We found a match
Your institution may have access to this item. Find your institution then sign in to continue.
- Title
ENVIRONMENTAL NOISE CLASSIFICATION USING THE GMM APPROACH.
- Authors
OLTEANU, Elena; SEGARCEANU, Svetlana; GAVAT, Inge
- Abstract
Environmental noise recognition is becoming doubtlessly an essential component of computer science and robotics as it simulates the important function of human hearing. It has applications in fields such as industry, urban or ecology monitoring solutions, allowing to identify flaws or threats in industrial processes or ecological aspects. We propose, and experiment several methods, based on the GMM approach, for sound recognition. The first approach applies the sheer GMM to several feature sets, including LPC and Mel-cepstral features, for closed-set identification. The second one is GMMUBM, applied on the same types of features, to carry out open-set sound identification. It is intended to consolidate the recognition process by classifying uninteresting sounds outside the significant classes. We considered three types of noise: chainsaw, vehicles, and typical forest noise but intend to extend our research to more categories and to design a system meant to detect environmental threats.
- Subjects
NOISE (Work environment); COMPUTER science; ROBOTICS; GAUSSIAN mixture models; INVESTIGATIONS
- Publication
Acta Electrotehnica, 2019, Vol 60, Issue 1/2, p183
- ISSN
1841-3323
- Publication type
Article