Bird species recognition using support vector machines
EURASIP Journal on Applied Signal Processing
VoxNet: An Interactive, Rapidly-Deployable Acoustic Monitoring Platform
IPSN '08 Proceedings of the 7th international conference on Information processing in sensor networks
Design and evaluation of a hybrid sensor network for cane toad monitoring
ACM Transactions on Sensor Networks (TOSN)
Monitoring frog communities: an application of machine learning
IAAI'96 Proceedings of the eighth annual conference on Innovative applications of artificial intelligence
IEEE Transactions on Audio, Speech, and Language Processing
Camazotz: multimodal activity-based GPS sampling
Proceedings of the 12th international conference on Information processing in sensor networks
| Hi-index | 0.00 |
Large scale freshwater monitoring networks can passively capture sound for species detection or classification. The sheer volume of acoustic recordings in such systems requires in-network classification. Most of the recent work on bio-acoustic in-network classification targets narrowband or short-durations signals, which renders it unsuitable for classifying species that emit broadband short-duration signals. This paper proposes a method for broadband sound based classification for large scale aquatic monitoring networks. The method is based on the extraction of a small set of spectral and temporal features. We collect empirical fish sounds, using the case study of the spotted tilapia (Tilapia mariae) which is an invasive freshwater fish species in Australia, and extract spectral and temporal features with our method. We then evaluate the classification accuracy and precision of these features for detecting tilapia sounds against the performance of existing narrowband sound features. The results show that using logistic regression with our limited feature set yields the best performance. Surprisingly, performance slightly improves when we downsample the signal from 44.1 to 16 kHz, indicating that our method is well-suited for classification on embedded devices. We quantify the computational benefits of our approach for enabling broader long-term in-situ species tracking in underwater environments.