Fusion of speech techniques for automatic environmental sound recognition

摘要

Environmental sound recognition is currently an important field of computer science and robotics, and a useful component in content-based search, security or environmental protection, framed in a new discipline called Automatic Environment Sound Recognition. We will evaluate how suitable is speech specific technology to be applied to environmental sound recognition, and how the performance can be grown by extending the feature space or by fusion of several sound technologies results. The problem addresses an environmental issue: preventing deforestation by detecting specific acoustic signal such as chainsaw, in a forest environment. The experiments consider three classes of forest specific sounds and are meant to detect the chainsaw sounds.