture and trade Intelligent Computer Systems for medical and industrial
applications, founded in 1991. The topic of the product development is the
method of real-time recognition of signals. The success in different fields of
applications are the motivation for further development of the new method. The
real-time recognition is realized by using the following methods of
SoftComputing: (1) Patented method of Virtual Sources, (2) Neural Networks, (3)
Fuzzy-logic, (4) Evolutionary algorithms, (5) Non Linear Dynamics, (6) Data Base
and Knowledge Management Systems.
Established
in 1991, Tilde is a leading Baltic IT company specializing in localization,
multilingual and Internet software. Tilde’s goal is to be the leading provider
of top-quality IT products for users in the Baltic countries, considering the
local language, culture and legislation specifics, as well as to provide the
best translation and localization services in the Baltics. Tilde’s experience
in language technologies has condensed in software for Latvian and Lithuanian
languages. Our software includes proofing tools, multilingual electronic
dictionaries, thesauruses, localized fonts and templates. In language
technologies the aim of Tilde is to provide language technologies for the
languages of the Baltic countries that would be equivalent to the support for
the major languages of the world. Tilde now expands work in new areas of NLP –
machine translation, information retrieval, speech synthesis and
recognition.
The
ABS GmbH was established in 1990 as an engineering enterprise. Currently we are
putting into practice our own and external innovative product ideas that are
based on scientific research. Application of Speech Processing and Design and
manufacture of electronic components using highly sophisticated processors
including basic and application-oriented research and the detailed product
management i.e. production and sales. Specific fields of competence, knowledge
and experiences are (1) speech-oriented feature extraction and pattern
classification of audio signals, (2) robust speech recognition for command
control, including keyword spotting and continuos phrase recognition.
STI
is a Spanish software company created in 1981. STI was born as a company
specialised in database development. Company products have been largely
commercialised in Spain, but also in Europe, Africa, and Latin-America. Today
STI is acting as a system developer and integrator and information technology
consultant. STI's skills include: databases development; declarative, functional
and object oriented languages; applications based on client-server architecture;
multimedia information retrieval; intelligent user interfaces; low-level
programming and Web based applications. The company has a large number of
installations both in the private and the public sectors, in Spain and abroad.
VIC
is an independent R&D enterprise in the field of smart algorithmic solutions
for automatic speech recognition, sound and image processing, echo cancellation,
noise suppression, sound and speech generation as well as data management.
Conventional techniques for speech processing do not sufficiently take into
account the special properties of speech as a signal, specially adapted to human
production and perception abilities. The research strategy of VIC lies in the
development and optimization of speech solutions for the embedded sector with
very low resources consumption. VIC understand the relations between speech
production, physical properties of speech and robust speech perception as key
knowledge.
VMU
was established as University of Lithuania in 1922 and is the source of
Lithuanian culture. The university accords special attention to Humanities and
to Social Sciences. The research area is new information and knowledge based
technologies, which will impact industries and corporations. We are developing
systems, which allow determine needs of customers, popular products and also
sell products in virtual environment. We are developing new models and tools,
which will provide intelligent access to heterogeneous, distributed information,
enabling software products (agents) to mediate between user needs and the
information sources available. These new virtual engineering information
services allow research organizations and industries collaborate, exchange
information and knowledge.
Universidad
Politécnica de Madrid (UPM) is one of the oldest Technical Universities in
Spain as several of its Engineering Schools date as back as the late eighteenth
century. During its lifetime the UPM has been involved in more than 250 projects
in the following subprogrammes: ACTS, APAS, BIOTECH, BRITE/EURAM, ENVIRONMENT,
ESPRIT, FAIR, INCO, INTAS, JOULE/THERMIE, NUCLEAR FUSION, TELEMATICS, TMR, and
TRANSPORT, among others. The span of the different actions carried out describe
an important activity in the field of R+D and technology transfer to the
industry, ranging from big companies to SME’s. In particular, as directly
related with the current proposal the following actions could be cited: (1)
IMDEX: Integrated System for Multimedia Indexing, Matching & Retrieval
(ESPRIT), (2) IVORY: Integrated Voice Recognition sYstem (ESPRIT), (3) ISAEUS:
Speech training for deaf and hearing impaired people (TELEMATICS).
