Upload
others
View
11
Download
0
Embed Size (px)
Citation preview
F R A U N H O F E R I N S T I T U T E F O R D I g I TA l M E D I A T E c H N O l O g y I D M T
1
SIP-Toolbox – ModelIng SPeeCH InTellIgIbIlITY
Speech plays an outstanding role in human
communication. The intelligibility of speech
signals depends very much on the acoustic
conditions, e.g. on background noise or
reverberation, the capacity of the hearing
system, or the transmission channel by
which speech is transmitted (such as a
telecommunication system). Several factors
influence the perceived speech quality,
in particular intelligibility, loudness or the
overall sound quality.
There are a number of applications
requiring a reliable prediction of speech
quality. For example, developers of public
address systems or designers of office areas
benefit from models predicting speech in-
telligibility since they can replace costly and
time-consuming subjective measurements.
Additionally, software engineers make use
of such models to distinguish between
different algorithms or parameters.
Our SIP-Toolbox (Speech Intelligibility
Prediction Toolbox) has been developed as
1 Many applications, e.g. the de-
velopment of public adress systems
require reliable models for the pre-
diction of speech intelligibility.
Fraunhofer Institute for
Digital Media Technology IDMT
Ehrenbergstr. 31
98693 Ilmenau
Project Group Hearing, Speech and
Audio Technology
Branch Lab Oldenburg
Haus des Hörens
Marie-Curie-Straße 2
26129 Oldenburg
Contact
Jan Rennies
Telefon +49 441 2172-433
www.idmt.fraunhofer.de
a solution for such applications.
Evaluation of Speech Quality
Currently, several models are available to
predict speech intelligibility, loudness or
quality. The problem, however, is that for
a given acoustic situation, the models may
predict different results or may even be
inapplicable. It is therefore up to the user to
decide which model to use for the desired
application context.
photo acknowledgement: istockphoto.com
1
General view of the Toolbox
1
2 Speech intelligibility depends
very much on specific situations.
The SIP-Toolbox provides quick and
easy modeling of speech intelligi-
bility for different situations.
SIP-Toolbox
The SIP-Toolbox developed in the
project group Hearing, Speech and Audio
Technology at the Fraunhofer IDMT offers
a quick and easy prediction of the main
factors affecting speech quality for different
situations. It can be used to easily compare
the different models and thereby assist the
user to select the most suitable models
for his or her application. The SIP-Toolbox
designed in MATLAB® offers versatile
utilities to import, process and represent
data. For users with no access to Matlab®,
we also offer a stand-alone version of the
SIP-Toolbox.
Technical Features
Speech intelligibility models
The SIP-Toolbox contains standardized
models of speech intelligibility, e.g.
the Articulation Index (AI), the Speech
Intelligibility Index (SII) and the Speech
Transmission Index (STI). Additionally,
extended models are integrated which
can also predict the influence of hearing
impairment and binaural hearing.
Technical and perceptual models
Further models included in the SIP-toolbox
range from simple energy-based measures
(e.g. signal-to-noise ratios, reverberation
time) to complex speech quality measures
based on auditory models (e.g. PEMO-Q).
Additionally, loudness (e.g. DIN 45631/A1)
and other psychoacoustic quantities can be
calculated.
Graphical user interface
For arbitrary speech and noise signals, all
models can be quickly compared in a con-
cise graphical interface. The user can easily
control all important parameters such as
level or signal-to-noise ratio. An option
for batch processing allows a convenient
processing of large amounts of data.
Representation of signals
The SIP-Toolbox offers a visual and
acoustical representation of the signals
providing for example a direct impression
of the effect of reverberation. Arbitrary
acoustic situations can be integrated into
the SIP-Toolbox by means of their impulse
responses, which allows simulations of
the influence of a transmission system or
reverberation on speech quality.
customization
Current research of the Fraunhofer
project group Hearing, Speech and Audio
Technology focuses on the development
of improved and more generally applicable
speech intelligibility models. These models
are developed to account for the effects of
binaural hearing, temporal fluctuations or
hearing impairment, which are not or only
insufficiently included in current standard
models. The new methods can be integra-
ted into the SIP-Toolbox in a modular way.
Furthermore, we offer to objectively evalua-
te your own signals using the models of the
SIP-Toolbox. If desired, we can also perform
subjective listening tests based on your
custom condition with normal-hearing and
hearing-impaired listeners in our specially
equipped measurement facilities.
Demo-Version
If you are interested in our SIP-Toolbox,
we kindly ask you to contact us via mail
or phone. We will provide you with a
free demo-version of the software. For
further information and pictures about the
SIP-Toolbox please visit our website http://
www.idmt.fraunhofer.de/de/research_to-
pics/sip_toolbox_eng.html.
2
Speech Intelligibility