Identifying and recognizing noticeable sounds from physical measurements and their effect on soundscape
Invited paper
Ghent University
Tuesday 2 june, 2015, 16:20 - 16:40
0.3 Copenhagen (49)
Abstract:
In contrast to the classical noise control, the soundscape approach analyzes
the person-environment interaction in more detail including positive as well
as negative effects. Environmental sound is often a by-product of the
environment and listening to it is rarely the purpose of being in a place.
Therefore, noticing and inhibition-of-return play an important role in the
theoretical model for people’s perception. The proposed model extends from
an initial physiological response to environmental sound over noticing,
identifying, and recognizing to appraisal within a context of personal
beliefs and expectations. Consequently, it attempts to encompass the whole
interaction of the person and the environment from sensory inputs to actions
related to the response on the environment.
During the recent years, environmental monitoring and sound monitoring as
its part have experienced a technology driven growth to which various
governing bodies have shown a significant interest. However, the challenge
now presents itself in the analysis of the acquired big data especially when
it comes to perception. Several aspects of the above mentioned theoretical
model for perception of environmental sound have been implemented in the
computational models for this purpose. The models are based on the
artificial neural network structure that mimics many of the low level neural
processes occurring in the human brain. However, the models do not attempt
to make a simulation of a complete brain, which is still well out of reach
even for the most advanced computer architectures. This contribution will
focus in particular on the object formation and attention processes in an
attempt to predict which sounds would be noticed by the user of a space and
how this will affect the soundscape. Examples from urban parks and
residential areas will be shown to illustrate how accurately the model based
on physical inputs solely can match the human response.