UCV
was set up in 1990 after changes in Romanian society, simultaneously with
setting up of the Faculty of Automation and Computers. Computer Engineering
Department had a rich international activity being involved in several Tempus,
Socrates, Leonardo da Vinci projects. The main areas of interest cover Digital
Design, Programming Languages, Data base Design, Operating Systems, Algorithms,
Microprocessors, Microcomputers & Microcontrollers, Input/Output Systems,
Artificial Intelligence, Computer Networks, Data Communications, Network
Management, Digital Testing etc.
SHU
is one of the largest in the UK specialising in a range of engineering
disciplines including Mechanical,
Enterprise and Manufacturing, Materials, Electronic and Electrical Engineering
and Control Engineering. The research activity is focused mainly in the areas of
the development and application of Digital Signal Processing and Artificial
Intelligence techniques to signal analysis, electronics, sensor array signal
analysis, and telecommunication engineering projects, and to Optical
Communications. The SHU is pursuing research on signal recognition utilising
signal processing and artificial neural network methods.
BUY
is leading faculty in former and present Yugoslavia in the field of signal and
speech processing. It has been involved in almost all theoretical and practical
projects from this field during last 50 years. Its scientists and professors
achieved remarkable results in the field of speech modeling, coding and
recognition. For instance, they developed first domestic channel, predictive and
CELP vocoders; first speech isolated spoken digit recognition system, first
system for recognition of Serbian phonemes etc. Former fellows of these two
institutions are leading researchers in many important scientific centers all
over the world.
Each
partner of RecoPhone is a high ranked specialist in his field, related to the
activity of RecoPhone. The established groups, worked in speech recognition in
frequency domain will be integrated in the RecoPhone for information and
cooperation. This way the limitations of the frequency domain method will be
demonstrated and the higher accuracy of the time domain method will be
established.
The
integration activities of RecoPhone will take place over 5 years. The
recognition will be achieved by using the following methods: (1) Neural networks
for feature extraction and pattern recognition. (2) Fuzzy-logic for pattern
recognition. (3) Evolutionary algorithms for classification. (4) Special
heuristic methods for solutions without algorithms. (5) Dynamic lists for
internal management of signal description in real-time. (6) Data Base systems
for external management of signal description. (7) Knowledge Management System
for knowledge acquisition.
The
improved quality of speech recognition will create new ways of working and new
types of business, and will provide solutions to major societal challenges such
as healthcare, environment, safety, mobility and employment, and have far
reaching implications on our everyday life.
The
development of language recognition technology and the opening of the market
with the support of the European Community by RecoPhone will create new jobs in
information and communication technologies in Europe, especially in the Newly
Associated States (NAS) and the new German countries. The development of this
new method of communication between human and computer requires the European
dimension. The RecoPhone will be
developed with multi-cultural input (including NAS). The project will lead to
new results in many fields even including improved understanding of the function
of hearing and new quality of ABIs.
In
the RecoPhone-project different applications which will prove the performance of
the method will be implemented. The potential market for this new method extends
from scientific methods of speech recognition to the development of specific
intelligent systems, optimised for human - computer interactions. Examples are:
(1) Real time recognition of spoken language from commands up to natural spoken
language, (2) high security systems for integrated speaker and keyword
recognition, (3) Integration in well known speech recognition systems (IBM,
Philips…W3C - VoiceXML), (4) Intelligent speech compression algorithms by
phoneme recognition and transmission (as SMS), (5) portable keyboard free
computer devices, served by natural speech.
Language
processing by computers is important for the EU because of
the large number of
different languages. RecoPhone is able to overcome the
communication barrier in providing new, robust
solutions for speech processing and information retrieval and - exchange
between Citizens of Europe.