An annotated dataset of bioacoustic sensing and features of mosquitoes

Abstract

As vectors of malaria, dengue, zika, and yellow fever, mosquitoes are considered one of the more severe worldwide health hazards. Widespread surveillance of mosquitoes is essential for understanding their complex ecology and behaviour, and also for predicting and formulating effective control strategies against mosquito-borne diseases. One technique involves using bioacoustics to automatically identify different species from their wing-beat sounds during flight. In this dataset, we collect sounds of three species of mosquitoes: Aedes Aegypti, Culex Quinquefasciatus & Pipiens, and Culiseta. These species were collected and reproduced in the laboratory of the Natural History Museum of Funchal, in Portugal, by entomologists trained to recognize and classify mosquitoes. For collecting the samples, we used a microcontroller and a mobile phone. The dataset presents audio samples collected with different sampling rates, where 34 audio features characterize each sound file, making it is possible to observe how mosquito populations vary heterogeneously. This dataset provides the basis for feature extraction and classification of flapping-wing flight sounds that could be used to identify different species.

A Review of Automated Bioacoustics and General Acoustics Classification Research

“…We confined the scope of the survey to studies that used machine-learning algorithms for sound classification. It is worth noting that several studies also used image recognition techniques classify animals [ 31 , 32 , 33 , 34 , 35 , 36 , 37 , 38 ]. Those studies fell outside the scope of this review.…”

A Low-Cost Multi-Purpose IoT Sensor for Biologging and Soundscape Activities

“…For the user to replicate our experiment, we used the dataset present in [ 27 ]. We extracted the audio features from the audio files used in that work and applied the various machine learning techniques for the species Aedes Aegypti and Culex Quinquefasciatus .…”

“…This approach aims to improve the efficiency of real-time mosquito abundance modeling and predict high-risk areas of infestation and breeding. During 2019 and 2020, 10 prototype devices were deployed in Madeira Island with the default recording program [ 27 ]. A new deployment is planned in Madeira Island and also in Brasil.…”

HumBugDB: A Large-scale Acoustic Mosquito Dataset

“…Due to the directionality of the recording method, typical sample durations are encountered from "only a few hundredths of a second" [Chen et al, 2014] to approximately half a second [Fanioudakis et al, 2018]. The approach therefore does not capture the acoustical properties of mosquito sound in free flight which aid mosquito detection in purely acoustic approaches [Vasconcelos et al, 2020]. Furthermore, these datasets survey lab-grown mosquito colonies which do not capture the biodiversity of mosquitoes encountered in the wild [Huho et al, 2007, Hoffmann andRoss, 2018].…”

“…Acoustic approaches Vasconcelos et al [2020] motivated their release by stating that none of the published datasets include environmental noise, which is essential to fully characterise mosquitoes in real-world scenarios. The dataset consists of 300 ms snippets, amounting to 15 minutes of recordings.…”